Social Science Research Network Electronic Paper Collection at:

http://ssrn.com/abstract=762385

Public Law & Legal Theory Working Paper Series

This paper can be downloaded without charge from:

Georgetown University Law Center

Working Paper No. 762385

JOHN MIKHAIL

Investigating Intuitive Knowledge of the Prohibition ofIntentional Battery and the Principle of Double Effect

Aspects of the Theory of Moral Cognition:

Business, Economics and Regulatory PolicyResearch Paper No. 762385

1

© John Mikhail, 2002 All rights reserved

Aspects of the Theory of Moral Cognition: Investigating Intuitive Knowledge of the Prohibition of Intentional Battery

and the Principle of Double Effect

John Mikhail 1

Abstract: Where do our moral intuitions come from? Are they innate? Does the brain contain a module specialized for moral judgment? Questions like these have been asked in one form or another for centuries. In this paper, we take them up again, with the aim of clarifying them and developing a specific proposal for how they can be empirically investigated. The paper presents data from six trolley problem studies of over five hundred individuals, including one group of Chinese adults and one group of American children, which suggest that adults and children ages 8­12 rely on intuitive or unconscious knowledge of specific moral principles to determine the permissibility of actions that require harming one person in order to prevent harm to others. Significantly, the knowledge in question appears to be merely tacit: when asked to explain or justify their judgments, experimental subjects were consistently incapable of articulating the operative principles on which their judgments appear to have been based. We explain these findings with reference to an analogy to human linguistic competence. Just as normal persons are typically unaware of the principles guiding their linguistic intuitions, so too are they often unaware of the principles guiding their moral intuitions. These studies pave the way for future research by raising the possibility that specific poverty of the stimulus arguments can be formulated in the moral domain. Differences between our approach to moral cognition and those of Piaget (1932), Kohlberg (1981), and Greene et al. (2001) are also discussed.

1. Introduction

Where do our moral intuitions come from? Are they innate? Does the brain contain a

module specialized for moral judgment? Does the human genetic program contain instructions

for the acquisition of a sense of justice or moral sense? Questions like these have been asked in

one form or another for centuries. In this paper we take them up again, with the aim of clarifying

them and developing a specific proposal for how they can be empirically investigated.

2

In Section 1, we summarize our approach to the theory of moral cognition and explain

some basic elements of our theoretical framework. We also introduce examples of the

perceptual stimuli used in our research and discuss some of the properties of the moral intuitions

they elicit. In Sections 2­7, we present the results of six trolley problem studies designed to

investigate the moral competence of adults and of children ages 8­12; in particular, their intuitive

or unconscious knowledge of the prohibition of intentional battery and the principle of double

effect. In Section 8, we provide a general discussion of our findings and contrast our approach to

moral cognition with those of Piaget (1932/1965), Kohlberg (1981), and Greene, Sommerville,

Nystrom, Darley & Cohen (2001). Section 9 is an Appendix containing both the stimulus

materials used in our experiments and our subjects’ responses to them.

1.1 Theoretical Framework

Like many theorists, we begin from the assumption that the theory of moral cognition

may be usefully modeled on aspects of the theory of linguistic competence (see, e.g., Chomsky,

1978; Cosmides & Tooby, 1994; Dwyer, 1999; Goldman, 1993; Harman, 2000; Mahlmann,

1999; Mikhail, 2000; Mikhail, Sorrentino & Spelke, 1998; Rawls, 1971; Stich, 1993). Our

research is thus organized, in the first instance, around three questions, close analogues of the

fundamental questions in Chomsky’s (1986) framework for the investigation of human language.

(1) (a) What constitutes moral knowledge? (b) How is moral knowledge acquired? (c) How is moral knowledge put to use?

A brief overview of some of the concepts and terminology we use to clarify these

questions may be helpful. In our framework, the answer to (1a) is given by a particular moral

grammar or theory of moral competence: a theory of the mind/brain of a person who possesses a

system of moral knowledge, or what might be referred to informally as a “moral faculty,” “moral

3

sense” or “conscience.” The answer to (1b) is given by Universal Moral Grammar (UMG): a

theory of the initial state of the moral faculty—which, in keeping with conventional assumptions

of modularity (see, e.g., Fodor, 1983; Gazzaniga, 1992; Gazzaniga, Ivry & Magnum, 1998;

Pinker, 1997), we provisionally assume to be a distinct subsystem of the mind/brain—along with

an account of how the properties UMG postulates interact with experience to yield a mature

system of moral knowledge. The answer to (1c) is given by a theory of moral performance: a

theory of how moral knowledge enters into the actual representation and evaluation of human

acts and institutional arrangements, as well as other forms of actual conduct (see, e.g., Dwyer,

1999; Mikhail, 2000; compare Rawls, 1971; Nozick, 1968).

Following Chomsky (1965), we use the terms “observational adequacy,” “descriptive

adequacy” and “explanatory adequacy” to refer to increasing levels of empirical success a theory

of moral cognition might achieve. A moral theory is observationally adequate with respect to a

given set of moral judgments to the extent that it provides a correct description of those judgments

in some manner or other, for example, by listing them or by explicitly stating a set of principles

from which they can be derived. A moral theory is descriptively adequate with respect to the

mature individual’s moral competence to the extent that it correctly describes that system, in other

words, to the extent it provides a correct answer to (1a). Finally, a moral theory meets the

condition of explanatory adequacy to the extent it correctly describes the initial state of the moral

faculty and correctly explains how the properties of the initial state it postulates interact with

experience to yield a mature system of moral competence; in other words, to the extent that it

provides a correct answer to (1b) (Mikhail, 2000). 2

Unlike Kohlberg (1981), we distinguish sharply between an individual’s operativemoral

principles (those principles actually operative in her exercise of moral judgment) and her express

4

principles (those statements she makes in the attempt to describe, explain, or justify her

judgments). We make no assumption that the normal individual is aware of the operative

principles which constitute her moral knowledge, or that she can become aware of them through

introspection, or that her statements about them are necessarily accurate. On the contrary, we

hypothesize that just as normal persons are typically unaware of the principles guiding their

linguistic or visual intuitions, so too are they often unaware of the principles guiding their moral

intuitions. In any event, the important point is that, as with language or vision, the theory of moral

cognition must attempt to specify what the properties of moral competence actually are, not what a

person may report about them (Haidt, 2001; Mikhail, 2000; Mikhail, Sorrentino & Spelke, 1998).

Finally, we follow Chomsky (1995), Lewontin (1990), Marr (1982), and other

commentators in assuming that the problems of descriptive and explanatory adequacy possess a

certain logical and methodological priority over more complicated inquiries into the neurological

and evolutionary foundations of moral cognition and behavior. Hence we carefully distinguish

(1a)­(1c) from two further questions a complete theory of moral cognition must answer:

(1) (d) How is moral knowledge physically realized in the brain? (e) How did moral knowledge evolve in the species?

Although many researchers have addressed questions like these, their efforts seem at this

juncture to be somewhat premature. Just as our ability to ask well­focused questions about the

evolution and physical bases of language depends on solving the problems of descriptive and

explanatory adequacy in the linguistic domain (Chomsky, 1995; Hauser, Chomsky & Fitch,

2002), so too is our understanding of (1d) and (1e) advanced by achieving reasonably correct

solutions to questions like (1a) and (1b) in the moral domain. Put simply, we cannot profitably

ask how moral knowledge evolved in the species or where it resides in the brain until what

constitutes moral knowledge and how it is acquired are better understood.

5

1.2 Perceptual Stimuli and Perceptual Model

Research in the Piagetian tradition has attempted to answer questions like (1a) and (1b)

by investigating the developing child’s mental representations of the “subjective” and

“objective” elements of moral judgment, the former consisting of the goals and intentions of an

action, the latter consisting of an action’s effects and material consequences. In Piaget’s

(1932/1965) original studies, children were found to base their moral judgments on mental

representations of effects, not intentions, until around age nine. More recently, many

investigators have suggested that these findings were an artifact of the methods and assessment

procedures Piaget employed. Some researchers (e.g., Baird, 2001; Berndt & Berndt, 1975;

Costanzo, Coie, Grumet & Farnhill, 1973; Lilliard & Flavell, 1990; Nelson, 1980) have

discovered that children as young as three use information about motives and intentions when

making moral judgments, if that information is made explicit and salient. Moreover, a

considerable body of research on infant cognition (e.g., Gergely, Nadasdy, Csibra & Biro, 1995;

Johnson, 2000; Meltzoff, 1995; Woodward, Sommerville & Guajardo, 2001) suggests that even

young infants are predisposed to interpret the actions of animate agents in terms of their goals

and intentions.

Our research seeks to build on these prior studies by investigating how experimental

subjects reconstruct and utilize information about intentions and effects when evaluating

“morally complex acts” – that is, acts and omissions which are comprised of multiple intentions

and which generate both good and bad effects. To illustrate, consider the following examples of

the so­called “trolley problem” and related thought experiments invented by Foot (1967) and

Thomson (1985).

6

The Trolley Problem Charlie is driving a train when the brakes fail. Ahead five people are working on the track with their backs turned. Fortunately, Charlie can switch to a side track, if he acts at once. Unfortunately, there is also someone on that track with his back turned. If Charlie switches his train to the side track, he will kill one person. If Charlie does not switch his train, he will kill five people.

Is it morally permissible for Charlie to switch his train to the side track?

The Transplant Problem Dr. Brown has five patients in the hospital who are dying. Each patient needs a new organ in order to survive. One patient needs a new heart. Two patients need a new kidney. And two more patients need a new lung. Dr. Brown can save all five patients if he takes a single healthy person and removes her heart, kidneys, and lungs to give to these five patients. Just such a healthy person is in Room 306. She is in the hospital for routine tests. Having seen her test results, Dr. Brown knows that she is perfectly healthy and of the right tissue compatibility. If Dr. Brown cuts up the person in Room 306 and gives her organs to the other five patients, he will save the other five patients, but kill the person in Room 306 in the process. If Dr. Brown does not cut up the person in Room 306, the other five patients will die.

Is it morally permissible for Dr. Brown to cut up the person in Room 306?

The Bystander Problem Edward is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Edward sees what has happened: the train driver saw five workmen men ahead on the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the five men; the banks are so steep that they will not be able to get off the track in time. Fortunately, Edward is standing next to a switch, which he can throw, that will turn the train onto a side­track. Unfortunately, there is one person standing on the side­track, with his back turned. Edward can throw the switch, killing the one; or he can refrain from doing this, letting the five die.

Is it morally permissible for Edward to throw the switch?

The Footbridge Problem Nancy is taking her daily walk near the train tracks when she notices that the train that is approaching is out of control. Five men are walking across the tracks. The train is moving so fast that they will not be able to get off the track in time. Nancy is standing next to a man, whom she can throw in front of the train, thereby preventing it from killing the men. Nancy can throw the man, killing him but saving the five men; or she can refrain from doing this, letting the five die.

Is it morally permissible for Nancy to throw the man?

7

As we discuss below, when experimental subjects were presented with these scenarios,

they judged Charlie’s turning the train in The Trolley Problem to be permissible, Dr. Brown’s

cutting up the patient in the Transplant Problem to be impermissible, Edward’s throwing the

switch in the Bystander Problem to be permissible, and Nancy’s throwing the man in the

Footbridge Problem to be impermissible (Table 1). These responses confront us with a

potentially surprising contrast between the Trolley and Bystander Problems, on the one hand,

and the Transplant and Footbridge Problems, on the other. In the former problems, saving five

people at the cost of killing one person is thought to be permissible. In the latter problems, by

contrast, saving five at the cost of killing one is held to be impermissible.

Table 1: Moral Intuitions of Trolley, Transplant, Bystander, and Footbridge Problems Problem Action Good Effect Bad Effect Deontic

Status Trolley Charlie’s turning the train Preventing 5 deaths 1 Death Permissible Transplant Dr. Brown’s cutting up the patient Preventing 5 deaths 1 Death Impermissible Bystander Edward’s throwing the switch Preventing 5 deaths 1 Death Permissible Footbridge Nancy’s throwing the man Preventing 5 deaths 1 Death Impermissible

These facts lead us to speculate about the cognitive mechanisms the mind employs in

responding to these four scenarios. In the first instance, they lead us to ask the following

question: what are the operative principles of moral competence that are responsible for these

divergent responses? The problem is more difficult than it may seem at first. On the one hand,

comparatively simple deontological and consequentialist moral principles (e.g., “If an act causes

death, then it is wrong,” “If the consequences of an act are better than the consequences of any of

available alternative, then it is required,” etc.) are incapable of explaining the pattern of

intuitions elicited by these problems. For example, a simple deontological principle forbidding

8

all killing would generate the intuition that Charlie’s switching tracks in the Trolley Problem and

Edward’s switching tracks in the Bystander Problem are impermissible. But these actions are

judged to be permissible. Likewise, a simple utilitarian principle requiring agents to perform

actions with the best foreseeable consequences would presumably generate the intuition that Dr.

Brown’s cutting up the patient in the Transplant Problem and Nancy’s throwing the man in the

Footbridge Problem are obligatory, or at least permissible; yet these actions are judged to be

impermissible.

On the other hand, conditional principles whose antecedents simply restate those action­

descriptions found in the stimulus (e.g., “If an act is of the type ‘throwing the switch,’ then it is

permissible”; “If an act is of the type ‘throwing the man’ then it is impermissible,”) are also

descriptively inadequate. This is because they lead us to make inaccurate predictions of how

these action­descriptions will be evaluated when they are embedded in materially different

circumstances. For example, as we discuss below, when the costs and benefits in the Bystander

Problem are manipulated, so that an action described as “throwing the switch” will save $5

million of equipment at the cost of killing one person, individuals judge the action so described

to be impermissible. Likewise, when the circumstances of the Footbridge Problem are modified

so that the action described as “throwing the man” is presumed to involve consensual touching,

subjects judge the action to be permissible. In general, it is easy to show that the action­

descriptions used in these problems are “morally neutral” (Baird, 2001; Nelson, 1980), in the

sense that the permissibility judgments they elicit are circumstance­dependent.

Since the circumstances of an action can vary along an indefinite number of dimensions

(e.g., D’Arcy, 1963; Donagan, 1977; Lyons, 1965; Stone, 1964), the conclusion to which we

quickly are led by considerations like these is that any attempt to explain the moral intuitions

9

elicited by these examples by means of a simple stimulus­response model is doomed at the start.

Although each of these moral intuitions is occasioned by an identifiable stimulus, how the mind

goes about interpreting these hypothetical fact patterns, and separating the actions they depict

into those that are permissible and those that are not, is not something revealed in any obvious

way by the surface properties of the stimulus itself. Instead, an intervening step between

stimulus and response must be postulated: a pattern of organization of some sort that is imposed

on the stimulus by the mind itself. Hence a simple perceptual model such as the one in Figure 1

is inadequate for explaining these moral intuitions. Instead, as is the case with language

perception (Chomsky, 1964), an adequate perceptual model must, at a minimum, look more like

the one in Figure 2.

Fig. 1: Simple Perceptual Model for Moral Judgment

INPUT ?

PERMISSIBLE

IMPERMISSIBLE

10

? PERMISSIBLE

IMPERMISSIBLE

STRUCTURAL DESCRIPTION

Perceptual Response: Moral Judgment

Unconscious Mental Representation

INPUT ?

Conversion Rules

Stimulus: Fact Pattern

Deontic Rules

Fig. 2: Expanded Perceptual Model for Moral Judgment

The expanded perceptual model in Figure 2 implies that, like grammaticality judgments,

permissibility judgments do not necessarily depend on the surface properties of an action­

description, but on more fundamental properties of how that action is mentally represented. Put

differently, it suggests that the problem of descriptive adequacy in the theory of moral cognition

may be divided into at least two parts: (a) the problem of determining the nature of the

computational principles (i.e., “deontic rules”) operative in the exercise of moral judgment, and

11

(b) the problem of determining the representational structures (i.e., “structural descriptions”)

over which those computational operations are defined.

What are the properties of these intervening mental representations? In our view, it

seems reasonable to suppose that morally cognizable fact patterns are mentally represented in

terms of abstract categories like act, consequence, and circumstance; agency, motive, and

intention; proximate and remote causes; and other familiar concepts that are the stock in trade of

philosophers, lawyers, and jurists (Mikhail, 2000; see also Donagan, 1977; Sidgwick, 1907). But

which specific concepts does the system of moral cognition in fact use? In what manner, i.e.,

according to what principles or rules, does it use them? Answers to questions like these, if

available, would begin to solve the problem of descriptive adequacy.

1.3 Our Hypothesis

Our hypothesis is that the moral intuitions generated by the Trolley, Transplant, Bystander,

and Footbridge problems and structurally similar thought experiments (henceforth, “trolley

problems”) can be best explained by postulating intuitive knowledge of specific moral principles,

including the prohibition of intentional battery and the principle of double effect. The former is

a familiar principle of both common morality and the common law proscribing acts of

unpermitted, unprivileged bodily contact, that is, of touching without consent (Prosser, 1941;

Shapo, 2003). The latter is a complex principle of justification, narrower in scope than the

traditional necessity or “choice of evils” defense, which in its standard formulation holds that an

otherwise prohibited action may be permissible if the act itself is not wrong, the good but not the

bad effects are intended, the good effects outweigh the bad effects, and no morally preferable

alternative is available (Mikhail, 2000; see also Fischer & Ravizza, 1992). Both of these

12

principles require clarification, but taken together and suitably elaborated they can be invoked to

explain the relevant pattern of intuitions in a relatively simple and straightforward manner. The

key structural difference between the two sets of examples is that, in Transplant and Footbridge

problems, the agent commits a series of distinct trespasses prior to and as a means of achieving

his good end, whereas in the Trolley and Bystander problems, these violations are subsequent

and foreseen side effects. Figures 3 and 4 illustrate this difference in the case of the Footbridge

and Bystander problems.

D’s throwing the man at t (0)

D’s committing battery at t (0)

D’s preventing the train from killing the men at t (+n+o)

D’s killing the man at t (+n+p)

D’s causing the train to hit the man at t (+n)

D’s committing battery at t (+n)

Fig. 3: Mental Representation of Footbridge Problem

Side Effects

End

Means

D’s committing homicide at t (+n+p)

13

Our computational hypothesis holds that when people encounter the Footbridge and

Bystander problems, they spontaneously compute unconscious representations like those in

Figures 3 and 4. 3 Note that in addition to explaining the relevant intuitions, this hypothesis has

further testable implications. For example, we can investigate the structural properties of the

underlying representations by asking subjects to evaluate certain probative descriptions of the

relevant actions. Descriptions using the word “by” to connect individual nodes of the tree in the

downward direction (e.g., “D turned the train by throwing the switch,” “D killed the man by

turning the train”) will be deemed acceptable; by contrast, causal reversals using “by” to connect

nodes in the upward direction (“D threw the switch by turning the train,” “D turned the train by

killing the man”) will be deemed unacceptable. Likewise, descriptions using the phrase “in order

to” to connect nodes in the upward direction along the vertical chain of means and ends (“D

D’s throwing the switch at t (0)

D’s turning the train at t (+n)

D’s preventing the train from killing the men at t (+n) D’s causing the train

to hit the man at t (+n+o)

D’s committing battery at t (+n+o) End

Fig. 4: Mental Representation of Bystander Problem

Side Effects

D’s killing the man at t (+n+o+p)

D’s committing homicide at t (+n+o+p)

Means

14

threw the switch in order to turn the train”) will be deemed acceptable. By contrast, descriptions

of this type linking means with side effects (“D threw the switch in order to kill the man”) will

be deemed unacceptable. In short, there is an implicit geometry to these representations, which

an adequate theory can and must account for.

Our hypothesis is interesting and controversial for several reasons. First, while many

theorists have suggested that the principle of double effect may be part of a descriptively

adequate theory of trolley intuitions (e.g., Harman, 1977, 2000), and of human morality

generally (e.g., Nagel, 1986; Quinn, 1993), no prior experimental studies have directly tested this

assumption. The experiments by Petrinovich and his colleagues (Petrinovich & O’Neill, 1996;

Petrinovich, O’Neill & Jorgensen, 1993), which utilize trolley problems, do not adequately

clarify this issue, because of their focus on behavioral predictions (e.g., asking participants to

answer the question “What would you do?”) rather than on deontic judgments per se (e.g., asking

participants to answer the question “Is X morally permissible?”). Likewise, Greene et al. (2001),

who also use trolley problems as probes, also appear to leave this issue unresolved (see §8.2.3).

Second, our hypothesis is significant because, if it is true, it implies that the mental

operations involved in the exercise of moral judgment are more complex than is commonly

thought. For the principle of double effect, for example, to be operative in its standard

formulation, adults and children must possess a list of intrinsically wrong acts, a set of rules for

generating morally cognizable act­representations, and a calculus of some sort for computing—

and comparing the probabilities of—an action’s good and bad effects. They must also have the

cognitive resources to distinguish the “act itself” from its effects and further consequences, to

distinguish the act’s “foreseen effects” from its “intended effects,” and, more generally, to

differentiate the act’s causal and intentional properties from those of its alternatives. Further,

15

they must compute act­representations in terms of properties like ends, means, and side effects,

even though the stimulus contains no direct evidence of these properties. In short, our

hypothesis implies that “ordinary” people—not just trained lawyers or philosophers—possess a

complex sense of justice that incorporates subtle elements of a fully articulated legal code,

including abstract theories of causation and intention.

Finally, our hypothesis raises interesting and novel questions for the theory of moral

development. Specifically, it leads us to ask whether children are explicitly taught the principle

of double effect, and if not, whether the principle or some variant of it is in some sense innate.

As Harman (2000) explains, this question naturally arises as soon as one settles on an

explanation of the structure of our moral intuitions that makes reference to this principle. “An

ordinary person was never taught the principle of double effect,” Harman observes, and “it is

unclear how such a principle might have been acquired by the examples available to the ordinary

person. This suggests that [it] is built into . . . morality ahead of time” (Harman, 2000, p. 225).

Similar reasoning may be thought to apply to the prohibition of intentional battery, at least as

that prohibition is defined and utilized here. 4 On reflection, it seems doubtful that children are

affirmatively taught to generate the specific representations presupposed by this principle to any

significant extent. We thus seem faced with the possibility that certain moral principles emerge

and become operative in the exercise of moral judgment that are neither explicitly taught, nor

derivable in any obvious way from the data of sensory experience. In short, we appear

confronted with an example of what Chomsky calls the phenomenon of the “poverty of the

stimulus” in the moral domain (Dwyer, 1999; Mikhail, 2000; compare Chomsky, 1986).

16

CHILD’S LINGUISTIC

DATA

LINGUISTIC GRAMMAR

CHILD’S MORAL DATA

MORAL GRAMMAR

English Japanese Zapotec Malagasy Arabic …… ……

How much diversity?

UG

UMG

?

?

Figure 5: Acquisition Models for Language and Morality

The “argument from the poverty of the moral stimulus” (Mikhail, 2000) can be depicted

graphically by means of an acquisition model similar to the one Chomsky (1964) initially proposed

in the case of language (Figure 5). In the linguistic version of this model, Universal Grammar

(UG) “may be regarded as a theory of innate mechanisms, an underlying biological matrix that

provides a framework within which the growth of language proceeds,” and proposed principles of

UG “may be regarded as an abstract partial specification of the genetic program that enables the

child to interpret certain events as linguistic experience and to construct a system of rules and

principles on the basis of that experience” (Chomsky, 1980, p. 187). Likewise, in the case of

17

moral development, Universal Moral Grammar (UMG) may be regarded as a theory of innate

mechanisms that provides the basic framework in which the development of moral competence

unfolds, and specific principles of UMG may be regarded as a partial characterization the innate

function that maps the developing child’s relevant moral experience (her “moral data”) into the

mature state of her acquired moral competence (i.e., her “moral grammar”).

The linguistic grammars children acquire are hopelessly underdetermined by the data

available to them as language learners; linguists therefore postulate a significant amount of innate

knowledge to fill this gap (e.g., Baker, 2001; Pinker, 1994). Further, because every normal human

child can and will learn any of the world’s natural languages simply by being placed in an

appropriate environment, UG must be rich and specific enough to get the child over the learning

hump, but not so specific as to preclude her ability to acquire every human language (Chomsky,

1986). Turning to UMG, it is unclear whether a similar situation and a similar tension between

descriptive and explanatory adequacy obtains. Nevertheless, the acquisition model we have

sketched, though abstract, can be made more concrete by considering the specific example of

trolley intuitions. If a computational moral grammar does in fact enter into the best explanation of

these intuitions, then two further questions arise within the framework of this model: First, what

are the properties of the moral grammars that people do in fact acquire, and how diverse are they?

Second, what informational gaps, if any, can be detected between the inputs and outputs of the

model? That is, what if any principles of moral grammar are acquired for which the environment

contains little or no evidence? According to the argument from the poverty of the moral stimulus,

if specific principles emerge and become operative in the course of normal moral development,

but the acquisition of these principles cannot be explained on the basis of the child’s moral data,

then the best explanation of how children acquire these principles may be that they are innate, in

18

Chomsky’s dispositional sense (Chomsky, 1986; see also Baker, 2001; Dwyer, 1999; Mikhail

2002; Pinker, 1994; Spelke, 1998).

2. Experiment 1

Having introduced some elements of our theoretical framework, we turn directly to a

discussion and analysis of our experimental findings. At the outset of our investigations, we

were interested in a variety of questions that might be asked about thought experiments like the

trolley problems and the moral intuitions they elicit, including the following: First, are these

intuitions widely shared? Are they shared across familiar demographic categories like gender,

race, nationality, age, culture, religion, or level of formal education? Second, what are the

operative principles? How precisely can we characterize the relevant mental operations and to

what extent are they open to conscious introspection? Third, how are the operative principles

learned or acquired? What might examples like these eventually tell us about moral

development and the acquisition of the moral sense?

Our first study attempted to address only a subset of these questions, including (1)

whether and to what extent these intuitions are widely shared; (2) what are the operative

principles; and (3) whether the operative principles are open to conscious introspection.

2.1 Method

2.1.1 Participants

Participants were 40 adult volunteers from the M.I.T. community between the ages of 18­

35. The group consisted of 19 women and 21 men.

19

2.1.2 Stimuli and Procedure

Eight scenarios were used, all of which were adapted from Foot (1967), Thomson (1986),

and Harman (1977) (see §9 for the complete text of these scenarios; see also Mikhail, 2000). In

all eight scenarios, an agent must choose whether to perform an action that will result in one

person being killed and five other persons, who would otherwise die, being saved.

The scenarios were divided according to our hypothesis into two groups. Four scenarios,

which were modeled on the Transplant and Footbridge Problems, described a choice between (a)

committing an intentional battery in order to prevent five other people from dying, knowing that

the battery will also constitute a foreseeable but non­intentional homicide, and (b) refraining

from doing so, thereby letting the five die. Four other scenarios, which were modeled on the

Trolley and Bystander Problems, described a choice between (a) doing something in order to

prevent five people from dying, knowing that the action will constitute a foreseeable but non­

intentional battery and a foreseeable but non­intentional homicide, and (b) refraining from doing

so, thereby letting the five die.

The morally salient difference between the two sets of cases, in other words, concerned

the type of battery embedded in the agent’s action plan. In the first group of scenarios, the

battery was intentional, embedded within the agent’s action plan as a means (henceforth

“Intentional Battery”). In the second group, the battery was foreseeable (but not intentional),

embedded within the agent’s action plan as a side effect (henceforth “Foreseeable Battery”).

Each participant received a written questionnaire containing one scenario. The

participant was first instructed to read the scenario and to judge whether or not the proposed

action it described was “morally permissible.” The participant was then asked on a separate page

of the questionnaire to provide reasons explaining or justifying his or her response. Twenty

20

participants were given an Intentional Battery scenario. The other twenty participants were

given a Foreseeable Battery scenario. The assignment of participants to scenario type was

random.

2.2 Results

2.2.1 Judgments

The main results of Experiment 1 are presented in Figure 6. 2 of 20 participants in the

Intentional Battery condition judged the action constituting intentional battery to be permissible.

By contrast, 19 of 20 participants in the Foreseeable Battery condition scenario judged the action

constituting foreseeable battery to be permissible. This difference is significant: x 2 (1, N=40) =

28.96, p < .001, suggesting that the scenarios evoke different action representations whose

properties are morally salient. 5

Male and female responses in Experiment 1 are presented in Figure 7. Of the 10 men

given an Intentional Battery scenario, 2 judged the action constituting intentional battery to be

permissible and 8 judged it to be impermissible. Of the 10 women given an Intentional Battery

scenario, all 10 judged the action constituting intentional battery to be impermissible.

Meanwhile, all 11 of the men and 8 of the 9 women who were given a Foreseeable Battery

scenario judged the action constituting foreseeable battery to be permissible. These differences

are also significant, x 2 (1, N=19) = 15.44, p < .001 (women) and x 2 (1, N=21) = 14.6, p < .001

(men), suggesting that there are no significant gender differences in the way the two types of

scenario are mentally represented and morally evaluated.

21

0 2 4 6 8 10 12 14 16 18 20

Intentional Battery

Foreseeable Battery

Permiss. Impermiss.

Act Type

Subjects

Figure 6: Moral Judgments of Two Act Types in Experiment 1 (Intentional Battery vs. Foreseeable Battery)

X 2 (1, N=40) = 29.0, p < .001

0

2

4

6

8

10

12

Men Women Men Women

Permiss. Impermiss.

Intentional Battery

Subjects

Figure 7: Judgments of Act Types in Experiment 1 by Gender (Intentional Battery vs. Foreseeable Battery)

Foreseeable Battery

X 2 (1, N=19) = 15.44, p < .001 (women).

X 2 (1, N=21) = 14.6, p < .001 (men).

22

2.2.2 Justifications

Subjects’ expressed principles—the responses they provided to justify or explain their

judgments—were also coded and analyzed. Three categories of increasing adequacy were used

to classify these responses: (1) no justification, (2) logically inadequate justification, and (3)

logically adequate justification. Responses that were left completely blank were categorized

under the heading of “no justification.” Responses that were not blank but which failed to state a

reason, rule, or principle—or to identify any feature whatsoever of the given scenario—that

could in principle generate the corresponding judgment were classified as logically inadequate

justifications. Finally, responses that did state a reason, rule, or principle, or did otherwise

identify at least one feature of the given scenario—even one that was obviously immaterial,

irrelevant, arbitrary, or ad hoc—that could in principle generate the corresponding judgment

were classified as logically adequate justifications.

Utilizing this taxonomy, two researchers independently coded a subset of justifications

and achieved an inter­observer reliability of 89% (n=36). One researcher then coded the

complete set of justifications collected in Experiment 1. 32.5% (13/40) of participants gave no

justification, 17% (7/40) provided logically inadequate justifications, while only 50% (20/40)

provided logically adequate justifications. Furthermore, many of the logically adequate

justifications consisted of simple deontological or consequentialist principles that were evidently

incapable of generating the conflicting pattern of intuitions in Experiment 1. These justifications

thus failed the test of observational adequacy in the sense defined in §1.1. These findings

together with the data on expressed justifications gathered in our remaining studies are discussed

again in §8.

23

2.3 Discussion

Experiment 1 was designed to achieve several different objectives. First, it was meant to

investigate a set of untested empirical claims implicit in the philosophical and legal literature

about how the trolley problems are mentally represented and morally evaluated. In their

accounts of trolley problems, philosophers and legal theorists often take for granted the deontic

status readers will assign to a given action sequence (e.g., Fischer & Ravizza, 1992; Katz, 1987;

Thomson, 1985). Prior to our studies, however, no controlled experiments had directly tested

these assumptions or attempted to extend them to broader populations. Instead, prior

experimental research using trolley problems as probes (Petrinovich & O’Neill, 1996;

Petrinovich et al., 1993) had left these issues largely unresolved. As we predicted, conventional

assumptions about the deontic intuitions elicited by these problems were confirmed, and the

intuitions themselves were widely shared.

Second, Experiment 1 was designed to investigate whether the participants in our

experiments could, when asked, provide coherent and well­articulated justifications for their

judgments about individual trolley problems. Based on informal observation, as well as theory­

dependent considerations arising from the linguistic analogy—in particular, the inaccessible

status of principles of grammar—we predicted that many or most of our subjects be incapable of

doing so. This prediction also held: even under an extremely liberal coding scheme, according to

which a justification was deemed logically adequate if it picked out at least one distinguishing

feature of the given scenario, even one that was obviously immaterial, irrelevant, arbitrary, or ad

hoc, that could in principle “serve as part of the premises of an argument that arrives at the

matching judgments” (Rawls, 1971, p. 46), only 50% of the participants in our study provided

logically adequate justifications for their judgments. Additionally, as indicated, many of these

24

justifications were inadequate to account for the pattern of intuitions generated in Experiment 1

and thus failed the test of observational adequacy in the sense defined in §1.1. This suggested

that a within­subject design would elicit considerable fewer logically adequate justifications than

a between­subject design, because in the former condition subjects would be required to

reconcile and explain two contrary intuitions by means of an overarching rationale or principle.

On this basis, we decided to utilize a within­subject design in Experiment 2 (see §3).

A further objective of Experiment 1 was to investigate our hypothesis that the moral

intuitions generated by the trolley problems could be explained by postulating intuitive

knowledge of the prohibition of intentional battery and the principle of double effect. As

interpreted here, the combined effect of mechanically applying these principles to these scenarios

would be to permit throwing the switch and turning the train in the Trolley and Bystander

conditions but to prohibit cutting up the patient and throwing the man in the Transplant and

Footbridge conditions. This is how participants did, in fact, respond in these conditions, thus

confirming to a limited extent our hypothesis about operative principles.

Finally, Experiment 1 was also meant to begin the process of investigating the potential

universality of a certain class of moral intuitions, such as those elicited by the trolley problems,

by determining whether one sample of adult men and women would share intuitive responses to

these problems. Again, based upon informal observation, as well as various theory­dependent

considerations (Mikhail, 2000), we predicted that there would be no statistically significant

gender differences. This prediction also held—a finding that is at least potentially in conflict

with the claims of Gilligan (1982) and others that men and women typically differ in how they

evaluate moral problems.

25

In sum, the findings of Experiment 1 constitute evidence that one component of moral

knowledge, deontic knowledge, consists of a system of rules or principles (a “moral grammar”)

capable of generating and relating mental representations of various elements of an action plan

(Mikhail et al., 1998). Our findings also constitute evidence that the moral grammar contains

principles capable of distinguishing intentional battery (battery embedded within an agent’s

action plan as a means) and foreseeable battery (battery embedded within an agent’s action plan

as a side effect), as well as a further principle, such as the principle of double effect or some

comparably complex ordering principle (Donagan, 1977), prohibiting intentional battery but

permitting foreseeable battery in the context of cases of necessity such as the trolley problems.

Because subjects displayed only a limited ability to provide adequate justifications of their

intuitions, Experiment 1 also implies that, as is the case with linguistic intuitions, the principles

generating moral intuitions are at least partly inaccessible to conscious introspection. Finally,

our findings also suggest that at least some moral intuitions are widely shared, irrespective of

gender.

3. Experiment 2

In Experiment 1, we discovered an apparent difference between the way intentional

battery and foreseeable battery are mentally represented and morally evaluated, at least in the

context of cases of necessity such as the trolley problems. We also discovered that the moral

competence of both men and women appears to consist, at least in part, of intuitive or

unconscious knowledge of the prohibition of intentional battery and the principle of double

effect. Experiment 2 was designed to bring additional evidence to bear on these hypotheses, in

three different ways.

26

The first way was to investigate the concept of battery that was used in our analysis of

Experiment 1. In Experiment 1, we drew on established legal doctrine in assuming that battery

could in effect be defined as “unpermitted or unprivileged contact with a person,” that is, as

contact without consent (Prosser, 1941; Shapo, 2003). Moreover, we followed the traditional

law of tort in assuming that the notion of unprivileged contact “extends to any part of the body,

or to anything which is attached to it” and includes any touching of one person by another or by

“any substance put in motion by him” (Hilliard, 1859). In Experiment 2, we investigated this

concept of battery by modifying one of the Intentional Battery scenarios used in Experiment 1,

so that an action described as “throwing the man,” which previously constituted battery, no

longer did so, because under the modified circumstances the action would likely be represented

as consensual. We did this by constructing a scenario in which a runaway trolley threatens to kill

a man walking across the tracks and the only way to save the man is to save him is to throw him

out of the path of the train, thereby seriously injuring him.

The second way we extended the results of Experiment 1 was to investigate our subjects’

knowledge of the consequentialist provision of the principle of double effect. As stated in §1,

the principle of double effect is a complex principle of justification requiring, among other

things, that the intended and foreseen good effects of an action outweigh its foreseen bad effects.

Our implicit assumption in Experiment 1 was that each of the scenarios used in that experiment

was mentally represented by our subjects as satisfying that condition. In particular, we took for

granted in Experiment 1 that individuals represented preventing the deaths of five people as an

intended and foreseen good effect that outweighed the foreseen bad effect of the death of one

person.

27

In Experiment 2, we tested our subjects’ presumed knowledge of this consequentialist

provision of the principle of double effect directly by modifying one of the Foreseeable Battery

scenarios used in Experiment 1, so that an action described as “throwing the switch,” which

previously generated intended and foreseen good effects outweighed by its foreseen bad effects,

no longer did so, but rather generated the opposite result of its intended and foreseen good

effects being outweighed by its foreseen bad effects. We did this by taking one version of the

standard Bystander Problem and substituting a valuable thing ­ “five million dollars of railroad

equipment lying idle across the tracks” ­ for what was previously described as “five men walking

across the tracks.” We reasoned that our subjects would take for granted that the moral worth of

a person is greater than that of a valuable thing.

Finally, our third modification was to convert the experimental design of Experiment 1

from a between­subject design to a within­subject design. Although fully half of the participants

in Experiment 1 provided logically adequate justifications of their judgments, on inspection it

was clear that many of these justifications were unable to explain the alternating pattern of

intuitions generated in Experiment 1. For example, many of these nominally adequate responses

took the form of either simple deontological justifications (e.g., “killing is wrong”) or simple

utilitarian justifications (“five versus one”) that were manifestly incapable of explaining the

essential contrast between the Foreseeable Battery scenarios (i.e., the Trolley and Bystander

Problems), on the one hand, and the Intentional Battery scenarios (i.e., the Transplant and

Footbridge Problems), on the other. Hence in Experiment 2 we decided to examine this issue

and to begin to explore related issues such as framing effects by asking participants to provide

judgments and justifications of a pair of trolley problems presented to them simultaneously.

28

3.1 Method

3.1.1 Participants

Participants were 65 adult volunteers between the ages of 22­35, 55 of whom were

graduate students at the John F. Kennedy School of Government at Harvard University, and 10

of whom were employees of the Cable News Network in Washington, DC. There were 33

women and 30 men. Two subjects did not provide information about their gender.

3.1.2 Stimuli and Procedure

Five scenarios were used, one of which was the Trolley Problem presented in §1.2. The

other four scenarios, which were our main concern in Experiment 2, are reprinted below. Each

was derived from a template that was designed to allow for systematic control and variation in

this and future investigations (Mikhail, 2000). Differences between the last three scenarios and

the first are underlined and italicized here to make them more noticeable. Participants were

given questionnaires without these markings.

The Bystander Problem (Hank: Foreseeable Battery) Hank is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Hank sees what has happened: the driver of the train saw five men walking across the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the five men. It is moving so fast that they will not be able to get off the track in time. Fortunately, Hank is standing next to a switch, which he can throw, that will turn the train onto a side track, thereby preventing it from killing the men. Unfortunately, there is a man standing on the side track with his back turned. Hank can throw the switch, killing him; or he can refrain from doing this, letting the five die.

Is it morally permissible for Hank to throw the switch?

The Footbridge Problem (Ian: Intentional Battery) Ian is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Ian sees what has happened: the driver of the train saw five men walking across the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the five men. It is moving so fast that

29

they will not be able to get off the track in time. Fortunately, Ian is standing next to a heavy object, which he can throw onto the track in the path of the train, thereby preventing it from killing the men. Unfortunately, the heavy object is a man, standing next to Ian with his back turned. Ian can throw the man, killing him; or he can refrain from doing this, letting the five die.

Is it morally permissible for Ian to throw the man?

The Modified Bystander Problem (Karl: Bad Effects Outweigh Good Effects) Karl is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Karl sees what has happened: the driver of the train saw five million dollars of new railroad equipment lying idle across the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the equipment. It is moving so fast that it will be destroyed. Fortunately, Karl is standing next to a switch, which he can throw, that will turn the train onto a side track, thereby preventing it from destroying the equipment. Unfortunately, there is a man standing on the side track with his back turned. Karl can throw the switch, killing him; or he can refrain from doing this, letting the equipment be destroyed.

Is it morally permissible for Karl to throw the switch?

The Modified Footbridge Problem (Luke: Consensual Contact) Luke is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Luke sees what has happened: the driver of the train saw a man walking across the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the man. It is moving so fast that he will not be able to get off the track in time. Fortunately, Luke is standing next to the man, whom he can throw off the track out of the path of the train, thereby preventing it from killing the man. Unfortunately, the man is frail and standing with his back turned. Luke can throw the man, injuring him; or he can refrain from doing this, letting the man die.

Is it morally permissible for Luke to throw the man?

In the first scenario, the agent (“Hank”) must choose whether to throw a switch in order to

prevent a runaway train from killing five people, knowing that doing so will cause the train to

run down and kill an innocent bystander (henceforth “Foreseeable Battery”). In the second

scenario, the agent (“Ian”) must choose whether to throw a man in front of a runaway train in

order to prevent the train from killing five people (henceforth “Intentional Battery”). In the third

scenario, the agent (“Karl”) must decide whether to throw a switch in order to prevent a runaway

30

train from destroying five million dollars of equipment, knowing that doing so will kill an

innocent bystander (henceforth “Bad Effects Outweigh Good Effects”). Finally, in the fourth

scenario, the agent (“Luke”) must decide whether to throw a man walking across the tracks out

of the path of the train, knowing that doing so will injure him (henceforth “Consensual

Contact”).

Unlike Experiment 1, which used a between­subject design, Experiment 2 employed a

within­subject design. Each of the 65 participants received a written questionnaire containing

two scenarios, including one or more of the four scenarios reprinted above. 6 Participants were

first asked whether the proposed actions were “morally permissible” and then to explain or

justify their responses. 25 participants were given the Intentional Battery scenario, 25

participants were given the Foreseeable Battery scenario, 25 participants were given the Bad

Effects Outweigh Good Effects scenario, and 25 participants were given the Consensual Contact

scenario. The assignment of participants to scenario type was random.

3.2 Results

3.2.1 Intentional Battery vs. Foreseeable Battery

We present the main results of Experiment 2 in stages, beginning with the comparison

between intentional and foreseeable battery (Figure 8). 2 of 25 subjects in the Intentional Battery

condition judged the action constituting intentional battery (“throwing the man”) to be

permissible. Meanwhile, 19 of 25 subjects in the Foreseeable Battery condition judged the

action constituting foreseeable battery (“throwing the switch”) to be permissible. This difference

is significant: x 2 (1, N=50) = 24.4, p < .001.

31

0

5

10

15

20

25

Intentional Battery

Foreseeable Battery

Permiss. Impermiss.

Act Type

Subjects

Figure 8: Moral Judgments of Two Act Types in Experiment 2 (Intentional Battery vs. Foreseeable Battery)

X 2 (1, N=50) = 24.4, p < .001

0

2

4

6

8

10

12

Men Women Men Women

Permiss. Impermiss.

Intentional Battery

Subjects

Figure 9: Judgments of Two Act Types in Experiment 2 by Gender (Intentional Battery vs. Foreseeable Battery)

Foreseeable Battery

Act Type

X 2 (1, N=21) = 11.4, p < .001 (women).

X 2 (1, N=27) = 13.38, p < .001 (men).

32

Male and female participants who were given these two scenarios showed a similar

pattern of responses (Figure 9). 2 of 14 men and 0 of 11 women in the Intentional Battery

condition judged throwing the man the action constituting intentional battery (“throwing the

man”) to be permissible. Meanwhile, 11 of 13 men and 6 of 10 women in the Foreseeable

Battery condition judged the action constituting foreseeable battery (“throwing the switch”) to be

permissible. These differences are also significant, x 2 (1, N=21) = 11.4, p < .001 (women) and

x 2 (1, N=27) = 13.38, p < .001 (men).

3.2.2 Good Effects Outweigh Bad Effects vs. Bad Effects Outweigh Good Effects

Next, we describe the results of Experiment 2 in terms of the weighing of good and bad

effects (Figure 10). As indicated, 19 of 25 subjects who were given the Hank scenario (now re­

categorized as “Good Effects Outweigh Bad Effects”) judged Hank’s throwing the switch to be

permissible. By contrast, none of the 25 subjects who were given the Karl scenario (“Bad

Effects Outweigh Good Effects”) judged Karl’s throwing the switch to be permissible. This

difference is significant: x 2 (1, N=50) = 30.65, p < .001.

Men’s and women’s responses followed the same pattern (Figure 11). 11 of 13 men and

6 of 10 women judged throwing the switch to be permissible in the Good Effects Outweigh Bad

Effects condition. By contrast, 0 of 14 men and 0 of 11 women held throwing the switch to be

impermissible in the Bad Effects Outweigh Good Effects condition. These results are also

significant, x 2 (1, N=21) = 11.4, p < .001 (women) and x 2 (1, N=27) = 19.99, p < .001 (men).

33

0

5

10

15

20

25

Bad Effect > Good Effect

Good Effect > Bad Effect

Permiss. Impermiss.

Act Type

Subjects

Figure 10: Moral Judgments of Two Act Types in Experiment 2 (Good Effects vs. Bad Effects)

X 2 (1, N=50) = 30.65, p < .001

0

2

4

6

8

10

12

Men Women Men Women

Permiss. Impermiss.

Bad Effect > Good Effect

Subjects

Figure 11: Judgments of Act Types in Experiment 2 by Gender (Good Effects vs. Bad Effects)

Good Effect > Bad Effect

Act Type

X 2 (1, N=21) = 11.4, p < .001 (women).

X 2 (1, N=27) = 19.99, p < .001 (men).

34

3.2.3 Intentional Battery vs. Consensual Contact

Third, we examine the comparison between intentional battery and consensual contact

(Figure 12). As indicated, 2 of 25 subjects in the Intentional Battery condition (“Ian”) judged the

action constituting intentional battery (“throwing the man”) to be permissible. By contrast, 24 of

25 subjects in the Consensual Contact condition (“Luke”) judged the action constituting

consensual contact (“throwing the man”) to be permissible. This difference is significant: x 2 (1,

N=50) = 38.78, p < .001.

Again, male and female responses conformed to the same pattern (Figure 13). 2 of 14

men and none of the 11 women in the Intentional Battery condition judged throwing the man to

be permissible. By contrast, all 7 of the men and all 16 of the women in the Consensual Contact

condition judged throwing the man to be permissible, x 2 (1, N=27) = 27.0, p < .001 (women) and

x 2 (1, N=21) = 14.0, p < .001 (men). 7

3.2.4 Justifications

Finally, we turn to our subjects’ expressed justifications, that is, the responses they

provided to justify or explain their judgments. Because we utilized a within­subject design in

Experiment 2, we expected that these justifications would be significantly less adequate than the

corresponding justifications in Experiment 1, which relied on a between­subject design. In

addition, we predicted that subjects presented with both an Intentional Battery scenario and a

Foreseeable Battery scenario, in particular, would not be able to justify their conflicting

intuitions.

35

0

5

10

15

20

25

Intentional Battery

Consensual Contact

Permiss. Impermiss.

Act Type

Subjects

Figure 12: Moral Judgments of Two Act Types in Experiment 2 (Intentional Battery vs. Consensual Contact)

X 2 (1, N=50) = 38.78, p < .001

0 2

4 6

8

10 12

14

16

Men Women Men Women

Permiss. Impermiss.

Intentional Battery

Subjects

Figure 13: Judgments of Act Types in Experiment 2 by Gender (Intentional Battery vs. Consensual Contact)

Consensual Contact

Act Type

X 2 (1, N=27) = 27.0, p < .001 (women).

X 2 (1, N=21) = 14.0, p < .001 (men).

36

Both of these predictions were confirmed. First, 35.4% (23/65) of participants gave no

justification and 38.5% (25/65) provided logically inadequate justifications, while only 20.0%

(13/65) provided logically adequate justifications. This contrasts sharply with the Experiment 1,

in which 50% of subjects provided logically adequate justifications.

Second, only 10% (1/10) of those subjects who were given both the Intentional Battery and

Foreseeable Battery scenarios and who attempted to provide some sort of explanation for their

judgments provided logically adequate justifications. The other 90% (9/10) provided logically

inadequate justifications. Further, as Table 2 reveals, this group’s expressed principles were

widely divergent. Many participants merely restated the problem they were asked to resolve or

otherwise provided answers which were non­responsive. Moreover, several participants

appeared puzzled by the nature and strength of their intuitions and by how those intuitions

shifted as a result of apparently minor and inconsequential differences in the relevant action

descriptions.

3.3 Discussion

The results of Experiment 2 corroborate and extend those of Experiment 1. First, they

lend further support to the hypothesis that both men and women possess intuitive or unconscious

knowledge of the prohibition of intentional battery and the principle of double effect. By

imputing knowledge of these principles to our subjects, we can explain and predict their moral

intuitions. Specifically, we can explain why their intuitions flip so predictably when the standard

Bystander Problem is modified so that the costs of throwing the switch outweigh its benefits and

the standard Footbridge Problem is modified so that throwing the man no longer constitutes

intentional battery.

37

Table 2: Justifications for Bystander and Footbridge Pair of Problems in Experiment 2

Problem Type of Battery

Judgment Justification

Bystander Foreseeable Permissible

Footbridge Intentional Impermissible

Very odd. I don't know why I chose differently in the second scenario. The end result is the same. I just chose my gut response­­and now am intrigued with how to reconcile them.

Bystander Foreseeable Permissible

Footbridge Intentional Impermissible

It's amazing that I would not throw a person but throw a switch to kill a person. I really wish there was more I could do for the 1 guy on the other track.

Bystander Foreseeable Permissible

Footbridge Intentional Impermissible

In either case, the moral decision rule depends on how close to the active killing of the man is.

Bystander Foreseeable Permissible

Footbridge Intentional Impermissible

Not acceptable to decide to risk someone else's life to save others.

Bystander Foreseeable Permissible

Footbridge Intentional Impermissible

I know­­five lives are five lives­­it's all about the guts. That's what it comes down to. Blaise Pascal got it all wrong.

Bystander Foreseeable Permissible

Footbridge Intentional Impermissible

The man, Hank can here actively influence a sequence of events which will limit damage (# of deaths). In the second event, he cannot throw another man onto the tracks because he will actively and deliberately kill an innocent bystander. Really an impossible choice.

Bystander Foreseeable Permissible

Footbridge Intentional Impermissible

Moral actors may be forced to make a decision between two passive choices where both will end rights. But to make action over passive choices requires another kind of analysis and degree of benefit.

Bystander Foreseeable Permissible

Footbridge Intentional Impermissible

In the first scenario it would be permissible to act as a utilitarian optimizer. In the second rights come into question.

Bystander Foreseeable Permissible

Footbridge Intentional Permissible

I believe that the ultimate question is that of lives lost. Some would argue that Hank and Ian would be morally justified in not stopping the train. While this may be true, it does not necessitate that it be morally unjustified to stop the train.

Bystander Foreseeable Impermissible

Footbridge Intentional Impermissible

For the first scenario, I wanted to draw a distinction between "is it permissible for him to throw the switch" and "does he have a duty to throw the switch," though I don't know if that would have changed my answer.

38

Second, the results of Experiment 2 suggest that individuals have limited conscious

access to these principles (or to whichever principles are actually responsible for generating their

intuitions). Even under a liberal coding scheme, only 20% of subjects provided logically

adequate justifications for their judgments. Further, only 10% did so when asked to explain the

most challenging pair of moral intuitions, namely, the perceived contrast between the Bystander

and Footbridge problems.

Third, Experiment 2 provides some initial evidence of framing effects. Most notably,

only 76% (19/25) of respondents in the Foreseeable Battery condition judged Hank’s throwing

the switch to be permissible, a much lower percentage than the 95% (19/20) of participants who

gave this response in Experiment 1. These effects were slightly less pronounced in males than in

females, but they were discernible in both groups: 85% (11/13) of men gave this response, as

compared with 100% (11/11) in Experiment 1, whereas 60% (6/10) of women gave this

response, as compared with 89% (8/9) in Experiment 1. These sample sizes are obviously quite

small, and it therefore would be premature to draw any firm conclusions about these effects at

this point. It seems likely, however, that a more systematic investigation of framing effects in

larger populations would yield significant results, perhaps including significant gender

differences. Nevertheless, the main pattern of intuitions Experiment 2 fell in line with those of

Experiment 1, in that both men and women in the aggregate recognized the relevant distinctions

among the Bystander, Footbridge, Good Effects Outweigh Bad Effects, and Consensual Contact

Problems. Hence Experiment 2 provides additional evidence that at least some moral intuitions

and the principles that generate them are widely shared, irrespective of demographic variables

like gender.

39

4. Experiment 3

Participants in Experiments 1­2 included persons from countries other than the United

States, including Belgium, Canada, Columbia, Denmark, France, Germany, Israel, Japan, Korea,

Lebanon, Mexico, and Puerto Rico. Nonetheless, only one or a few individuals from each of

these countries were represented, and the majority of participants were United States citizens or

members of other Western nations. Accordingly, Experiment 3 was designed to investigate the

moral intuitions of a “non­Western” population.

4.1 Method

4.1.1 Participants

Participants were 39 adult volunteers ages 18­65 from the broader Cambridge,

Massachusetts community, all of whom had emigrated from China within the previous five years

and most of whom had done so within the previous two years. The group included 19 women

and 19 men; 1 participant did not volunteer information about his or her gender. 8

4.1.2 Stimuli and Procedure

Same as Experiment 2, except that participants in this study were not asked to justify

their judgments. 14 participants were given the Intentional Battery scenario, 16 participants

were given the Intentional Battery scenario, 15 participants were given the Bad Effects Outweigh

Good Effects scenario, and 16 participants were given the Consensual Contact scenario. The

assignment of participants to scenario type was random.

40

4.2 Results

4.2.1 Intentional Battery vs. Foreseeable Battery

Once again, we present the results of Experiment 3 in stages, beginning with the

comparison between intentional and foreseeable battery (Figure 14). 2 of 14 subjects in the

Intentional Battery condition judged the action constituting intentional battery (“throwing the

man”) to be permissible. Meanwhile, 11 of 14 subjects in the Foreseeable Battery condition

judged the action constituting foreseeable battery (“throwing the switch”) to be permissible.

This difference is significant: x 2 (1, N=28) = 11.72, p < .001.

4.2.2 Good Effects Outweigh Bad Effects vs. Bad Effects Outweigh Good Effects

Due to the limited number of subjects in Experiment 3, we refrain from analyzing our

responses by gender. Instead, we turn directly to the comparison between good and bad effects

(Figure 15). 11 of 14 subjects in the Good Effects Outweigh Bad Effects condition judged

throwing the switch to be permissible. Meanwhile, only 1 of 15 subjects in the Bad Effects

Outweigh Good Effects condition judged throwing the switch to be permissible. This difference

is significant: x 2 (1, N=29) = 16.81, p < .001.

4.2.3 Intentional Battery vs. Consensual Contact

Third, we examine the contrast between intentional battery and consensual contact

(Figure 16). 2 of 16 subjects in the Intentional Battery condition judged throwing the man to be

permissible. Meanwhile, 14 of 16 subjects in the Consensual Contact condition judged throwing

the man to be permissible. This difference is significant: x 2 (1, N=32) = 18.0, p < .001.

41

0

2

4

6

8

10

12

Intentional Battery

Foreseeable Battery

Permiss. Impermiss.

Act Type

Subjects

Figure 14: Moral Judgments of Two Act Types in Experiment 3 (Intentional Battery vs. Foreseeable Battery)

X 2 (1, N=28) = 11.72, p < .001

0

2

4

6

8

10

12

14

Bad Effect > Good Effect

Good Effect > Bad Effect

Permiss. Impermiss.

Act Type

Subjects

Figure 15: Moral Judgments of Two Act Types in Experiment 3 (Good Effects vs. Bad Effects)

X 2 (1, N=29) = 16.81, p < .001

42

4.3 Discussion

The results of Experiment 3 suggest that the central findings of Experiments 1­2 are not

limited to persons educated or raised in the United States or other Western nations. Instead, they

suggest at least some operative principles of moral competence, including the prohibition of

intentional battery and the principle of double effect, are transnational and may be universal.

While claims of universality are often controversial and should be made with care, this

hypothesis is consistent with the role these principles already play in international law (i.e., the

“law of nations”). For example, the principle of double effect’s implied norm of non­combatant

immunity—that is, its prohibition against directly targeting civilians, together with its qualified

0

2

4

6

8

10

12

14

Intentional Battery

Consensual Contact

Permiss. Impermiss.

Act Type

Subjects

Figure 16: Moral Judgments of Two Act Types in Experiment 3 (Intentional Battery vs. Consensual Contact)

X 2 (1, N=32) = 18.0, p < .001

43

acceptance of harming civilians as a necessary side effect of an otherwise justifiable military

operation—has long been part of customary international law and is codified in Article 48 of the

First Protocol (1977) to the 1949 Geneva Conventions (e.g., Henkin, Pugh, Schacter & Smit,

1993, p. 364­65). Likewise, the principle of double effect’s implied norm of proportionality is

also part of customary international law and is codified in Articles 22­23 of the Hague

Convention of 1907 (e.g., Henkin et al., 1993, p. 368). Further, many important legal doctrines,

in both American law and the domestic law of other nations, turn on an analysis of purpose and

the distinction between intended and foreseen effects (Mikhail, 2002). Hence it is perhaps not

surprising to discover that thought experiments like trolley problems, which implicate these

concepts, elicit widely shared moral intuitions from individuals of different cultural

backgrounds.

Nevertheless, while Experiment 3 provides some initial support for the existence of moral

universals, this support is obviously quite limited. More empirical investigation on a much wider

scale is necessary before specific claims about universality could be defensible. In the context of

our hypothesis, what would perhaps be most compelling in this regard would be to collect

additional evidence on trolley intuitions from individuals from around the world, in particular

those from markedly different cultural, social, religious, and socioeconomic backgrounds. To do

this, one would presumably need to translate these thought experiments into different languages.

One might also need to modify them in culturally specific ways, insofar as certain inessential

elements of the scenarios (e.g., trolleys) may be unfamiliar. We do not attempt these extensions

in this paper but merely identify them as objectives of future research which flow naturally from

the studies presented here. 9

44

5. Experiment 4

Experiments 1­3 suggest that the moral competence of adults includes the prohibition of

intentional battery and the principle of double effect. By attributing intuitive knowledge of these

principles to our subjects, we can explain and predict their moral intuitions.

As Table 3 indicates, the computations presupposed by this explanation can be

reconstructed in the simple form of series of yes­no questions or decision tree. Presented with a

presumptively wrong action, such as those harmful actions at issue in the Trolley, Transplant,

Bystander, and Footbridge problems, the decision­maker first asks whether the proposed action’s

good effects outweigh its bad effects. If the answer is no, then the decision­maker concludes the

action is impermissible. If the answer is yes, then the decision­maker next asks whether the

action involves committing a battery as a means to achieve a given end. If the answer is no, then

the decision­maker concludes that the action is permissible. If the answer is yes, then the

decision­maker concludes that the action is impermissible.

Table 3: Explanation of Trolley, Transplant, Bystander, and Footbridge Intuitions as a Function of the Principle of Double Effect Problem (Agent) Homicide? Battery? Good Effects

Outweigh Bad Effects?

Battery as a Means?

Deontic Status

Trolley (Charlie) Yes Yes Yes No Permissible Transplant (Dr.Brown) Yes Yes Yes Yes Impermissible Bystander (Denise) Yes Yes Yes No Permissible Footbridge (Nancy) Yes Yes Yes Yes Impermissible Bystander (Hank) Yes Yes Yes No Permissible Footbridge (Ian) Yes Yes Yes Yes Impermissible Modified Bystander (Karl) Yes Yes No No Impermissible Modified Footbridge (Luke) No No Yes No Permissible

45

Table 3 illustrates that our central findings up to this point can be explained in the

foregoing terms. However, our findings are also consistent with an alternative explanation,

according to which trolley intuitions do not depend primarily on the mental state properties of an

agent’s action plan, but on its temporal properties, in particular whether its bad effects (or its

prima facie wrongs such as battery) are mentally represented as occurring before or after its

good effects. In particular, our central findings could be equally explained by the so­called

“Pauline Principle,” which holds that “it is impermissible to do evil that good may come”

(Anscombe, 1970; Donagan, 1977). Suitably formalized, a temporal interpretation of this

principle would in effect compute as “impermissible” any action plan which represents either a

bad effect or a battery occurring before a good effect. As Table 4 reveals, all but one of the

impermissible act representations examined thus far possess this property, the lone exception,

Karl’s throwing the switch in the Modified Bystander Problem, being explainable on other

grounds. 10 Hence the Pauline Principle (or some suitable formalization of it) also constitutes

(part of) an observationally adequate explanation of the results of Experiments 1­3.

Table 4: Explanation of Trolley, Transplant, Bystander, and Footbridge Intuitions as a Function of the Pauline Principle Problem (Agent) Homicide? Battery? Good Effects

Outweigh Bad Effects?

Battery or Bad Effects Prior to Good Effects?

Deontic Status

Trolley (Charlie) Yes Yes Yes No Permissible Transplant (Dr.Brown) Yes Yes Yes Yes Impermissible Bystander (Denise) Yes Yes Yes No Permissible Footbridge (Nancy) Yes Yes Yes Yes Impermissible Bystander (Hank) Yes Yes Yes No Permissible Footbridge (Ian) Yes Yes Yes Yes Impermissible Modified Bystander (Karl) Yes Yes No No Impermissible Modified Footbridge (Luke) No No Yes No Permissible

46

Experiment 4 was designed to investigate this logically possible alternative, as well as to

provide an additional check on the abstract concept of battery utilized in Experiments 1­3. To

accomplish these objectives, we constructed the following two new scenarios (Mikhail, 2000):

The Looped Track Problem—Intentional Battery (Ned) Ned is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Ned sees what has happened: the driver of the train saw five men walking across the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the five men. It is moving so fast that they will not be able to get off the track in time. Fortunately, Ned is standing next to a switch, which he can throw, that will temporarily turn the train onto a side track. There is a heavy object on the side track. If the train hits the object, the object will slow the train down, thereby giving the men time to escape. Unfortunately, the heavy object is a man, standing on the side track with his back turned. Ned can throw the switch, preventing the train from killing the men, but killing the man. Or he can refrain from doing this, letting the five die.

Is it morally permissible for Ned to throw the switch?

The Looped Track Problem—Foreseeable Battery (Oscar) Oscar is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Oscar sees what has happened: the driver of the train saw five men walking across the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the five men. It is moving so fast that they will not be able to get off the track in time. Fortunately, Oscar is standing next to a switch, which he can throw, that will temporarily turn the train onto a side track. There is a heavy object on the side track. If the train hits the object, the object will slow the train down, thereby giving the men time to escape. Unfortunately, there is a man standing on the side track in front of the heavy object, with his back turned. Oscar can throw the switch, preventing the train from killing the men, but killing the man. Or he can refrain from doing this, letting the five die.

Is it morally permissible for Oscar to throw the switch? 11

In the first scenario (“Ned”), battery is embedded within the agent’s action plan as a means. In

the second scenario (“Oscar”), battery is embedded within the agent’s action plan as a side

effect. Unlike the scenarios used in Experiments 1­3, however, the Ned­Oscar pair is not

distinguishable in terms of their morally neutral basic actions (e.g., “throwing the switch” vs.

“throwing the man”) or the temporal properties of their good effects, bad effects, and batteries.

47

Instead, five fundamental properties are held constant between these two scenarios: (1) good

effects, (2) bad effects, (3) ultimate purpose or goal, (4) morally neutral basic action (“throwing

the switch” in each case), and (5) the temporal order of good effects, bad effects, and batteries.

Further, both are “impersonal” scenarios in the sense defined by Greene and colleagues (Greene

et al., 2001). The Ned­Oscar pair is therefore the purest “minimal pair” of scenarios used thus

far in our investigations.

5.1 Method

5.1.1 Participants

Participants were 309 adult volunteers ages 18­35 from the M.I.T. community. Because

the postulated difference between the relevant scenarios was quite subtle, we greatly increased

our sample sizes in order to be able to detect statistically significant differences in their

underlying representations. For the purposes of this study, we did not actively collect

information on participants’ gender. However, a retrospective analysis of participants’ names

indicated that at least 119 men and at least 117 women participated in this study. An additional

73 individuals did so whose gender was not readily ascertainable in this manner.

5.1.2 Stimuli and Procedure

Two scenarios were used. In one (“Ned”), battery was embedded within the agent’s

action plan as a means (henceforth “Intentional Battery”). In the other (“Oscar”), battery was

embedded within the agent’s action plan as a side effect (henceforth “Foreseeable Battery”). In

both scenarios, good effects, bad effects, ultimate purpose or goal, and morally neutral basic

48

action (“throwing the switch”) were held constant. In addition, temporal order of good effects,

bad effects, and batteries were also held constant.

A between­subject design was utilized. Each participant received a questionnaire

containing one written scenario, accompanied by a diagram designed to make the scenario fully

comprehensible (on file with author). The participant was instructed to read the scenario and to

determine whether the proposed action it described was “morally permissible.” Unlike

Experiments 1 and 2, the participants were not asked to provide justifications for their

judgments; nevertheless, many individuals did provide justifications on their own initiative, and

these responses are analyzed below. 159 individuals were given the Intentional Battery scenario

(“Ned”) and 150 were given the Foreseeable Battery scenario (“Oscar”). The assignment of

participants to scenario type was random.

5.2 Results

5.2.1 Judgments

The main results of Experiment 4 are summarized in Figure 17. 76/159 or 48% of

participants who were given the Intentional Battery scenario (“Ned”) judged throwing the switch

to be permissible. Meanwhile, 93/150 or 62% of participants who were given the Foreseeable

Battery scenario (“Oscar”) judged throwing the switch to be permissible. This difference is

significant: x 2 (1, N=302) = 6.52, p < .025.

Judgments of men and women are presented in Figure 18. 54% (29/54) of men and 45%

(32/71) of women in the Intentional Battery condition judged throwing the switch to be

permissible. Meanwhile, 72% (47/65) of men and 59% (27/46) of women held the same action

to be permissible in the Foreseeable Battery condition. These contrasts are significant for men,

49

0 5 10 15 20 25 30 35 40 45 50

Men Women Men Women

Permiss. Impermiss.

Intentional Battery

Subjects

Figure 18: Judgments of Act Types in Experiment 4 by Gender (Intentional Battery vs. Foreseeable Battery)

Foreseeable Battery

X 2 (1, N=117) = 2.07, p < .2 (women).

X 2 (1, N=119) = 4.42, p < .05 (men).

0 10 20 30 40 50 60 70 80 90 100

Intentional Battery

Foreseeable Battery

Permiss. Impermiss.

Act Type

Subjects

Figure 17: Moral Judgments of Two Act Types in Experiment 4 (Intentional Battery vs. Foreseeable Battery)

X 2 (1, N=302) = 6.52, p < .025

50

x 2 (1, N=119) = 4.42, p < .025, but not for women, x 2 (1, N=117) = 2.07, p < .20. Hence, based

on this data, the null hypothesis that these scenarios are indistinguishable is falsified for men but

not for women. However, larger sample sizes would presumably support the same conclusion

with respect to women. These results also indicate a slight but discernible trend in which, in the

aggregate, men appear more willing than women to permit throwing the switch in these

circumstances.

5.2.2 Justifications

Although we did not ask for justifications, 49 subjects provided some sort of verbalized

explanation of their judgments on their own initiative. 30 of these 49 responses, or 9.9%

(30/302) of the overall total, were logically adequate justifications, while 19 of these 49

responses, or 6.3% (19/302) of the overall total, were logically inadequate. Meanwhile, 83.8%

(253/302) of subjects in Experiment 4 gave no justification.

5.3 Discussion

According to our hypothesis, when individuals encounter hypothetical fact patterns like the

Trolley, Transplant, Footbridge and Bystander problems, they spontaneously compute

unconscious representations of the relevant actions in terms of ends, means, and side effects.

They also distinguish battery as a means from battery as a side effect, prohibiting the former but

permitting the latter in the specific circumstances depicted by these problems. Consequently, we

were led to make two related predictions about the Ned and Oscar pair of scenarios. First, we

predicted that subjects would perceive an intuitive distinction between these scenarios, even

though their overt differences are quite subtle, because in only one of them (Ned) does the agent

51

intend to commit a battery as a means of furthering his good end. In the Oscar scenario, by

contrast, we assumed that subjects would compute a representation according to which battery is

not a means but a foreseen side effect. Second, we predicted that, on the basis of this distinction

between means and side effects, subjects would judge Oscar’s act of throwing the switch to be

permissible but Ned’s act of throwing the switch to be impermissible.

The results of Experiment 4 confirmed the first prediction. Although the differences

between these fact patterns are minimal, our subjects did in fact distinguish the two scenarios to

a statistically significant extent. That is, we were able to falsify the null hypothesis that these

scenarios are intuitively indistinguishable. This implies that, despite sharing the five

fundamental properties described above, the Ned and Oscar scenarios trigger distinct mental

representations whose properties are morally salient. This in turn lends at least some support to

the hypothesis that the operative distinction between these scenarios is the distinction between

battery as a means and battery as a side effect.

By contrast, the second prediction did not hold, or rather, it held only to a limited extent.

Although a majority (62%) of those participants in the Foreseeable Battery condition held

Oscar’s throwing the switch to be permissible, while a minority (47%) of those in Intentional

Battery condition held Ned’s throwing the switch to be permissible, the contrast between these

percentages was less sharp than in our previous studies. Further, the Ned responses were no

different than chance in this regard. This was also a departure from our prior studies, in which

the number of participants holding acts constituting intentional battery to be permissible was

small enough to warrant the claim that, as a general matter, individuals regard these acts to be

impermissible.

52

Nevertheless, Experiment 4 did confirm an intuitive distinction between this pair of

cases, despite their close similarities. Further, several explanations of these comparatively

anomalous results suggest themselves and raise interesting problems for future research. We

briefly mention two such possibilities here, leaving their investigation for another occasion.

First, although trolley intuitions are normally quite sharp (Thomson, 1986), it is also a

familiar observation that they “begin to fail above a certain level of complexity” (Nagel, 1986, p.

174). Indeed, some trolley problems are so complex and bizarre that they do not appear to be

particularly useful given the central aims and methods of cognitive science (e.g., Unger, 1996).

While the Ned­Oscar pair arguably does not fit the latter category, these scenarios also are

undeniably more complex and difficult to process than the problems used in our prior studies.

Indeed, this is one reason we provided participants in this study with a diagram to facilitate

comprehension. Considerations like these suggest that the comparatively anomalous results of

Experiment 4 may be understood as the predictable effect of increasing the amount of relevant

information subjects are required to process in the course of fastening upon a morally salient

structural description. In the case of language, it is well understood that certain non­linguistic

factors, such as memory limitations and other general limits on how the mind processes

information, can interrupt or interfere with the parsing of linguistic expressions. This, of course,

is one reason why linguists draw the competence­performance distinction and a related

distinction between grammaticality and acceptability judgments (Chomsky, 1965; Haegeman,

1994; Mikhail, 2002). A similar situation has been thought to obtain in the moral domain

(Dwyer, 1999; Mikhail, 2000). If so, then it is possible that the comparatively anomalous

findings of Experiment 4 can be explained within the framework of a moral competence­

performance distinction.

53

Second, a disparity between intended and actual result (LaFave, 2003) may also help

explain these findings. The standard operational test for distinguishing necessary means from

unnecessary side effect is a counterfactual test, according to which one asks whether the actor

would have acted differently if the negative result could have been avoided. By this measure,

Ned but not Oscar is presumed to intend a battery a means to achieving his end, because Ned’s

objectives include causing the train to hit the man, whereas Oscar’s objectives do not. Put

differently, if circumstances were altered and the man were no longer on the side track, then

presumably Ned would not throw the switch, because his immediate purpose in doing so is to

cause the train to hit the man. By contrast, Oscar presumably would still throw the switch in

these circumstances, because his immediate purpose is to cause the train to hit the object, of

which the man is standing in front.

The disparity between intended and actual result complicates this analysis. This concept

refers to the fact that, although Oscar’s intentions do not include saving the men by causing the

train to hit the man, the actual result of his action will likely be just that, because unless the man

jumps off the track in time, Oscar’s throwing the switch will cause the train to hit the man before

the train hits the object. If so, then, from an ex post perspective, there is a sense in which it will

be true that Oscar committed a battery as a means to saving the five men, even though, from an

ex ante perspective, it was not his intention to do so.

Significantly, this potential disparity between intended and actual result is not present in

any of the other Foreseeable Battery scenarios used in our investigations. Instead, in these other

scenarios, the intended good result of saving the five men necessarily occurs before, and

independently of, the foreseen bad result of committing a battery by causing the train to hit the

man. 12 Hence this important structural difference between the Oscar scenario and the

54

Foreseeable Battery scenarios utilized in our previous experiments may also help explain the

comparatively anomalous results of Experiment 4.

6. Experiment 5

As interpreted here, the principle of double effect is a complex principle of justification

which holds that an otherwise prohibited action may be permissible if (1) the act itself is not

wrong, (2) the good but not the bad effects are intended, (3) the good effects outweigh the bad

effects, and (4) no morally preferable alternative is available. In Experiments 1­4, we focused

mainly on the first and third provisions of this principle. In Experiment 5, we turned our

attention to the second and fourth provisions.

6.1 Method

6.1.1 Participants

Participants were 60 adult volunteers ages 18­65 from the Washington, D.C. area. The

group consisted of 32 women and 28 men.

6.1.2 Stimuli and Procedure

Four scenarios were used, one of which was the Bystander Problem (see §3.1.2) and the

other three of which were the following modifications of this problem:

Modified Bystander Problem # 2 (Mark: Intentional Homicide) Mark is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Mark sees what has happened: the driver of the train saw five men walking across the tracks and slammed on the brakes, but the brakes failed, and the driver fainted. The train is now rushing toward the five men. It is moving so fast that they will not be able to get off the track in time. Fortunately, Mark is standing next to a switch, which he can throw, that will turn the train onto a side track, thereby preventing it from killing the men. Unfortunately, there is a man on the side track. Mark can throw the switch, killing him; or he can refrain from doing this, letting the men die.

55

Mark then recognizes that the man on the side track is someone who he hates with a passion. “I don’t give a damn about saving those five men,” Mark thinks to himself, “but this is my chance to kill that bastard.”

Is it morally permissible for Mark to throw the switch in order to kill the man on the side track? 13

Modified Bystander Problem #3 (Richard: Better Alternative) Richard is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Richard sees what has happened: the driver of the train saw five men walking across the tracks and slammed on the brakes, but the brakes failed, and the driver fainted. The train is now rushing toward the five men. It is moving so fast that they will not be able to get off the track in time. Fortunately, Richard is standing next to a switch, which he can throw, that will turn the train onto a side track, thereby preventing it from killing the men. Unfortunately, there is a man on the side track with his back turned. Richard can throw the switch, killing him; or he can refrain from doing this, letting the men die.

By pulling an emergency cord, Richard can also redirect the train to a third track, where no one is at risk. If Richard pulls the cord, no one will be killed. If Richard throws the switch, one person will be killed. If Richard does nothing, five people will be killed.

Is it morally permissible for Richard to throw the switch?

Modified Bystander Problem #4 (Steve: Disproportional Death) Steve is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Steve sees what has happened: the driver of the train saw a man walking across the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the man. It is moving so fast that he will not be able to get off the track in time. Fortunately, Steve is standing next to a switch, which he can throw, that will turn the train onto a side track, thereby preventing it from killing the man. Unfortunately, there are five men standing on the side track with their backs turned. Steve can throw the switch, killing the five men; or he can refrain from doing this, letting the one man die.

Is it morally permissible for Steve to throw the switch?

In the first scenario, the agent (“Mark”) contemplates whether to throw the switch, not for the

purpose of saving the men, but for the purpose of killing the man on the side track (henceforth

“Intentional Homicide”). 14 In the second scenario, the agent (“Richard”) must choose whether to

throw the switch in order to prevent the train from killing five at the cost of killing one, but there

56

is a third option that will result in no one being killed (henceforth, “Better Alternative”). Finally,

in the third scenario, which was designed to replicate and extend the results of Experiments 2­3

concerning the manipulation of good and bad effects, the agent (“Steve”) must choose whether to

throw a switch in order to prevent a runaway train from killing one person, knowing that doing

so will cause the train to kill five other people (henceforth “Disproportional Death”).

A within­subject design was utilized. Each of the 60 participants received a written

questionnaire containing two scenarios, including (1) one of these three scenarios and (2) the

original Bystander Problem (Hank). Hence there were three conditions, with 20 participants

assigned randomly to each condition. Participants were asked to read the scenarios and decide

whether the proposed actions were “morally permissible” and then to explain or justify their

responses.

6.2 Results

6.2.1 Intentional Homicide v. Bystander Problem

We present the results of Experiment 5 in stages, beginning with the comparison between

the Intentional Homicide Problem and the original Bystander Problem (Figure 19). 4 of 20

subjects in this condition judged the action constituting intentional homicide (“Mark’s throwing

the switch”) to be permissible. Meanwhile, 16 of 20 subjects in this condition judged the same

action (“Hank’s throwing the switch”) in the original Bystander Problem to be permissible. This

difference is significant: x 2 (1, N=20) = 14.4, p < .001.

57

0 2 4 6 8 10 12 14 16

Intentional Homicide

Bystander Problem

Permiss. Impermiss.

Act Type

Subjects

Figure 19: Moral Judgments of Two Act Types in Experiment 5 (Intentional Homicide vs. Bystander Problem)

X 2 (1, N=20) = 14.4, p < .001

0 2 4 6 8 10 12 14 16

Better Alternative

Bystander Problem

Permiss. Impermiss.

Act Type

Subjects

Figure 20: Moral Judgments of Two Act Types in Experiment 5 (Better Alternative vs. Bystander Problem)

X 2 (1, N=20) = 8.12, p < .01

58

6.2.2 Better Alternative v. Bystander Problem

Due to the limited number of subjects in Experiment 5, we refrain from analyzing our

responses by gender. Instead, we turn directly to the comparison between the Better Alternative

Problem and the Bystander Problem and (Figure 20). 15 of 20 subjects in this condition judged

Hank’s throwing the switch to be permissible in the Bystander Problem. By contrast, only 6 of

20 subjects judged the same action (“Richard’s throwing the switch”) to be permissible in the

presence of a better alternative. This difference is significant: x 2 (1, N=20) = 8.12, p < .01.

6.2.3 Disproportional Death v. Bystander Problem

Third, we turn to the comparison between the Disproportional Death Problem and the

Bystander Problem and (Figure 21). 3 of 20 subjects in this condition judged the action

0 2 4 6 8 10 12 14 16 18

Disproportional Death

Bystander Problem

Permiss. Impermiss.

Act Type

Subjects

Figure 21: Moral Judgments of Two Act Types in Experiment 5 (Disproportionate Death vs. Bystander Problem)

X 2 (1, N=20) = 7.03, p < .01

59

generating Disproportional death (“Steve’s throwing the switch”) to be permissible. By contrast,

11 of 20 subjects in this condition judged the same action (“Hank’s throwing the switch”) in the

original Bystander Problem to be permissible. This difference is significant: x 2 (1, N=20) = 7.03,

p < .01.

6.2.4 Justifications

Finally, we turn to participants’ justifications for their judgments. The fact that we

utilized a within­subject design led us to expect that justifications in Experiment 5 would be less

adequate than those in Experiment 1 (which used a between­subject design) and more like those

in Experiment 2 (which used a within­subject design). This prediction was confirmed. Fifty­

three percent (32/60) of participants gave no justification, 23% (14/60) provided logically

inadequate justifications, while 23% (14/60) provided logically adequate justifications—a figure

much closer to the percentage of logically adequate justifications in Experiment 2 (20%) than in

Experiment 1 (50%).

6.3 Discussion

The results of Experiment 5 lend further support to our hypothesis that the moral

competence of adults includes intuitive knowledge of the principle of double effect. By

attributing knowledge of this principle to our subjects, we can explain why their deontic

intuitions change when the standard Bystander Problem is modified such that (1) the bad but not

the good effects are intended, or (2) a morally preferable alternative is available. Additionally,

the results of Experiment 5 confirm our previous finding that individuals’ deontic intuitions are

also susceptible to systematic manipulation of good and bad effects.

60

Experiment 5 also provides additional evidence of framing effects. For example, 80%

(16/20) of participants in the Intentional Homicide condition (in which participants were given

both the Intentional Homicide and Bystander Problems in a within­subject design) and 75%

(15/20) of participants in the Better Alternative condition (in which participants were given both

the Better Alternative and Bystander problems in a within­subject design) judged throwing the

switch in the Bystander Problem to be permissible. Further, only 55% (11/20) of participants in

the Disproportional Death condition (in which participants were given both the Disproportional

Death and Bystander problems in a within­subject design) judged throwing the switch in the

Bystander Problem to be permissible. These percentages, and particular the last figure, contrast

sharply with the 95% of respondents who judged throwing the switch in the Bystander Problem

to be permissible in the between­subject design utilized in Experiment 1. This in turn suggests

that whether individuals are prepared to permit foreseeable battery or homicide on broadly

utilitarian grounds in these circumstances may depend on how that question is framed. However,

because the number of participants in Experiment 5 was again relatively small, we refrain from

drawing any firm conclusions about framing effects at this stage of our inquiry and merely

identify this issue as a topic for future research.

7. Experiment 6

Experiments 1­5 were designed to investigate the moral competence of adults only. In

Experiment 6, we extended this inquiry in a provisional way to include the moral competence of

children ages 8­12. Our objectives in this regard were limited. First, we wished to determine

whether children in this age group had moral intuitions about a pair of relatively simple trolley

problems that were similar to the intuitions of adults. Second, and more generally, we wished to

61

explore the potential of using these and similar thought experiments to investigate the moral

competence of young children. A central premise of both Piaget’s (1932/1965) and Kohlberg’s

(1981, 1984) theory of moral development is that the moral conceptions of adults and children

consist of fundamentally different principles. A corollary is the assumption that moral

development is something that happens gradually over the course of one’s lifetime, and thus

should be investigated by means of longitudinal studies. In this study, we began testing these

assumptions by presenting a group of children with two of the scenarios we used in studying the

moral intuitions of adults. Finally, we wanted to investigate whether these children’s moral

intuitions could be explained with reference to the prohibition of intentional battery and the

principle of double effect. That is, we wished to discover whether children, like adults, would

treat these two cases of necessity differently, depending on whether battery is used as an

intended means to a given end or as a foreseen side effect. In this way, we sought to inquire

whether these principles emerge and become operative relatively early in mental development,

thereby raising the possibility that specific poverty of the stimulus questions could be formulated

in the moral domain.

7.1 Method

7.1.1 Participants

Participants were 30 children ages 8­12 who were recruited with parental consent from

four metropolitan areas: Cambridge, Massachusetts; Knoxville, Tennessee; Toledo, Ohio; and

the District of Columbia. There were 14 girls and 16 boys.

7.1.2 Stimuli and Procedure

62

Two scenarios were used. In the first scenario (“Dr. Brown”), which was modeled on the

Transplant Problem, battery was embedded within an agent’s action plan as a means (henceforth

“Intentional Battery”). In the second scenario (“Charlie”), which was modeled on the Trolley

Problem, battery was embedded within the agent’s action plan as a side effect (henceforth

“Foreseeable Battery”) (see §9 for actual text).

A between subject design was utilized. Each of the 30 children was given a written

questionnaire containing either the Intentional Battery or the Foreseeable Battery scenario. The

child was first instructed to read the scenario and then to decide whether the proposed action it

described was “wrong.” For the purposes of this experiment, we took for granted the standard

assumption in deontic logic that “wrong” is logically equivalent to “not morally permissible”

(Prior, 1955; Von Wright, 1951) and reasoned that the children would have an easier time

answering a question using the term “wrong” than one using the phrase “morally permissible.”

The child was also asked on a separate page to provide an explanation for his or her response. 15

children were given the Intentional Battery scenario and 15 were given the Foreseeable Battery

scenario. Assignment of participant to scenario type was random.

7.2 Results

7.2.1 Judgments

The main results of Experiment 6 are summarized in Figure 22. 6 of 15 children in the

Intentional Battery condition judged the action constituting intentional battery to be permissible.

By contrast, 14 of the 15 children in the Foreseeable Battery condition judged the action

constituting unintentional battery to be permissible. This difference is significant: x 2 (1, N=30) =

9.6, p < 0.01.

63

7.2.2 Justifications

Due to the limited number of participants in this experiment, we refrain from analyzing

participants’ responses by gender. 15 Instead, we turn to children’s expressed justifications for

their judgments. These justifications were categorized according to the same coding scheme

used in our previous experiments. 46.7% (14/30) of participants gave no justification, 13.3%

(4/30) provided logically inadequate justifications, while 40% (12/30) provided logically

adequate justifications.

7.3 Discussion

Although the results of Experiment 6 are limited, they constitute at least some initial

evidence that the moral competence of 8­12 year old children includes the prohibition of

0

2

4

6

8

10

12

14

Intentional Battery

Foreseeable Battery

Permiss. Impermiss.

Act Type

Subjects

Figure 22: Moral Judgments of Two Act Types in Experiment 5 (Intentional Battery vs. Foreseeable Battery)

X 2 (1, N=30) = 9.6, p < .01

64

intentional battery and the principle of double effect. Put differently, they suggest that simple

deontological or consequentialist principles alone may be inadequate to describe the intuitive

moral knowledge of children ages 8­12. More generally, these results support the efficacy of

using trolley problems to investigate the moral competence of children, not only those of this age

group but possibly even much younger populations. This conclusion is significant in part

because trolley problems are qualitatively more complex than the questions used by researchers

in the Piagetian tradition (see §8.2.1).

Turning to adequacy of justifications, it is notable that the justifications offered by

children in Experiment 6 were only marginally less adequate than the corresponding

justifications offered by adults in Experiment 1 (which also relied upon relatively simple trolley

problems presented in a between­subject design). In particular, the percentage of children who

provided logically adequate justifications (40% or 12/30) compares favorably with the

percentage of adults (50% or 20/40) of adults who did so in Experiment 1. Nevertheless, as was

the case with adults, many of these logically adequate justifications were manifestly incapable of

accounting for the divergent pattern of intuitions elicited in this experiment, in which saving five

people at the cost of killing one person is felt to be permissible in one case but not the other.

Hence Experiment 6 provides further evidence that, like adults, children ages 8­12 possess

unconscious moral knowledge that may be largely inaccessible to deliberate introspection.

Two further tentative conclusions may be drawn from Experiment 6. First, together with

the results of our previous experiments, the results of Experiment 6 imply that at least some of

the operative principles of moral competence, such as the distinction between intended means

and foreseen side effect, are invariant throughout the course of moral development, at least

between ages 8­65. This conclusion runs counter to one of the most basic assumptions of both

65

Piaget’s (1932/1965) and Kohlberg’s (1981, 1984) approach to moral development, according to

which the adult’s and child’s moral competence are comprised of fundamentally different

principles. A corollary of this assumption is the view that moral development is something that

happens gradually over the course of one’s lifetime, and thus should be investigated by means of

longitudinal studies. The findings of Experiment 6 call at least some aspects of this investigative

procedure into question, raising the possibility that, like language, vision, and other cognitive

systems, moral development involves pre­determined critical stages, after which moral

competence more or less stabilizes.

Finally, the results of Experiment 6 also suggest, at least tentatively, that it may be

possible to formulate poverty of the stimulus arguments in the moral domain. While this

possibility is theoretically intriguing, we refrain from drawing any firm conclusions about it here.

Instead, we simply note that more experimental work must be done to determine whether certain

complex moral principles, such as those investigated here, are explicitly taught or otherwise

available to the developing child during the acquisition process (Harman, 2000; Mikhail, 2000).

Offhand, this seems unlikely, particularly in light of the discovery that adults do not explain or

justify their own moral intuitions with reference to these principles. Indeed, the fact that at least

some operative moral principles appear to be non­introspectible makes it plausible to suppose

that these principles are not taught to successive generations explicitly. Hence we may

reasonably assume, as a working hypothesis, that they are the developmental consequences of an

innate cognitive faculty (Mikhail, 2000; Mikhail et al., 1998). However, this assumption is

largely speculative and the issue requires more empirical investigation.

66

8. General Discussion

Taken together, the studies presented here constitute significant evidence that adults

possess intuitive or unconscious knowledge of complex moral principles, including the

prohibition of intentional battery and the principle of double effect. Additionally, Experiment 6

provides some evidence for inferring that the same may be true of ages 8 to 12. By attributing

this knowledge to experimental subjects, we can predict and explain their moral intuitions.

Because the imputed knowledge is intuitive and not fully open to conscious introspection, we can

also advance a tentative explanation of why relatively few individuals appear capable of

providing logically adequate justifications of their judgments, even on an extremely permissive

interpretation of what counts as logically adequate, and why virtually no individuals appear

capable of providing observationally adequate justifications, that is, justifications from which the

systematic pattern of intuitions elicited by these experiments can be mechanically derived (see

§1.1). The explanation, simply put, is that moral cognition appears to involve unconscious

computation, that is, mental operations which are not consciously accessible. In this respect,

moral cognition may be compared to other cognitive capacities, such as language, vision, object

perception, and face recognition, all of which also involve unconscious computation. Human

language, in particular, is well understood to depend on “mental processes that are far beyond the

level of actual or even potential consciousness” (Chomsky, 1965, p. 8). Hence the discrepancy

between moral judgments and justifications observed here is in keeping with the non­transparent

character of mental activity generally. Within the expository framework we have adopted, the

phenomena can be explained rather easily with reference to the analogy to linguistic competence:

Just as normal persons are typically unaware of the principles guiding their moral linguistic

intuitions, so too are they often unaware of the principles guiding their moral intuitions. 16

67

In §8.2, we elaborate on these findings and place them within a broader context by

comparing and contrasting our approach to the theory of moral cognition with those of Piaget

(1932/1965), Kohlberg (1981), and Greene et al. (2001). Before doing so, however, we turn

directly in §8.1 to a more extensive discussion of the unconscious mental representations that

appear to be triggered by thought experiments like the trolley problems.

8.1 Describing the Operative Principles

In Figure 2 (see §1.2), we sketched an expanded perceptual model for moral judgment,

according to which permissibility judgments do not necessarily depend on the surface structure

of an action­description, but on how that action is mentally represented. The main theoretical

problem within this framework is an information­processing problem, namely: How do people

manage to compute a full structural description of the action that incorporates certain properties,

such as ends, means, side effects, and prima facie wrongs like battery, when the stimulus

contains no direct evidence for these properties? This is similar in principle to determining how

people manage to extract a three­dimensional representation from a two­dimensional stimulus in

the theory of vision (e.g., Marr, 1982), or to determining how people to recognize the word

boundaries in an undifferentiated auditory stimulus in the theory of language (e.g., Chomsky &

Halle, 1968). In our case, the question is how and why individuals make the inferences they do

about the various agents and actions in our examples, even when we deliberately deprive them of

direct evidence of those agents’ mental states and other morally salient properties.

As Figure 23 depicts, this problem may be divided into at least four main parts.

Presumably, to compute a morally cognizable structural description of a given action, one must

68

? PERMISSIBLE

IMPERMISSIBLE

STRUCTURAL DESCRIPTION

Perceptual Response: Moral Judgment

Unconscious Mental Representation

INPUT ?

Conversion Rules

Stimulus: Fact Pattern

Deontic Rules

Figure 2: Expanded Perceptual Model for Moral Judgment

69

?

Intentional Structure

“Conversion Rules”

?

Temporal Structure

Stimulus: Fact Pattern

Moral Structure

? ? INPUT

Causal Structure

Fig. 23: Components of Conversion from Stimulus to Structural Description

70

generate a complex mental representation of that action which encodes relevant information

about its temporal properties, its causal properties, its moral properties, and its intentional

properties. But how does the individual manage to extract the relevant cues from the stimulus

and convert what is given into a full structural description? The following is one hypothesis

(Mikhail, 2000).

8.1.1 Temporal Structure

The process appears to include the following steps. First, one must identify the morally

relevant action­descriptions contained in the stimulus and order them serially according to their

relative temporal properties. For example, in the Bystander Problem, one must recognize that

“Hank’s seeing what happened” occurs before “Hank’s throwing the switch,” which occurs

before “Hank’s killing the man” (Figure 24).

Fig. 24: Temporal Order of Three Act­Representations in the Bystander Problem

t (­m) t (0) t (+n)

[Hank’s seeing what happened] [Hank’s throwing the switch] [Hank’s killing the man]

71

There is an important convention that this time­line incorporates, which is to date an

action from its time of completion. An act that begins at t (0) and ends at t (+n) is in a sense

performed neither at t (0), nor at t (+n) , but in that period of time bounded by them. For present

purposes, we simplify this situation by following traditional jurisprudence in locating the time of

an action according to when it is completed (Salmond, 1902/1966).

8.1.2 Causal Structure

Second, one must interpret the morally relevant action­descriptions contained in the

stimulus in terms of their basic causal and other semantic properties. For example, one must

identify and interpret causative expressions such as “Hank killed the man,” “Hank prevented the

train from killing the men,” and “Hank let the men die” in terms of their underlying semantic

structures. Figure 25 illustrates both the surface and semantic structures of “Hank killed the

man.” Figures 26 and 27 do the same for the semantic structures of “Hank prevented the train

from killing the men” and “Hank let the men die,” respectively.

S S

NP VP Agent

N V NP Cause Effect

Det N Patient Event

Hank killed the man (person) (death)

Fig. 25: Surface and Semantic Structures of “Hank killed the man”

72

S1 Agent

Cause Effect

Neg S2 Agent

Cause Effect

Patient Event

(person) (death)

Fig. 26: Semantic Structure of “Hank prevented the train from killing the men”

S1 Agent

Cause Effect

Neg S2 Agent

Cause Effect

Neg S3 Agent

Cause Effect

Patient Event

(person) (death)

Fig. 27: Semantic Structure of “Hank let the men die”

73

Additionally, presumably by relying in part on temporal information (Kant, 1787/1953),

one must compute a representation of the causal structure of the relevant acts and omissions in

the form of a causal chain or ordered sequence of causes and effects. Figure 28 illustrates one of

three such causal chains at issue in the Bystander Problem, namely, the chain linking the agent’s

throwing the switch to the outcome of killing the man. At least two further causal chains must

be generated, the first linking the same action (“throwing the switch”) to the outcome of

preventing the train from killing the men, and the second connecting the forbearance of this

action (“not throwing the switch”) to the outcome of letting the men die. (In Figure 28, the effect

of causing the train to hit the men is placed in brackets to signify that this representation, unlike

the others, is not derived directly from an action­description contained in the stimulus, but rather

must be inferred from assumptions about how objects interact with one another, presumably in

accord with certain core knowledge of contact mechanics (Carey & Spelke, 1994; Spelke,

Breinlinger & Jacobson, 1992). In other words, the brackets identify one location in the causal

chain where the mind “supplies the missing information” (Pinker, 1997, p. 28) that killing the

man in these circumstances requires causing the train to come into contact with him.)

Fig. 28: Causal Chain Generated by Throwing Switch in Bystander Problem

(thing)

Patient

(move)

Event

Effect Cause Agent Cause Effect Cause Cause Effect

(thing)

Patient Event

(move) (thing)

Effect

Event Event Patient Patient

(contact) (person) (death)

switch Hank throw train turn man kill [train] [hit]

74

8.1.3 Moral Structure

Third, one must apply the moral principle “death is bad” to the semantic structures in

Figures 25­27, transforming each of those structures into a new one that encodes a bad effect.

Figure 29 illustrates this operation in the case of the semantic structure of “Hank killed the man.”

S S

Agent Agent

Cause Effect → Cause Bad Effect

Patient Event Patient Event

(person) (death) (person) (death)

Fig. 29: Moral Transformation of “Hank killed the man”

This operation states that an effect which consists of the death of a person is a bad effect, and

may be rewritten as such. Likewise, one must apply the logical principle “preventing a bad

effect is good” to the morally transformed structure of sentences like “Hank prevented the train

from killing the men,” thereby converting that structure into one that encodes a good effect.

S1 S1

Agent Agent

Cause Effect → Cause Good Effect

Neg S2 Neg S2

Agent Agent

Cause Bad Effect Cause Bad Effect

Fig. 30: Conversion of Transform of “Hank prevented the train from killing the man”

75

This operation states that an effect which consists of the negation of a bad effect is a good effect,

and may be rewritten as such (Figure 30).

Additionally, one must apply the logical principle “failing to prevent a bad effect is bad”

to the morally transformed structure of sentences like “Hank let the men die,” thereby converting

that structure into another one that encodes a second, overarching bad effect:

S1 S1 Agent Agent

Cause Effect Cause Bad Effect

Neg S2 → Neg S2 Agent Agent

Cause Good Effect Cause Good Effect

Neg S3 Neg S3 Agent Agent

Cause Bad Effect Cause Bad Effect

Fig. 31: Conversion of Transform of “Hank let the men die”

This operation states that an effect which consists of the negation of a good effect is a bad effect,

and may be rewritten as such (Figure 31).

8.1.4 Intentional Structure

Fourth, one must apply what may be called a “presumption of good intentions” to the

representational structures which have been generated up to this point, thereby converting them

into new structures that represent the intentional properties of the given action. That is, taking

the representation of an action with both good and bad effects as input, one must (absent

countervailing information of the type presented in the Intentional Homicide Problem in §6)

generate the intentional structure of the action by identifying the good effect as the intended

76

effect or goal of the action and the bad effect as the foreseen but unintended side effect. This

operation can also be represented graphically (Mikhail, 2000; see also Donagan, 1977; Goldman,

1970). In Figure 32, for example, the tree diagram on the left represents one part of the

underlying structure, not of a sentence, but of an action with both good and bad effects that we

may presume the mind spontaneously constructs upon encountering thought experiments like the

trolley problems. The arrow in Figure 32 indicates that this act tree must be converted into the

tree diagram on the right, which in turn signifies that the intended outcome of S’s action, her

ultimate aim or goal, is to achieve the good effect (as represented by the vertical chain of means

and ends connecting the base of the tree to its final end), whereas the bad effect is an unintended

side effect.

[Bad Effect] [Good Effect] [Bad Effect] [Good Effect]

→

[S’s V­ing at t (α) ] C [S’s V­ing at t (α) ] C

Fig. 32: Generation of Intentional Structure of Act with Good and Bad Effects

Note that some operation of this general character must be postulated to explain how the

mind/brain computes intended and non­intended effects, since—crucially—there is no mental

state information in the stimulus itself. In the operation depicted here, the presumption of good

intentions acts as a default principle which says, in effect, that unless evidence to the contrary is

presented, the decision­maker is to assume that the agent, S, is a person of good will, who

intends and pursues good effects and avoids bad ones. 17 In particular, the presumption directs

77

one to assume that S intends to prevent the train from killing the men and regards killing the man

as an unintended side effect (Figure 33).

[killing the man] [preventing the train [killing the man] [preventing the train from killing the men] from killing the men]

→

[throwing the switch] [throwing the switch]

Fig. 33: Generation of Intentional Structure in Trolley Problems

8.1.5 Additional Moral/Deontic Structure

The act­representations conceptualized thus far are complex in that they represent the

given actions in terms of morally salient properties like ends, means, and side effects, even

though the stimulus contains no direct evidence of these properties. Nevertheless, these

representations are necessary but not sufficient to distinguish the trolley problems in accord with

their corresponding moral intuitions. The tree structure depicted on the right side of Figure 33,

for example, applies equally to both of the looped­track scenarios used in Experiment 4, and the

general distinction it captures, between the goal of preventing the death of five persons and the

side effect of killing one person, is a property shared by most of the trolley problems we have

discussed so far. Hence these intuitions can be adequately explained only by attributing some

additional moral or deontic structure to the representations these problems elicit.

What additional structure is needed? The specific hypothesis we have advanced here is

that the relevant actions must be represented in terms of prima facie wrongs, such as battery or

homicide. For example, in the Footbridge Problem, one must derive a representation of battery

78

by inferring that (1) the agent must touch the man in order to throw him onto the track in the path

of the train, and (2) the man would not consent to being touched in this manner, because of his

desire for self­preservation (and because no evidence is given to the contrary). Utilizing

standard notation in action theory (e.g., Mikhail, 2000; Goldman, 1970) and deductive logic

(e.g., Leblanc & Wisdom, 1993), this line of reasoning can be formalized in the following

derivation (Figure 34):

1. [Ian’s throwing the man at t (0) ] C ________________ 2. [Ian’s throwing the man at t (0) ] Given 3. [Ian throws the man at t (0) ] 2; Linguistic Transformation 4. [Ian throws the man at t (0) ] ⊃ [Ian touches the man at t (0) ] Analytic 5. [Ian touches the man at t (0) ] 3, 4; Modus Ponens 6. [The man has not expressly consented to be touched at t (0) ] Given 7. [Ian throws the man at t (0) ] ⊃ [Ian kills the man at t (+n) ] Given 8. [[Ian throws the man at t (0) ] ⊃ [Ian kills the man at t (+n) ]] ⊃

[the man has not implicitly consented to be touched at t (0) ] . [the man would not consent to being touched at t (0) , if asked] Self­Preservation Principle 18

9. [the man has not implicitly consented to be touched at t (0) ] 7,8; Modus Ponens 10. [the man would not consent to be touched at t (0) , if asked] 7,8; Modus Ponens 11. [Ian touches the man without his express, implicit, or hypothetical 4,5,9,10

consent at t (0) ] 12. [Ian touches the man without his express, implicit, or hypothetical

consent at t (0) ] ⊃ [Ian commits battery at t (0) ] Definition 13. [Ian commits battery at t (0) ] 11,12; Modus Ponens 14. [Ian’s committing battery at t (0) ] Linguistic Transformation

Fig. 34: Derivation of Representation of Battery in the Footbridge Problem

Additionally, one must locate the representation of these prima facie wrongs which are

derived in this manner in the correct temporal, causal, and intentional location in one’s act tree,

thereby identifying whether they are an intended and/or prior means to a given end, or a

subsequent and/or foreseen side effect. For example, one must locate the batteries committed in

the Footbridge and Bystander Problems in the manner depicted in Figures 3­4, but not in Figures

35­36.

79

D’s throwing the switch at t (0)

D’s turning the train at t (+n)

D’s preventing the train from killing the men at t (+n) D’s causing the train

to hit the man at t (+n+o)

D’s committing battery at t (+n+o) End

Fig. 4: Mental Representation of Bystander Problem

Side Effects

D’s killing the man at t (+n+o+p)

D’s throwing the man at t (0)

D’s committing battery at t (0)

D’s preventing the train from killing the men at t (+n+o)

D’s killing the man at t (+n+p)

D’s causing the train to hit the man at t (+n)

D’s committing battery at t (+n)

Fig. 3: Mental Representation of Footbridge Problem

Side Effects

End

Means

80

D’s throwing the switch at t (0)

D’s turning the train at t (+n)

D’s committing battery t (0)

D’s preventing the train from killing the men at t (+n)

D’s causing the train to hit the man at t (+n+o)

Fig. 36: Inaccurate Representation of Bystander Problem

Side Effects

D’s killing the man at t (+n+o+p)

D’s committing battery at t (+n+p)

D’s throwing the man at t (0)

D’s committing battery at t (+n+o)

D’s killing the man at t (+n+p)

D’s causing the train to hit the man at t (+n)

D’s preventing the train from killing the men at t (+n+o)

Fig. 35: Inaccurate Representation of Footbridge Problem

Side Effects

End

Means

End

Means

81

Finally, once one has generated a full structural description of the agent’s action that

encodes information about its temporal, causal, moral, and intentional properties, as well as the

location of its prima facie wrongs such as battery, one must apply the relevant substantive moral

principles or deontic rules (e.g., the prohibition of intentional battery and the principle of double

effect) to that structural description.

While the foregoing account is incomplete, a complete theory of the steps converting

proximal stimulus to perceptual response by means of a morally cognizable structural description

can be given along these lines. In principle, a computer program could be written that could

execute these operations from start to finish. The theory presented here thus goes some way

toward achieving the first two of Marr’s (1982) three levels at which any perceptual information

processing task may be understood, namely, the computational level and the representation and

algorithm level. In this sense, the theory is arguably a major step forward in terms of satisfying

the demands of both observational and descriptive adequacy.

8.2 Contrast with Alternative Frameworks

8.2.1 Contrast with Piaget’s Framework: Complex vs. Simple Acts

From the point of view of the theory of moral cognition and moral development, the

thought experiments tested in these investigations are fascinating for several reasons. Perhaps

the most important reason is that they are qualitatively more complex than the stimulus materials

used by researchers working within the Piagetian tradition. As we mentioned in §1, in his

original studies Piaget (1932/1965) sought to address the problem of descriptive adequacy by

focusing attention on the “subjective” and “objective” elements of moral judgment. To this end,

he asked children to compare an action characterized by “good intentions” (e.g., helping mom set

the table) whose negative consequences were significant (e.g., breaking fifteen cups) with an

action characterized by “bad intentions” (e.g., taking a cookie from the cookie jar) whose

82

negative consequences were slight (e.g., breaking one cup), and to determine which agent was

“naughtier.” Piaget discovered that until around age nine, children tended to judge the agent

whose action is characterized by the “good intention/greater negative consequences”

combination to be “naughtier.” After age nine, this response pattern changed: the older children

tended to judge that the agent whose action is characterized by the “bad intention/slight negative

consequence” combination was “naughtier.” It was on this basis that Piaget concluded that

children based their moral judgments on effects, not intentions, until around age nine.

As subsequent researchers have noted, Piaget’s conclusions were unjustified, because he

used stories that co­varied two and sometimes three parameters at once. The exact role played

by intentions and consequences in the children’s judgments was therefore impossible to

determine. To remedy this situation, researchers modified Piaget’s procedure by presenting

children with stories which permitted the independent variation of intention and consequence

parameters along two dimensions, “good” and “bad.” In this way, subjects could be presented

with actions characterized by one of four possible combinations of features: (1) “good

intention/good effect,” (2) “good intention/bad effect,” (3) “bad intention/good effect,” and (4)

“bad intention/bad effect.”

In one of Nelson’s (1980) experiments, for example, children were given four different

stories to evaluate. In the first, a boy, who sees his friend is sad because he does not have

anything to play with, throws a ball toward his friend in order to play catch with him and cheer

him up (good intention). The friend catches the ball and is happy (good effect). In the second,

the boy throws the ball with the same intention (good intention) but ends up hitting his friend on

the head and making him cry (bad effect). In the third, the boy is mad at his friend that day and

throws the ball toward him in order to hit him with it (bad intention). However, the friend

83

catches the ball and is happy (good effect). Finally, in the fourth, the boy again throws the ball at

his friend in order to hit him with it (bad intention). This time he succeeds in hitting his friend

on the head with the ball and making him cry (bad effect). When Nelson presented these four

stories to her experimental subjects, she discovered that children as young as three years of age

utilized information about intentions when making moral judgments (Nelson, 1980).

Table 5. Moral Judgments Elicited by Trolley, Transplant, Bystander, and Footbridge Problems as a Function of Piaget’s Variables Problem (Agent) Good Effect Bad Effect Good Intention

(Ultimate Aim or Goal) Deontic Status

Trolley (Charlie) Preventing 5 deaths 1 Death Preventing 5 deaths Permissible Transplant (Dr.Brown) Preventing 5 deaths 1 Death Preventing 5 deaths Impermissible Bystander (Denise) Preventing 5 deaths 1 Death Preventing 5 deaths Permissible Footbridge (Nancy) Preventing 5 deaths 1 Death Preventing 5 deaths Impermissible Bystander (Hank) Preventing 5 deaths 1 Death Preventing 5 deaths Permissible Footbridge (Ian) Preventing 5 deaths 1 Death Preventing 5 deaths Impermissible

Although experiments such as Nelson’s, which permitted the independent variation of

intentions and effects, were an improvement on Piaget’s original studies, they nonetheless

limited their attention to what we might call “morally simple acts,” i.e., acts whose mental

representations are characterized by only one morally salient (good or bad) ultimate intention

(i.e., purpose or goal) and only one morally salient (good or bad) effect. By contrast, the

examples we investigate here are more complex, because they involve act­representations which

are comprised of multiple intentions and which generate both good and bad effects. Moreover, as

Table 5 reveals, three of the four variables in the Piagetian framework – good and bad effects

and good intention (i.e., ultimate aim, purpose or goal) – are held constant in the Trolley,

Transplant, Footbridge, and Bystander Problems, while the fourth – bad intention – is not

relevant. This suggests that some other property or properties are responsible for the divergent

moral judgments generated by these examples.

84

The key insight of our paper is that these intuitions can be adequately explained only by

drawing on complex moral and legal principles, such as the prohibition of intentional battery and

the principle of double effect. Piaget emphasized what he called the “juridical complexity” of

children’s moral judgments and observed that, to understand those judgments, psychologists

must familiarize themselves with “the common law” (Piaget, 1932/1965, pp. 20, 13). Yet

Piaget’s own stimulus materials often failed to test for common legal distinctions, such as those

between justification and excuse, recklessness and negligence, proximate and remote causation,

or—what appears most relevant here—intended and foreseen effects.

Piaget may not be alone in this regard. As Robinson and Darley (1995) have observed,

many social scientists seeking to describe social norms and “the community’s sense of justice”

appear to rely on definitions and concepts that are both descriptively inadequate and legally

inaccurate. As a result, they often underestimate the subtlety and complexity of laypersons’

intuitive grasp of legal concepts and distinctions. And they often beg important questions about

how moral knowledge is acquired. For example, in a recent wide­ranging law review article

criticizing the role of moral intuitions in legal analysis, Kaplow and Shavell insist that norms

“must be relatively simple because they must be imparted to children and applied without

sustained analysis” (Kaplow & Shavell, 2001, p. 1033). Kaplow and Shavell’s observation begs

the question whether and to what extent norms are “imparted to children” in any meaningful

sense. If morality is like language, some of the principles generating our moral intuitions may

turn out to be innate; indeed, as some commentators have argued, the principle of double effect

may turn out to be one such principle (Harman, 2000). In any case, we would simply observe

that it seems highly questionable for norms theorists like Kaplow and Shavell to invert the

logical relationship of descriptive and explanatory adequacy in this fashion. The success of

85

research programs in other cognitive domains suggests that the origin of our moral intuitions can

be adequately investigated only insofar as their structure is clearly understood.

8.2.2 Contrast with Kohlberg’s Framework: Operative vs. Express Principles

Many research programs in moral cognition do not distinguish sharply between moral

judgment and moral reasoning and tacitly assume that moral principles are introspectible.

Perhaps the most influential research program of this type is the one developed by Lawrence

Kohlberg (1958, 1981, 1984). On Kohlberg’s view, Piaget was correct to assume that the

capacity for moral judgment passes through a series of developmental stages. However, whereas

Piaget proposed two broad stages, Kohlberg’s theory postulates six. Further, although

Kohlberg’s theory is partially based on Piaget’s, his conclusions are far better supported

empirically.

The aspect of Kohlberg’s framework most relevant to our discussion concerns his focus

on the explicit statements people make to justify or explain their moral judgments. Kohlberg

assessed moral development by having trained researchers code an experimental subject’s stated

justifications for her decisions on a series of moral dilemmas. One of his best known puzzles is

the “Heinz dilemma,” which deals with the example of a man whose wife’s life can be saved

only by a medicine he cannot afford. Under these circumstances, would it be right for the man to

steal the drug? Kohlberg and his associates put these and similar questions to children and adults

of all ages, asking them to justify whatever decision they reached. By evaluating not the

decision itself but the justification accompanying it, Kohlberg (1981, 1984) claimed to discover

that moral development progresses through an unvarying sequence of six stages.

86

Kohlberg’s overall theory is rich and complex, and we make no effort to evaluate it

systematically here. Instead, we simply note that our investigations appear to confirm in rather

dramatic fashion that those commentators who have questioned Kohlberg’s methodological

decision to focus on verbalized justifications rather than moral intuitions themselves are correct

to criticize this aspect of his framework (e.g., Darley & Shultz, 1990; Haidt, 2001; Kagan, 1987;

MacNamara, 1990, 1999; Mikhail, 2000).

Unlike Kohlberg, we distinguished at the outset of our investigations between operative

moral principles and express principles (§1.1). We also assumed that just as normal persons are

typically unaware of the principles guiding their linguistic intuitions, so too are they often unaware

of the principles guiding their moral intuitions. Further, based on both informal observation and a

review of the relevant literature, we predicted that an empirical investigation of the trolley

problems would reveal that our subjects’ “locus of moral certitude” (Jonsen & Toulmin, 1988)

would be their intuitive judgments themselves, not their underlying justification.

Our studies confirmed this prediction. When the participants in our experiments were

asked to provide verbal rationales for their judgments, they were consistently incapable of

articulating the operative principles on which their judgments were based (Table 6). Indeed, in

sharp contrast with Kohlberg’s (1981, 1984) findings concerning the relevance of demographic

variables like gender, nationality, age, and education, but in accord with our expectations, our

subjects’ moral judgments were widely shared, irrespective of these factors. But, as quotations

like these reveal, our subjects’ expressed principles were widely divergent. More importantly,

they consistently failed to state the operative reasons for their judgments in any theoretically

compelling sense. They often said things that were incompatible with their own judgments, or

even internally incoherent. And they often appeared puzzled by the nature and strength of their

87

intuitions, and by the way those intuitions shifted when we introduced small changes in the

wording of the action­sequences we gave them, in order to evoke distinct mental representations.

Table 6: Adequacy of Justifications in Experiments 1­2 and 4­6 Experiment and Design

Degree of Logical Adequacy Number Responses

Percentage Responses

No Justification 13 32.5 Inadequate Justification 7 17.5

Experiment 1: Between Subject (n=40) Adequate Justification 20 50.0

No Justification 23 35.4 Inadequate Justification 25 38.5

Experiment 2: Within Subject (n=65) Adequate Justification 13 20.0

No Justification 260 84.1 Inadequate Justification 19 6.1

Experiment 4: Between Subject (n=309) Adequate Justification 30 9.7

No Justification 32 53.3 Inadequate Justification 14 23.3

Experiment 5: Within Subject (n=60) Adequate Justification 14 23.3

No Justification 14 46.7 Inadequate Justification 4 13.3

Experiment 6: Between Subject (n=30) Adequate Justification 12 40.0

A related point worth highlighting concerns a question of method underlying Kohlberg’s

controversial findings about the role of gender and other demographic variables in moral

development. As critics such as Gilligan (1982) have observed, the original research from which

Kohlberg derived his theory was based on a study of 84 American boys from suburban Chicago

whose development Kohlberg followed for a period of over twenty years. Although Kohlberg

claimed universality for his stage theory, subsequent research revealed that girls and other groups

who were not included in Kohlberg’s original sample tended to reach only the third or fourth

stage of his 6­stage sequence (Gilligan, 1982; see also Edwards, 1975; Holstein, 1976; Simpson,

88

1974). Critics like Gilligan have criticized these findings as the inevitable outcome of a research

program which is insensitive to the fact that, in moral matters, men and women often speak “in a

different voice” (Gilligan, 1982). Generally speaking, however, these critics have not strayed far

from Kohlberg’s paradigm and have continued to “measure” or otherwise evaluate the character

of an individual’s moral judgment by looking to her actual utterances.

We think the dissociation between moral judgments and justifications identified here

calls into question the entire approach of both Kohlberg and many of his critics. Simply put, our

studies suggest that these psychologists may have focused their attention on the wrong

phenomena. A comparison to the study of language and vision, neither of which constructs a

theory of development based on expressed justifications, brings the point into sharper focus.

Neither linguists nor vision theorists take the post hoc explanations of experimental subjects—

for example, statements explaining or justifying the intuition that a particular utterance is

ungrammatical—to be their primary source of data. Rather, their data is the subjects’ intuitions

themselves.

While our findings clearly establish a significant discrepancy between judgments and

justifications, it would be a mistake to conclude from what has been said thus far that express

principles have no evidentiary role to play in the theory of moral cognition. On the contrary, an

individual’s introspective reports may often provide important and even compelling evidence for

the properties of her moral competence (cf., Chomsky, 1965). Indeed, many of the justifications

offered by adults and even some of the justifications offered by young children in our

experiments were illuminating in that they revealed an intuitive grasp of a specific legal concept.

For example, several of the children who were given the Transplant Problem and asked to

explain their judgments in Experiment 6 referred on their own initiative to the crucial issue of

89

lack of consent. One participant, an eight year­old boy, who judged Dr. Brown’s cutting up the

patient to be wrong made the following comment:

“Okay, if given consent”

A second participant, an eleven year­old boy, said the following:

“I think that Dr. Brown should ask the person in Room 306 if they would like to be cut up to save the other peoples’ lives.”

A third participant, also an eleven year­old boy, said:

“I said no because it never said that she gave permission to kill her; to give away her body parts…I did not feel good about it because I would not like somebody to take my body parts.”

Finally, a fourth participant, a twelve year­old girl, said:

“I believe that it would be wrong to cut this 306 person up without them even knowing it. It would be different if Dr. Brown had asked this person if they would donate their organs and he had received their permission. That is why I would blame him if he took their life.”

As these remarks suggest, the distinction between operative and express principles does not

imply that researchers should disregard or discount the significance of verbalized justifications.

Although none of these children used the term “battery” to explain their judgments, their

comments clearly suggest an intuitive appreciation for one of the key elements of battery,

namely, lack of consent.

In sum, the distinction between operative and express principles appears to vitiate, or at

least seriously compromise, the Kohlberg paradigm, which arguably was the dominant approach

to the psychology of moral development in the second half of the twentieth century (e.g., Gilligan,

1982; Kohlberg, 1981, 1984; Rest, 1983; Turiel, 1983). Kohlberg’s stage theory of moral

development is one which primarily tracks the development of a person’s ability to express or

articulate moral principles. While this is an important skill, and perhaps corresponds with the

90

ability to engage in more complex acts of moral reasoning, it does not necessarily reveal the

properties of moral competence. On the contrary, as our subjects’ responses to the trolley

problems reveal, a person’s introspective verbal reports and general viewpoint about her moral

knowledge may sometimes be in error. Yet, at the same time, a subject’s reports may contain

important evidence of the speaker’s intuitive knowledge of a specific legal concept, knowledge

that runs the risk of being overlooked if the researcher is preoccupied with the search for

articulate justifications in the manner prescribed by Kohlberg. In the final analysis, the

important point is that, as is the case with a theory of language or vision, the goal of the theory of

moral cognition must be to explain an ideal observer’s actual moral intuitions and their

underlying cognitive mechanisms, rather than to account for an individual’s own statements,

explanations, or justifications of what she intuits and why. (For similar remarks on this

inadequacy of Kohlberg’s framework, see generally Darley & Shultz, 1990; Haidt, 2001; Kagan,

1987; MacNamara, 1990, 1999; Mikhail, 2000. For parallel remarks with respect to language,

see Chomsky, 1965).

8.2.3 Contrast with the Framework of Greene et al.: Computation vs. Emotion

A recent paper relying on trolley problems to investigate moral cognition which was

published after the studies presented here were largely concluded and which has received

widespread attention is Greene et al. (2001). In this section, we briefly comment on this paper

and contrast it with the investigation presented here.

To begin with, we note that the authors have written a highly stimulating paper. They

deserve credit for showing how a problem that has preoccupied philosophers can be studied

using the methods of brain imaging. However, the authors’ conclusion that the moral judgments

91

they examine are caused by differences in emotional engagement seems decidedly premature.

Their study appears to have serious methodological flaws which suggest this conclusion may be

suspect. 19 Additionally, Greene and colleagues have given insufficient consideration to the

competing hypothesis that their moral dilemmas elicit mental representations which differ in

their structural properties. Put simply, a computational theory of moral cognition has been ruled

out too soon.

The authors’ central thesis is that “the crucial difference between the trolley dilemma and

the footbridge dilemma lies in the latter’s tendency to engage people’s emotions in a way that the

former does not. The thought of pushing someone to his death is, we propose, more emotionally

salient than the thought of hitting a switch that will cause a trolley to produce similar

consequences, and it is this emotional response that accounts for people’s tendency to treat these

cases differently” (Greene et al., 2001, p. 2106). They advance a further generalization: “Some

moral dilemmas (those relevantly similar to the footbridge dilemma) engage emotional

processing to a greater extent than others (those relevantly similar to the trolley dilemma), and

these differences in emotional engagement affect people’s judgments” (Greene et al., 2001, p.

2106). Finally, on the basis of these observations, the authors predict and then confirm that

certain emotional centers of the brain are more active when subjects respond to the Footbridge

Problem than when they respond to the Trolley Problem.

These claims prompt three related observations. First, the authors’ data do not exclude

the possibility that the Footbridge and Trolley problems and related thought experiments engage

perceptual and cognitive processing in systematically different ways, and that it is these

differences, rather than (or in addition to) differences in emotion, that influence people’s moral

judgments. Rather, their data are consistent with assuming that people distinguish permissible

92

and impermissible actions for independent reasons and respond emotionally once these prior

determinations have been made.

Second, on the authors’ own view, some account of the process whereby subjects

interpret the verbal stimulus and extract informational cues is not merely possible, but necessary.

Indeed, Greene and colleagues presuppose just such an account, suggesting that people manage

to conclude that it is acceptable to sacrifice one person for the sake of five in the Trolley Problem

but not the Footbridge Problem by spontaneously analyzing these cases in terms of three

features: “whether the action in question (a) could reasonably be expected to lead to serious

bodily harm, (b) to a particular person or a member or members of a particular group of people

(c) where this harm is not the result of deflecting an existing threat onto a different party”

(Greene et al., 2001, p. 2107). Greene and colleagues explain their subjects’ moral judgments

and predict patterns of brain activity on the basis of these three features.

Third, the authors’ characterization of the function that maps verbal stimulus to moral

response is neither complete nor descriptively adequate. It is incomplete because we are not told

how people manage to recognize whether a given dilemma contains these features; surprisingly,

Greene and colleagues leave this first step in the perceptual process unanalyzed. More

importantly, the authors’ account is descriptively inadequate because it cannot explain simple but

compelling data to which a theory of moral cognition must answer.

Consider, for example, two marginal variations of the Footbridge and Bystander

problems, which we investigated in Experiments 2­3 and Experiment 5, respectively. In the first,

which we labeled the Consensual Contact Problem (see §3.1.2), a runaway trolley threatens to

kill a man walking across the tracks. The only way to save him is to throw him out of the path of

the train. Throwing the man, however, will seriously injure him. In the second, which we

93

labeled the Disproportional Death Problem (see §6.1.2), the same runaway trolley again

threatens to kill the man walking across the tracks. This time, the only way to save the man is to

throw a switch that will turn the trolley onto a side track, where it will kill five people.

Taken together, these two problems create a puzzle for Greene et al. (2001). Throwing

the man out of the path of the train is an action which “could reasonably be expected to lead to

serious bodily harm to a particular person . . . where this harm is not the result of deflecting an

existing threat onto a different party” (Greene et al., 2001, p. 2107). On the authors’ account,

therefore, it should be assigned to their “moral­personal” category and judged impermissible.

But a combined total of 93% (38/41) of participants in Experiments 2­3 thought this action was

permissible. Conversely, while turning a trolley onto a side track where it will kill five people

instead of one is an action which “could reasonably be expected to lead to serious bodily harm to

. . . a particular group of people,” it is also “the result of deflecting an existing threat onto a

different party” (Greene et al., 2001, p. 2107). On the authors’ account, therefore, it should be

assigned to their “moral­impersonal” category and judged permissible. Yet, 85% (17/20) of

respondents in Experiment 5 thought this action was impermissible. How did our subjects

manage to come to these conclusions? The answer cannot be the one proposed by Greene et al.

(2001).

As we have argued, a simpler and more powerful explanation of all of these moral

intuitions is ready to hand, one that grows out of the rationalism that Greene and colleagues too

hastily reject. We need only assume people possess tacit knowledge of specific moral principles

and the ability to compute mental representations of various actions in morally cognizable terms.

The operative reason why pushing the man in the Footbridge Problem is impermissible is

because it constitutes intentional battery. The operative reason why turning the trolley in the

94

Bystander Problem is permissible is because the battery and homicide it generates are foreseen

but non­intended side effects that are outweighed by the intended good effect of preventing the

train from killing the men. By contrast, the operative reason why turning the trolley in the

Disproportional Death Problem is impermissible is because its costs outweigh its benefits.

Finally, the operative reason why throwing the man out of the path of the train in the Consensual

Contact Problem is permissible is both because its benefits outweigh its costs and because this

action is not a battery at all. In light of the reasonable presumption that a person would consent

to being thrown and injured if that were necessary to save his life, the man’s hypothetical

consent to being thrown in these circumstances may be assumed.

Greene and colleagues raise an important objection to the computational approach. They

observe that in an unusual variant of the Bystander Problem invented by Judith Thomson

(Thomson, 1985), in which the side track leading to the one person loops around to connect with

the track leading to the five people, most people agree that it would be permissible to turn the

trolley, even though doing so would appear to violate the Kantian injunction against “using” a

person to achieve a worthy end (Greene et al., 2001, p. 2106). But the authors fail to note that

the original Bystander Problem and Thomson’s looped track example differ in their temporal,

causal, and counterfactual properties (Costa, 1987). More importantly, Greene and colleagues do

not explain why one must accept their tacit assumption that the trolley would be turned in

Thomson’s looped track example for any purpose other than the one motivating the agent in the

original Bystander Problem—to prevent the train from killing the men. If one refrains from

making this assumption, then the intuition that turning the train is permissible in these

circumstances can be explained along familiar lines.

95

Recall that the “looped track” scenarios we utilized in Experiment 4 were designed in

part to investigate just this issue. In Experiment 4, we discovered that even in this context

subjects remained sensitive to the distinction between intended means and foreseen side effect.

Moreover, they did so even though the process of recovering these intentional properties from

the impoverished stimulus is non­trivial. Further, the relevant data presumably cannot be

explained by appealing to varying levels of emotional engagement, because both scenarios

involve an “impersonal” action (Greene et al., 2001) described as “throwing the switch.” The

data can be explained, however, with reference to unconscious computational principles.

Finally, we emphasize that Greene and colleagues do not provide a clear procedure for

determining whether the three features they identify are contained in (or otherwise derivable

from) their stimuli. We are told that patterns of brain activity can be predicted on the basis of

whether an action can “reasonably be expected to lead to serious bodily harm” to a person or

group “where this harm is not the result of deflecting an existing threat onto a different party”

(Greene et al. 2001, p. 2107), but we are not told how a given stimulus is to be analyzed in these

terms. A further virtue of our computational alternative is that these crucial first steps in the

perceptual process are fully analyzed. That is, not only can our subjects’ judgments be generated

by fixed and recognizable moral principles, once appropriate mental representations have been

computed. The mental representations themselves can be derived from their corresponding

stimuli by a process that is purely mechanical (§8.1). In short, a complete and explicit theory of

the steps converting proximal stimulus to perceptual response can be given. In this sense, the

theory of moral cognition presented here is at least observationally if not descriptively adequate.

Greene et al. (2001) offer no comparable explanation of the conversion of proximal stimulus to

perceptual response. Hence their theory is not even observationally adequate.

96

8.3 Conclusion

This paper summarizes the results of six experiments performed on 543 individuals,

including 513 adults and 30 children ages 8­12. The results constitute evidence that both adults

and children ages 8­12 possess intuitive or unconscious knowledge of specific moral principles,

including the prohibition of intentional battery and the principle of double effect. Significantly,

this knowledge appears to be merely tacit: when participants were asked to explain or justify

their judgments, they consistently failed to provide adequate justifications for their judgments.

Our findings also suggest that at least some moral intuitions and the principles that generate them

are widely shared, irrespective of demographic variables like gender, nationality, age, and

education. Finally, our findings imply that long­standing questions in moral cognition may be

fruitfully investigated within the framework of theoretical models similar to those utilized in the

study of language and other cognitive systems. Specifically, we have shown how it may be

possible to pursue a “Galilean style” (Chomsky, 1980) of scientific explanation in this domain, in

which observable data are rigorously explained in terms of computational rules and

representations. Having gathered evidence that individuals possess intuitive knowledge of moral

principles, we are now in a better position to determine how this knowledge is acquired and

whether and to what extent it may be innate. Our study thus paves the way for future research

into the nature and origin of human moral intuitions.

97

Acknowledgments:

We thank the participants in our study and the following individuals who enabled us to recruit subjects and collect data: Cary Coglianese, Nagwa Hultquist, Laura Sandler, Amber Smith, Laila Waggoner, Kenneth Winston, Fei Xu, and Yaoda Xu. Izzat Jarudi and Martin Hewitt provided assistance coding and analyzing data. Thanks also to Kirsten Condry, who created the looped track diagrams used in Experiment 4, and to Paul Bloom, Noam Chomsky, Danny Fox, Steve Goldberg, Tom Grey, Marc Hauser, Lisa Heinzerling, Ray Jackendoff, Emma Jordan, Mark Kelman, Joshua Knobe, Don Langevoort, David Luban, Matthias Mahlmann, James McGilvray, Shaun Nichols, Philippe Schlenker, Mike Seidman, Steve Stich, Josh Tenenbaum, and Kathy Zeiler for helpful suggestions and encouragement. Research support was provided by the Department of Brain and Cognitive Sciences, Massachussets Institute of Technology; the Department of Psychology, New York University; the Department of Psychology, Harvard University; Stanford Law School; Georgetown University Law Center; and the Peter Wall Institute for Advanced Studies at the University of British Columbia.

References:

Anscombe, G. E. (1970). War and murder. In R. Wasserstrom (Ed.), War and morality (pp. 42­ 53). Los Angeles: Wadsworth Publishing Company.

Baker, M. (2001). The atoms of language: The mind’s hidden rules of grammar. Basic Books. Baird, J. A. (2001). Motivations and morality: Do children use mental state information to

evaluate identical actions differently? Paper presented to Biennial Meeting, Society for Research in Child Development, Minneapolis, MN.

Berndt, T. J., & Berndt, E. G. (1975). Children’s use of motives and intentionality in person perception and moral judgment. Child Development, 46, 904­912.

Carey, S. & Spelke, E. (1994). Domain­specific knowledge and conceptual change. In L. A. Hirschfeld & S. A. Gelman (Eds.), Mapping the mind: Domain specificity in cognition and culture (pp. 169­200). New York: Cambridge University Press.

Chomsky, N. (1964). Current issues in linguistic theory. The Hague: Mouton. Chomsky, N. (1965). Aspects of the theory of syntax. Cambridge, MA: MIT Press. Chomsky, N. (1978). Interview with Noam Chomsky. Linguistic Analysis, 4(4), 301­319. Chomsky, N. (1980). Rules and representations. New York: Columbia University Press. Chomsky, N. (1986). Knowledge of language: Its nature, origin, and use. Westport, CT: Praeger. Chomsky, N. (1995). The minimalist program. Cambridge, MA: MIT Press. Chomsky, N. & Halle, M. (1968). The sound pattern of English. Cambridge, MA: MIT Press. Cosmides, L. & Tooby, J. (1994). Beyond intuition and instinct blindness: Toward an

evolutionary rigorous cognitive science. Cognition, 50, 41­77. Costa, M. J. (1987). Another trip on the trolley. In J. M. Fischer & M. Ravizza (Eds.), Ethics:

Problems and principles (pp. 303­307). Fort Worth, TX: Harcourt Brace Jovanovich. Costanza, P. R., Coie, J. D., Grumet, J. F., & Farnhill, D. (1973). A

reexamination of the effects of intent and consequence on children’s moral judgments. Child Development, 44, 154­61.

D’Arcy, E. 1963. Human acts: An essay in their moral evaluation. Oxford: Clarendon Press.

98

Darley, J. M. & Shultz, T. R. (1990). Moral rules: Their content and acquisition. Annual Review of Psychology, 41, 525.

Dobbs, D. (2000). The law of torts. St. Paul, MN: West Group. Donagan, A. (1977). The theory of morality. Chicago: University of Chicago Press. Dwyer, S. (1999). Moral competence. In R. Stainton (Ed.), Philosophy and Linguistics. Boulder,

CO: Westview Press. Edwards, C. (1975). Societal complexity and moral development: A Kenyan study. 3 Ethos 505. Fischer, J.M. & Ravizza, M. (1992). Ethics: Problems and principles. Fort Worth, TX: Harcourt

Brace Jovanovich. Fodor, J. (1983). Modularity of mind: An essay on faculty psychology. Cambridge, MA: MIT

Press. Foot, P. (1967). The problem of abortion and the doctrine of double effect. In J. M. Fischer & M.

Ravizza (Eds.), Ethics: Problems and principles (pp. 59­66). Fort Worth, TX: Harcourt Brace Jovanovich.

Gazzaniga, M. S. (1992). Nature’s mind: The biological roots of thinking, emotions, sexuality, language, and intelligence. New York: Basic Books.

Gazzaniga, M. S., Ivry, R. B., & Mangum, G. R. (1998). Cognitive neuroscience: The biology of the mind. Cambridge, MA: MIT Press.

Gergely, G., Nadasdy, Z., Csibra, G., & Biro, S. (1995). Taking the intentional stance at 12 months of age. Cognition, 56, 165­93.

Gilligan, C. (1982). In a different voice. Cambridge: Harvard University Press. Ginet, C. (1990). On action. New York: Cambridge University Press. Goldman, A. (1970). A theory of human action. Princeton: Princeton University Press. Goldman, A. (1993). Ethics and cognitive science. Ethics, 103, 337­60. Greene, J., Sommerville, R., Nystrom, L., Darley, J., & Cohen, J. (2001, September 14). An

fMRI investigation of emotional engagement in moral judgment. Science, 293. Haegeman, L. (1994). Introduction to government and binding theory. Oxford: Basil Blackwell. Haidt, J. (2001). The emotional dog and its rational tail: A social intuitionist approach

to moral judgment. Psychological Review, 108, 814­834. Harman, G. (1977). The nature of morality: An introduction to ethics. New York: Oxford

University Press. Harman, G. (2000). Explaining value and other essays in moral philosophy. New York: Oxford

University Press. Hauser, M. D., Chomsky, N., & Fitch, T. (2002). The faculty of language: What is

it, who has it, and how did it evolve? Science, 298, 1569­1579. Hauser, M. D., Cushman, F., Young, L., Jin, R. K­X., & Mikhail, J. A dissociation between

moral judgments and justifications. Manuscript submitted for publication. Helmholtz, H. von. (1867/1962). Helmholtz’s treatise on physiological optics. (J.P.C. Southall,

Ed. and Trans.). New York: Dover. Henkin, L., Pugh, R. C., Schacter, O., & Smit, H. (1992). Basic documents supplement to

international law: Cases and materials. St. Paul, MN: West. Hilliard, F. (1859). The law of torts (Vols. 1­2). Boston: Little, Brown and Company. Holstein, C. (1976). Development of moral judgment: A longitudinal study of males and

females. Child Development, 47: 51­61. Johnson, S.C. (2000). The recognition of mentalistic agents in infancy. Trends in Cognitive

Science, 4(1), 22­28. Jonsen, A. R. & Toulmin, S. (1988). The abuse of casuistry: A history of moral

99

reasoning. Berkeley, CA: University of California Press. Kagan, J. (1987). Introduction. In J. Kagan & S. Lamb (Eds.), The emergence of morality in

young children (pp. ix­xx). Chicago: University of Chicago Press. Kant, I. (1787/1965). Critique of pure reason. (N. K. Smith, Trans.). St. Martin’s Press. Kaplow, L. & Shavell, S. (2001). Fairness versus welfare. Harvard Law Review, 114, 961­1061. Katz, J. (1972). Semantic theory. New York: Harper and Row. Katz, L. (1987). Bad acts and guilty minds: Conundrums of the criminal law. Chicago:

University of Chicago Press. Kohlberg, L. (1958). The development of modes of thinking and choices in years 10 to 16.

Unpublished doctoral dissertation, University of Chicago (on file with author). Kohlberg, L. (1981). Essays on moral development: Vol. 1. The philosophy of moral

development. New York: Harper and Row. Kohlberg, L. (1984). Essays on moral development: Vol. 2. The psychology of moral

development. New York: Harper and Row. LaFave, W. R. (2003). Principles of criminal law. St. Paul, MN: West Group. Leblanc, H. & Wisdom, W. (1993). Deductive logic (3rd ed.). Englewood Cliffs, NJ:

Prentice Hall. Lewontin, R. (1990). The evolution of cognition. In D. N. Osherson & E. E. Smith (Eds.),

An Invitation to Cognitive Science: Vol. 3 (pp. 229­46). Cambridge, MA: MIT Press. Lyons, D. (1965). Forms and limits of utilitarianism. Oxford: Clarendon Press. Macnamara, J. (1990). The development of moral reasoning and the foundations of geometry.

Journal for the Theory of Social Behavior, 21, 125­50. Macnamara, J. (1999). Through the rearview mirror: Historical reflections on psychology.

Cambridge, MA: MIT Press. Mahlmann, M. (1999). Rationalismus in der praktischen theorie: Normentheorie und

praktiche kompetenz. Baden­Baden: Nomos Verlagsgesellschaft. Marr, D. (1982). Vision. San Francisco: Freeman. Meltzoff, A. (1995). Understanding the intentions of others: Re­enactment of intended

acts by 18­month­old­children. Developmental Psychology, 31, 838­50. Mikhail, J. (2000). Rawls’ linguistic analogy: A study of the generative grammar model of

moral theory described by John Rawls in A Theory of Justice. Unpublished doctoral dissertation, Cornell University.

Mikhail, J. (2002). Law, science, and morality: A review of Richard Posner’s The problematics of moral and legal theory. Stanford Law Review, 54, 1057­1127.

Mikhail, J. (in press). Rawls’ linguistic analogy. Cambridge: Cambridge University Press. Mikhail, J., Sorrentino, C., & Spelke, E. (1998). Toward a universal moral grammar. In M. A.

Gernsbacher & S. J. Derry (Eds.), Proceedings, Twentieth Annual Conference of the Cognitive Science Society (p. 1250). Mahwah, NJ: Lawrence Erlbaum Associates.

Nagel, T. (1986). The view from nowhere. In J. M. Fischer & M. Ravizza (Eds.), Ethics: Problems and principles (pp. 165­79). Fort Worth, TX: Harcourt Brace Jovanovich.

Nelson, S. 1980. Factors influencing young children’s use of motives and outcomes as moral criteria. Child Development, 51, 823.

Nozick, R. (1968). Moral complications and moral structures. Natural Law Forum, 13, 1­50. Piaget, J. (1932/1965). The moral judgment of the child. New York: The Free Press. Petrinovich, L. & O’Neill, P. (1996). Influence of wording and framing effects on

moral intuitions. Ethology and Sociobiology, 17, 145­171. Petrinovich, L., O’Neill, P., and Jorgensen, M. (1993). An empirical study of moral intuitions:

100

Toward an evolutionary ethics. Journal of Personality and Social Psychology, 64(3), 467­478.

Pinker, S. (1994). The language instinct: How the mind creates language. New York: Harper­ Collins.

Pinker, S. (1997). How the mind works. New York: Norton. Prior, A.N. (1955). Formal logic. Oxford: Clarendon Press. Prosser, W. (1941). Casebook on torts. Minneapolis, MN: University of Minnesota Press. Quinn, W. S. (1993). Morality and action. Cambridge: Cambridge University Press. Rawls, J. (1971). A theory of justice. Cambridge, MA: Harvard University Press. Rest, J. (1983). Morality. In J. Flavell & E. Markman (Eds.), Manuel of child psychology: Vol. 3.

Cognitive development (4th ed.). New York: Wiley. Robinson, P. H. & Darley, J. M. (1995). Justice, liability, and blame: Community views

and the criminal law. San Francisco: Westview Press. Salmond, J. (1902/1996). Salmond on jurisprudence (12th ed., P.J. Fitzgerald, Ed.). London:

Sweet & Maxwell. Shapo, M. S. (2003). Principles of tort law. St. Paul, MN: Thomson­West. Sidgwick, H. (1907). The methods of ethics. Indianapolis, IN: Hackett. Simpson, E. (1974). Moral development research: A case study of scientific cultural bias. Human

Development, 17, 81­106. Spelke, E. (1998). Nativism, empiricism, and the origins of knowledge. Infant

Behavior and Development, 21. Spelke, E.S., Breinlinger, K., and Jacobson, K. (1992). Origins of knowledge. Psychological

Review, 99, 605­32. Stich, S. (1993). Moral philosophy and mental representation. In M. Hechter et al. (Eds.),

The origin of values. New York: Aldine De Gruyter. Stone, J. (1964). Legal system and lawyers’ reasonings. Stanford, CA: Stanford University Press. Thomson, J. J. (1985). The trolley problem. In J. M. Fischer & M. Ravizza (Eds.), Ethics:

Problems and principles (pp. 67­76). Fort Worth, TX: Harcourt Brace Jovanovich. Thomson, J. J. (1986). Rights, restitution, and risk. Cambridge, MA: Harvard University

Press. Turiel, E. (1983). The development of social knowledge: Morality and convention.

Cambridge, MA: Cambridge University Press. Unger, P. (1996). Living high and letting die: Our illusion of innocence. New York: Oxford

University Press. Von Wright, G. H. (1951). An essay in modal logic. Amsterdam: North­Holland

Publishing Company. Woodward, A. L., Sommerville, J. A., & Guajardo, J. J. (2001). How infants make sense of

intentional action. In B. Malle, L. Moses & D. Baldwin (Eds.), Intentions and intentionality: Foundations of social cognition (pp.149­169). Cambridge, MA: MIT Press.

101

9. Appendix

In this Appendix, we reproduce for each experiment both the scenarios we used and our

subjects’ responses to them.

9.1 Experiment 1

9.1.1 Scenarios

1. Transplant Problem ­ Intentional Battery (Dr. Brown) Dr. Brown has five patients in the hospital who are dying. Each patient needs a new organ in order to survive. One patient needs a new heart. Two patients need a new kidney. And two more patients need a new lung. Dr. Brown can save all five patients if he takes a single healthy person and removes her heart, kidneys, and lungs to give to these five patients. Just such a healthy person is in Room 306. She is in the hospital for routine tests. Having seen her test results, Dr. Brown knows that she is perfectly healthy and of the right tissue compatibility. If Dr. Brown cuts up the person in Room 306 and gives her organs to the other five patients, he will save the other five patients, but kill the person in Room 306 in the process. If Dr. Brown does not cut up the person in Room 306, the other five patients will die.

Is it morally permissible for Dr. Brown to cut up the person in Room 306?

2. Footbridge Problem—Intentional Battery (Frank) Frank is on a footbridge over the train tracks. He knows trains, and can see that the one approaching the bridge is out of control. On the track under the bridge there are five people; the banks are so steep that they will not be able to get off the track in time. Frank knows that the only way to stop an out­of­control train is to drop a very heavy weight into its path. But the only available, sufficiently heavy weight is a large man wearing a backpack, also watching the train from the footbridge. Frank can shove the man with the backpack onto the track in the path of the train, killing him; or he can refrain from doing this, letting the five die.

Is it morally permissible for Frank to shove the man? Y or N

3. Footbridge Problem—Intentional Battery (Ian) Ian is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Ian sees what has happened: the driver of the train saw five men walking across the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the five men. It is moving so fast that they will not be able to get off the track in time. Fortunately, Ian is standing next to a heavy object, which he can throw onto the track in the path of the train, thereby preventing it from killing the men. Unfortunately, the heavy object is a man, standing next to Ian with his back turned. Ian can throw the man, killing him; or he can refrain from doing this, letting the five die.

Is it morally permissible for Ian to throw the man?

102

4. Footbridge Problem—Intentional Battery (Nancy) Nancy is taking her daily walk near the train tracks when she notices that the train that is approaching is out of control. Five men are walking across the tracks. The train is moving so fast that they will not be able to get off the track in time. Nancy is standing next to a man, whom she can throw in front of the train, thereby preventing it from killing the men. Nancy can throw the man, killing him but saving the five men; or she can refrain from doing this, letting the five die.

Is it morally permissible for Nancy to throw the man?

5. Trolley Problem—Foreseeable Battery (Charlie) Charlie is driving a train when the brakes fail. Ahead five people are working on the track with their backs turned. Fortunately Charlie can switch to a side track, if he acts at once. Unfortunately there is also someone on that track with his back turned. If Charlie switches his train to the side track, he will kill one person. If Charlie does not switch his train, he will kill five people.

Is it morally permissible for Charlie to switch his train to the side track? Y or N

6. Trolley Problem—Foreseeable Battery (Denise) Denise is a passenger on a train whose driver has just shouted that the train's brakes have failed, and who then fainted of the shock. On the track ahead are five people; the banks are so steep that they will not be able to get off the track in time. The track has a side track leading off to the right, and Denise can turn the train onto it. Unfortunately there is one person on the right hand track. Denise can turn the train, killing the one; or she can refrain from turning the train, letting the five die.

Is it morally permissible for Denise to switch the train to the side track? Y or N

7. Bystander Problem—Foreseeable Battery (Edward) Edward is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Edward sees what has happened: the train driver saw five workmen men ahead on the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the five men; the banks are so steep that they will not be able to get off the track in time. Fortunately, Edward is standing next to a switch, which he can throw, that will turn the train onto a side­track. Unfortunately, there is one person standing on the side­track, with his back turned. Edward can throw the switch, killing the one; or he can refrain from doing this, letting the five die.

Is it morally permissible for Edward to throw the switch? Y or N

8. Bystander Problem—Foreseeable Battery (Hank) Hank is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Hank sees what has happened: the driver of the train saw five men walking across the tracks and slammed on the brakes, but the brakes failed and the driver fainted.

103

The train is now rushing toward the five men. It is moving so fast that they will not be able to get off the track in time. Fortunately, Hank is standing next to a switch, which he can throw, that will turn the train onto a side track, thereby preventing it from killing the men. Unfortunately, there is a man standing on the side track with his back turned. Hank can throw the switch, killing him; or he can refrain from doing this, letting the five die.

Is it morally permissible for Hank to throw the switch? Y or N

9.1.2 Responses

A. Intentional Battery Scenarios Problem Judgment Justification Adequacy Footbridge (Ian)

Impermissible Obviously, there is no right or wrong answer to this scenario. Both options presented are not "optimal" ones. In a sense the question may be interpreted if one life is more valuable than five lives. All lives were innocent bystanders in the scenario and either way, the killing of 5 versus 1 is still the loss of a life.

Adequate

Transplant (Bob)

Impermissible The moral permissibility of this action is dependent on the healthy patient's own judgments as to whether she would be willing to offer her body in this way. If she did offer permission, then things would be different.

Adequate

Footbridge (Frank)

Impermissible Based on the wording of the scenario, I am left to assume that Frank does not know the large man or the five people on the tracks. As such, he is a third­party, passive spectator to the upcoming tragedy, and has no moral position at stake­­he is not responsible for the train being out of control or the presence of the five people in its path. But by causing the death of the large man, he becomes actively involved, in a morally negative way. Because the train is inanimate and makes no moral choice in killing the five, this scenario is not a moral dilemma for anybody. In considering the trade of one life for five, Frank assumes that his immoral act would fight the greater immorality of the train, thereby raising the moral equilibrium of the situation. In fact, because the train has no morals, he only lowers the moral equilibrium of the scenario. It is never morally permissible to lower the moral equilibrium. . . . I am fascinated that it took me 5 seconds to determine my answer, and 45 minutes to justify it, above. I have never given any thought to a moral equilibrium before, or whether all my decisions consider raising or lowering it, but as I think about choices I've made, it seems that my view of the "right" choice is tied to the idea of how it balances the morality of the situation. I wonder when this was instilled in me.

Adequate

Footbridge (Ian)

Impermissible In this case, Ian has the morally superior option of throwing himself in front of the train instead of the innocent bystander. Even if Ian is too light to stop the train, he cannot throw the bystander in front of the train without bearing moral responsibility for the bystander's death, even though the act would save five lives.

Adequate

104

Footbridge (Nancy)

Impermissible Shoving the man is tantamount to [her] taking some sort of responsibility for the otherwise randomly occurring situation. I don't feel that it is appropriate for her to decide to "sacrifice" this man, who is in as safe a position as she is, simply because only one person will be injured instead of 5. She could drop a heavy rock or the man's backpack, if that would be sufficient, but to choose on her own to put this man's life in jeopardy or cause him physical harm seems beyond the realm of what one human being should be allowed to do. "She cannot play God!" A nice simple crystallization of much more complicated situations that probably occur all the time in the world, on different levels and in different contexts.

Adequate

Transplant (Bob)

Impermissible As it was written (with the donor patient healthy), it was very easy to consider [the doctor's] action immoral. However, if she had been terminally ill and in a vegetative state, I would have had to think a little harder (though of course I would want to know the wishes of her and her family before deciding).

Adequate

Transplant (Bob)

Impermissible The scenario present is one which, to my thinking[/feeling], leaves the participants with [no] feeling of ambiguity [in] dilemma. My response was immediate, clear. Despite the mathematical gain in lives saved, to sacrifice one healthy life to save the lives of five others is unacceptable.

Adequate

Footbridge (Nancy)

Impermissible What if the driver was able to stop in time before hitting the five people but the heavy man was injured anyway? Besides, she does not have the right to risk his life without his permission—if he wanted to, he could jump in front of the train (or she could) but she can't force him.

Adequate

Footbridge (Ian)

Impermissible Ian is not justified in deciding who gets to live or die. I'm not sure how plausible it is for a train to be stopped by one man, no matter how heavy he is.

Adequate

Footbridge (Ian)

Impermissible Because the five workers will be killed in our accident where no one will be blamed, but the heavy man will have to be deliberately murdered to save the workers. After all, accidents happen and people die—the force of nature. It's more acceptable than a murder.

Adequate

Footbridge (Nancy)

Impermissible Replace: 5 with 100, people with school children => no judgment (or even: "no" becomes "yes")

Inadequate

Transplant (Bob)

Impermissible In principle (absolute secrecy etc.) should do for greater good.

Inadequate

Footbridge (Frank)

Impermissible He can jump himself, if he wants to... Inadequate

Footbridge (Nancy)

Permissible I dislike large men with backpacks, but some would [like me] to be pushed myself.

Adequate

105

B. Foreseeable Battery Scenarios Problem Judgment Justification Adequacy Trolley (Denise)

Permissible I don't have a strong preference. It seems that either way something bad will happen, and you can argue either action or inaction is the moral route. But there isn't time to decide in the real situation, so there is no moral obligation to act one way or the other.

Adequate

Bystander (Hank)

Permissible I would think that Hank's intentions should be examined. If Hank's intention was fed by personal desire (i.e., desire to participate in the act of killing), then there would be a different light thrown on the subject.

Adequate

Trolley (Denise)

Permissible Since D. knows what is going to happen if she does nothing, she would be remiss not to act. In a perverse sort of way it is better to kill one person than five, all else being equal. On the other hand, if the one person on the spur sees the train and believes s/he is safe and fails to move, while the 5 on the main track know what's happening, the situation might be different. But if there's not time for people to run off a train track, how is there time to switch tracks safely?

Adequate

Trolley (Charlie)

Permissible I am a utilitarian, so the scenario of 1 death is preferable to five deaths. Such decisions are made every day, although the differences between mortality in alternate scenarios are in "statistical" terms, rather than attached to particular people.

Adequate

Trolley (Charlie)

Permissible Given that in either outcome someone will die (and barring any way to alert any of the people on the tracks), it seems morally justifiable to take an action which lessens the loss of life.

Adequate

Trolley (Denise)

Permissible Denise has only two choices: 1) action turning to the side track and killing one person or 2) inaction which will lead to five deaths. I think that the only viable choice is action, turning the train. By doing so, she will save four lives that would have been lost. The scenario's wording makes it seem that Denise is in some way responsible for killing the person on the right side­track rather than being a hero for saving the lives of five people.

Adequate

Bystander (Edward)

Permissible By yes, I mean EITHER option is “morally permissible.” The other would have been just as morally permissible. Either option is sub­optimal. I feel that 5 lives lost equals one life lost without further info. Both are awful. But say the 5 people were serial murderers and rapists plus the one person was say, some saintly person, then one might consider their views on capital punishment and reevaluate.

Inadequate

Bystander (Edward)

Permissible I went through several stages in answering this and generally found it troubling/morally difficult. 1st reaction: yes, throw switch. 2nd reaction: no! Is active killing. 3rd reaction: ramp up to large city/bomb vs. countryside. Obviously would divert to countryside.

Inadequate

Bystander (Hank)

Permissible It was a difficult decision to make!! The words that stand out are “killing the man” vs. “letting the five die”...which is a greater responsibility? Do we determine the value of life by number (5 vs. 1)? I'm left looking for a better solution to the problem. Could Hank warn the single

Inadequate

106

man standing on the track? Is it really Hank's right to choose who will die that day? Does it make him more responsible for death in general now that he has played an active role? It also brings up the question of bystander apathy—are we perpetrators by not “acting”?

Trolley (Denise)

Permissible I think that the obvious answer is "Yes" (and in fact it is likely the answer you would predict that most people would choose). For that reason, I would word the question differently (so that it is less loaded). I would ask: "Is it morally permissible for Denise to refrain from turning the train."

Inadequate

Trolley (Charlie)

Permissible The gut reaction is to turn. In the given scenario, I read it as if I was the driver, moving quickly in a panic situation. I believe that I would turn off the main track, THEN see the 1 person working there. If the driver (me) had taken the time to look at the 1 person, then decide, the opportunity to...[rest is too light to read]

Inadequate

Bystander (Hank)

Permissible Compare with following scenario: Hank is a physician with five sick patients. Each needs an organ transplant to live: one needs a lung, one a kidney, one a spleen, one a heart, one a brain (this is hypothetical). Fortunately, there is a healthy visitor (with a backpack, coincidentally) who has a healthy set of organs—lung, kidney, spleen, heart, brain. Is it morally permissible for Dr. Hank to sacrifice the backpack guy for organs?

Inadequate

Bystander (Hank)

Impermissible The reason I wrote no is because I don't believe that one can evaluate the worth of human lives by numbers alone. The scenario outlined here, if it mandated one of two actions, then I would choose to pull the switch, not because it is morally permissible but because it is the best available option.

Inadequate

107

9.2 Experiment 2

9.2.1 Scenarios

1. Bystander Problem—Foreseeable Battery (Hank) Hank is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Hank sees what has happened: the driver of the train saw five men walking across the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the five men. It is moving so fast that they will not be able to get off the track in time. Fortunately, Hank is standing next to a switch, which he can throw, that will turn the train onto a side track, thereby preventing it from killing the men. Unfortunately, there is a man standing on the side track with his back turned. Hank can throw the switch, killing him; or he can refrain from doing this, letting the five die.

Is it morally permissible for Hank to throw the switch?

2. Footbridge Problem—Intentional Battery (Ian) Ian is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Ian sees what has happened: the driver of the train saw five men walking across the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the five men. It is moving so fast that they will not be able to get off the track in time. Fortunately, Ian is standing next to a heavy object, which he can throw onto the track in the path of the train, thereby preventing it from killing the men. Unfortunately, the heavy object is a man, standing next to Ian with his back turned. Ian can throw the man, killing him; or he can refrain from doing this, letting the five die.

Is it morally permissible for Ian to throw the man?

3. Modified Bystander Problem—Bad Effects Outweigh Good Effects (Karl) Karl is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Karl sees what has happened: the driver of the train saw five million dollars of new railroad equipment lying idle across the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the equipment. It is moving so fast that it will be destroyed. Fortunately, Karl is standing next to a switch, which he can throw, that will turn the train onto a side track, thereby preventing it from destroying the equipment. Unfortunately, there is a man standing on the side track with his back turned. Karl can throw the switch, killing him; or he can refrain from doing this, letting the equipment be destroyed.

Is it morally permissible for Karl to throw the switch?

4. Modified Footbridge Problem—Consensual Contact (Luke) Luke is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Luke sees what has happened: the driver of the train saw a man walking across the tracks and slammed on the brakes, but the brakes failed and the driver fainted.

108

The train is now rushing toward the man. It is moving so fast that he will not be able to get off the track in time. Fortunately, Luke is standing next to the man, whom he can throw off the track out of the path of the train, thereby preventing it from killing the man. Unfortunately, the man is frail and standing with his back turned. Luke can throw the man, injuring him; or he can refrain from doing this, letting the man die.

Is it morally permissible for Luke to throw the man?

5. Trolley Problem—Foreseeable Battery (Charlie) Charlie is driving a train when the brakes fail. Ahead five people are working on the track with their backs turned. Fortunately Charlie can switch to a side track, if he acts at once. Unfortunately there is also someone on that track with his back turned. If Charlie switches his train to the side track, he will kill one person. If Charlie does not switch his train, he will kill five people.

Is it morally permissible for Charlie to switch his train to the side track? Y or N

9.2.2 Responses

A. Pair #1: Bystander Problem (Hank) and Footbridge Problem (Ian) Problem Judgment Justification Adequacy Bystander Permissible

Footbridge Impermissible

The man, Hank can here actively influence a sequence of events which will limit damage (# of deaths). In the second event, he cannot throw another man onto the tracks because he will actively and deliberately kill an innocent bystander. Really­­an impossible choice.

Inadequate

Bystander Permissible

Footbridge Impermissible

It's amazing that I would not throw a person but throw a switch to kill a person. I really wish there was more I could do for the 1 guy on the other track.

Inadequate

Bystander Permissible

Footbridge Impermissible

In either case, the moral decision rule depends on how close to the active killing of the man is.

Inadequate

Bystander Permissible

Footbridge Impermissible

Not acceptable to decide to risk someone else's life to save others.

Inadequate

Bystander Permissible

Footbridge Impermissible

I know­­five lives are five lives­­it's all about the guts. That's what it comes down to. Blaise Pascal got it all wrong.

Inadequate

Bystander Impermissible

Footbridge Impermissible

For the first scenario, I wanted to draw a distinction between "is it permissible for him to throw the switch" and "does he have a duty to throw the switch," though I don't know if that would have changed my answer.

Inadequate

Bystander Permissible

Footbridge Permissible

I believe that the ultimate question is that of lives lost. Some would argue that Hank and Ian would be morally justified in not stopping the train. While this may be true, it does not necessitate that it be morally unjustified to stop the train.

Adequate

109

B. Pair #2: Bystander Problem (Hank) and Modified Bystander Problem (Karl) Problem Judgment Justification Adequacy Bystander Permissible

Modified Bystander

Impermissible

First scenario­­morally justifiable if he feels that a net savings of 4 lives (and knowingly taking an action to kill a man) would be better than non­action and therefore, saving the life of one. Second scenario­­life is more important than money.

Adequate

Bystander Permissible

Modified Bystander

Impermissible

For the first scenario, saving 5 at expense of 1 is better than saving 1 at expense of 5. For the second scenario, value of even 1 person's life is greater than $5 million, if only because equipment can be replaced.

Adequate

Bystander Permissible

Modified Bystander

Impermissible

It is morally permissible for Hank to throw the switch. This can be framed as a "binary" decision. If we assume that Hank's decision is 100% accurate, he will choose to throw rather than not to throw. Karl is thus directly responsible for the man's life. Not switching the train would make him morally responsible for the man's death (Throw switch => 1 killed; Don't throw switch => 5 killed).

Inadequate

Bystander Permissible

Modified Bystander

Impermissible

While a life should not always be preferred to $5 million, in this case the railroad equipment is probably not going to be put to any better use­­it will probably just serve the greedy interests of the railroad firm itself. (Furthermore, the railroad firm probably got the $5 million unjustly.)

Inadequate

C. Pair #3: Footbridge Problem (Ian) and Modified Footbridge Problem (Luke) Problem Judgment Justification Adequacy Footbridge Impermissible

Modified Footbridge

Permissible

One cannot place another in harm's way. Ian may choose to throw himself in front of the train only. It is morally permissible for Luke to risk himself if he so chooses. These are all individual, private decisions relating to control over one's self.

Adequate

Footbridge Impermissible

Modified Footbridge

Permissible

I guess I didn't think Ian should actively murder someone, but if Luke didn't mind risking his own life, that's fine. He shouldn't be morally obligated to do so, however.

Adequate

Footbridge Impermissible

Modified Footbridge

Permissible

For Question 1, it's OK to throw yourself on the tracks to save the five. It's not OK to kill somebody else (who's innocent) to spare your own life. For Question 2, it's OK to force people if you think it's in their best interest, in circumstances like this.

Adequate

D. Pair #4: Footbridge Problem (Ian) and Modified Bystander Problem (Karl) Problem Judgment Justification Adequacy Footbridge Impermissible

Modified Bystander

Impermissible

In scenario 1, the issue is essentially allowing 5 to die (passive) or killing 1 (active). I don't feel that utility justifies the killing of the one to save the 5. Similarly in scenario 2, rights not to be killed trump utility.

Adequate

Footbridge Impermissible In the first case, killing a person is much different than letting people live. You can chose to kill yourself if you

Adequate

110

Modified Bystander

Impermissible feel that sacrifice will benefit society, but you cannot choose to kill someone else without their consent. Second case, money is a cheap substitute for life. I realize that we cannot spend limitless money to save lives, but he cannot kill a person due to our choice to reduce costs.

Footbridge Impermissible

Modified Bystander

Impermissible

My [judgment] is intuitive and I realize not logically justifiable. My intuition is that throwing 1 man on track of a passing train will not stop the train. I also am reluctant to grade life and thus equate [the] value of one life as worth more than 5, even though I know this can be done.

Inadequate

E. Pair #5: Bystander Problem (Hank) and Modified Footbridge Problem (Luke) Problem Judgment Justification Adequacy Bystander Impermissible

Modified Footbridge

Permissible

While in situation 1 Hank might not have the moral authority to kill the man on the side track, it may be the best policy decision to do so.

Inadequate

Bystander Impermissible

Modified Footbridge

Permissible

Option 1: the utility of killing a man to save 5 does not warrant a transgression of one's right not to be killed unjustly. And I am not morally obligated (it is not right) to deliberately kill someone to save other lives.

Inadequate

F. Pair #6: Modified Bystander Problem (Karl) and Modified Footbridge Problem (Luke) Problem Judgment Justification Adequacy Modified Bystander

Impermissible

Modified Footbridge

Permissible

Money should not determine whether you save a life or not if those are the options given. $5,000,000 vs. killing life.

Inadequate

Modified Bystander

Impermissible

Modified Footbridge

Permissible

Luke is not morally required to throw the man, but it is permissible and even nice of him to do so.

Inadequate

Modified Bystander

Impermissible

Modified Footbridge

Permissible

On the 2nd question, clearly it is morally permissible for him to do so but he is not morally obligated to do so.

Inadequate

G. Pair #7: Trolley Problem (Charlie) and Bystander Problem (Hank) Problem Judgment Justification Adequacy Trolley Impermissible

Bystander Impermissible

It is hard for me to allow one person to make the choice regarding whether other people die, regardless of the number of people.

Adequate

Trolley Permissible

Bystander Permissible

Of course it's wrong to allow anyone to die in both scenarios, but killing one is not as bad as losing 5 lives.

Adequate

111

Trolley Permissible

Bystander Permissible

My immediate instinct would be to envision the families of 5 vs. 1

Inadequate

H. Pair #8: Trolley Problem (Charlie) and Footbridge Problem (Ian) Problem Judgment Justification Adequacy Trolley Permissible

Footbridge Impermissible

Very odd. I don't know why I chose differently in the second scenario. The end result is the same. I just chose my gut response­­and now am intrigued with how to reconcile them.

Inadequate

Trolley Permissible

Footbridge Impermissible

Moral actors may be forced to make a decision between two passive choices where both will end rights. But to make action over passive choices requires another kind of analysis and degree of benefit.

Inadequate

Trolley Permissible

Footbridge Impermissible

In the first scenario it would be permissible to act as a utilitarian optimizer. In the second rights come into question.

Inadequate

I. Pair #9: Trolley Problem (Charlie) and Modified Bystander Problem (Karl) Problem Judgment Justification Adequacy Trolley Permissible

Modified Bystander

Impermissible

I believe that it is always best to minimize the number of lives lost in a choice situation, through direct action. Although I have a problem with comparing any sum of money to human life, the situation could be more difficult for me if the equipment were so valuable it would save life in the future, or irreplaceable.

Adequate

Trolley Permissible

Modified Bystander

Impermissible

Yes for the first scenario is you will save five lives at the expense of one. And I would treat all lives equally. For the second scenario, life has to be greater than dollar amounts.

Adequate

Trolley Permissible

Modified Bystander

Impermissible

For the first scenario, first one [would be an utility reasoning?] (more utility for society if you kill one compare to five). For the second scenario, this answer depends on the value you put on life, which is equal to the utility of the person. I personally put a large value on life.

Adequate

Trolley Permissible

Modified Bystander

Impermissible

Doesn't the train have a horn? Inadequate

Trolley Permissible

Modified Bystander

Impermissible

If I am directly responsible for the outcome in the given situations ­ I will always choose these answers.

Inadequate

Trolley Permissible On the second case, I am able to place a boundary between the right to life and a utilitarian option. In the

Inadequate

112

Modified Bystander

Impermissible first case, I am dealing also with the right to life, but in this case I prefer to save 5 lives (is that utilitarian? I don't think is comparable).

J. Pair #10: Trolley Problem (Charlie) and Modified Footbridge Problem (Luke) Problem Judgment Justification Adequacy Trolley Permissible

Modified Footbridge

Permissible

It was not difficult to answer because the question was about "morally permissible." Ask me whether or not I would risk my life to save another person, I might say yes but if I was in Luke's situation (2nd scenario) I do not know what I would do.

Inadequate

Trolley Permissible

Modified Footbridge

Permissible

I had no doubt at all on my answers. Inadequate

Trolley Permissible

Modified Footbridge

Permissible

In the second case, it is morally "permissible" for Luke to throw the man. However, he is not morally obligated to do so as he may risk his own life in the process.

Inadequate

113

9.3 Experiment 3

9.3.1 Scenarios

Same as Experiment 2.

9.3.2 Responses

Subjects in Experiment 3 were not asked to provide justifications.

114

9.4 Experiment 4

9.4.1 Scenarios

1. Looped Track Problem—Intentional Battery (Ned) Ned is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Ned sees what has happened: the driver of the train saw five men walking across the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the five men. It is moving so fast that they will not be able to get off the track in time. Fortunately, Ned is standing next to a switch, which he can throw, that will temporarily turn the train onto a side track. There is a heavy object on the side track. If the train hits the object, the object will slow the train down, thereby giving the men time to escape. Unfortunately, the heavy object is a man, standing on the side track with his back turned. Ned can throw the switch, preventing the train from killing the men, but killing the man. Or he can refrain from doing this, letting the five die.

Is it morally permissible for Ned to throw the switch?

2. Looped Track Problem—Foreseeable Battery (Oscar) Oscar is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Oscar sees what has happened: the driver of the train saw five men walking across the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the five men. It is moving so fast that they will not be able to get off the track in time. Fortunately, Oscar is standing next to a switch, which he can throw, that will temporarily turn the train onto a side track. There is a heavy object on the side track. If the train hits the object, the object will slow the train down, thereby giving the men time to escape. Unfortunately, there is a man standing on the side track in front of the heavy object, with his back turned. Oscar can throw the switch, preventing the train from killing the men, but killing the man. Or he can refrain from doing this, letting the five die.

Is it morally permissible for Oscar to throw the switch?

9.4.2 Responses

A. Looped Track Problem – Intentional Battery (Ned) Judgment Justification Adequacy Impermissible The man has the same right to live as the 5 men. I think that Ned must

not interfere with what naturally is going to happen. Adequate

Impermissible He doesn’t have the right to impose his morals on others (without consent/agreement).

Adequate

Impermissible would be playing god . . . (besides, if Ned had time to hatch this plan, then how come the 5 men didn't have time to jump?)

The only decisions of morality you can make involve yourself. i.e. should I die to save them.

Adequate

115

Maybe the man wouldn't want to sacrifice his life to save the 5 Impermissible If Ned were able to save the 5 men without killing a sixth, I believe he

should. If there were some way to alert the man on the side track and let him escape (being that his back is turned and he has no reason to believe he is going to be hit by the train), then I could justify throwing the switch.

I could see how someone would believe it was moral in that 5 lives would be saved at the expense of one. And honestly, I don't know what I would do if I were Ned, but somehow choosing who is going to die . . . that is, deciding which life has more value, is not "morally permissible" in my value system.

Adequate

Impermissible Question: how can a man slow a train down?

I wouldn't know what to do. I do not think I have the right to judge the importance value of life based on the quantity, or the age. Look for different alternative. I do not have the right to take one's life in order to save 5 others. It would be wrong.

Adequate

Impermissible This question sucks. Who is to decide one life for many? One life is worth as much as the rest—or do they not weigh more for the greater need? Yuck. I can’t decide morality!

Inadequate

Impermissible But I think I would. (?) Inadequate Impermissible Although I believe Ned should hit the switch anyway for nonmoral

reasons, I also believe that both hitting and not hitting the switch is morally wrong. Assuming that Ned has absolute knowledge that one act kills five men and the opposite act kills one, there is no moral option. In addition, there is no "more" moral option. Taking action (or not taking action) that results in the death of human being is immoral regardless of whether the same action saves another human being or other human beings. The making of this decision is beyond the bailiwick of morality.

Inadequate

Impermissible (1) How could you know it would give them sufficient time to escape? Would you guarantee a man's death for the chance of saving a few others. Would you be so bold? What if you misjudged the viscosity?

Inadequate

Impermissible This situation is exactly alike, in every respect, to one in which a doctor has five patients on a given day, each needing a transplant of a different organ; without such a transplant, they will all die within a day. A sixth perfectly healthy patient comes in. The question is, assuming that the doctor can perform all five transplants perfectly, guaranteed, should he kill the healthy man to save the five? I must answer this no, hence my response to the question above.

Inadequate

Impermissible Respondent drew arrows from the five men to the train, and from there to the object on the other side of the tracks, and commented: Time is longer here than here.

Give more time a chance.

Inadequate

Permissible (1) But try yelling at the guy to get off the track

(2) It sucks but 5 lives to 1 right? What if those 5 are criminals and the 1 is a saint.

There's also the Fat Man in the Cave Problem. Where you're with a group of people exploring a cave and the entrance caves in trapping a fat man in the entrance (his rear is sticking out of the cave and torso in the cave). And you have one stick of dynamite. The tide is rising and you

Adequate

116

will all drown if you don't get out. Blow up the fat guy or not? One of the most interesting answers I've ever heard is that if the entire group decides to blow up the entrance thereby blowing up the fat man, then at least the person telling him could hold his hand. 2 lives for the group. Is that the ethical thing to do?

Permissible The problem as stated is imperfect in that Ned has some uncertainty about the situation ­ he cannot truly know that flipping the switch will trade one life for five (all six could die, or the driver could regain consciousness, etc.). However, the essence of the question is "kill one to save five?" I consider this "permissible" from a theoretical view; it is the logically superior choice given no other information. "Yes." In practice, though, I don't know if I would personally have the strength to directly kill another human under any circumstances; there may be an almost physical limit on my ability to do harm. I agree that doing ­100 points to save +500 has a net +400 balance, but my "evil" limit may be ­10, beyond which I might not be able to act.

Adequate

Permissible By throwing the switch, he would be "killing" the one man but he'll be saving the lives of the 5 men. That's the only way I can justify it. Otherwise, he'd be ignoring a grave problem and watching it happen. That is morally irresponsible.

Adequate

Permissible Yell at the 5 men. Inadequate Permissible Either way Ned will have nightmares about it. Inadequate Permissible "This is sick" Inadequate Permissible "This sucks" Inadequate Permissible But he shouldn’t because the 5 men know what’s coming. The single

man has his back turned and has no idea what is going on. But it is perfectly moral for him to throw the switch and spare 5 lives.

Inadequate

B. Looped Track Problem – Foreseeable Battery (Oscar) Judgment Justification Adequacy Impermissible I don't know. I don't think that Oscar has the right to decide who lives or

who dies, but he should do something. I guess, no. Adequate

Impermissible No, because Oscar has no right to play God. Adequate Impermissible Because Oscar is killing an individual himself, therefore being

accountable for a murder rather than a mishap that he was not responsible for.

Adequate

Impermissible The man on the side track might know that no train was due, so was justified in having his back turned—however, the 5 men crossing may have been doing so knowing about the potential risk.

Adequate

Impermissible No, since it would not be morally right to involve and kill an innocent bystander, even though it would save the lives of 5 men.

Adequate

Impermissible It’s never morally right to kill someone even if it means that 5 other lives can be saved.

Adequate

Impermissible My gut feeling is no, it's not morally permissible ­ he is playing God and human beings aren't "quantifiable" like that (5 more important than one). On the other hand, I think the action is to throw the switch.

Adequate

Impermissible If he throws the switch he is taking a deliberate action which will (or likely will) result in the death of an innocent bystander.

If he does nothing, he is allowing an act of god to proceed over which he has no control other than to decide to kill someone (morally unacceptable).

Adequate

117

The 5 potential victims are accountable for their own situation (i.e. standing on a train track which is in service).

Impermissible Morally I believe that humans do not have the right to decide to take the lives of other humans, and in this situation nature should be allowed to run its course. Therefore, I believe that it is morally impermissible to throw the switch. In other cases, however, such as the taking of one's own life (euthanasia), I believe that there are certain circumstances where the switch could morally be "thrown."

Adequate

Impermissible Respondent drew an arrow pointing to the five men and commented: Shouldn't have been there in the first place.

Adequate

Impermissible Those 5 people were wrong in being on the track. To kill a man that is innocent and not doing anything wrong is wrong. The other people are walking on the track in danger.

Adequate

Impermissible Since the man on the side track is innocent. The side track is not open for operation because there is a heavy object on the track. So it’s ok for the man standing on the side track.

Adequate

Impermissible Oscar is in no position to play God deciding whose life (lives) are more important.

Adequate

Impermissible It's not morally permissible for him to let the 5 men die if it were in his power to change that. Meanwhile, it is also not permissible for him to take this other man's life. I think Oscar is caught in a situation in which neither choice would be morally right; he is going to have to base his decision on some other standards other than morality.

Inadequate

Impermissible Overpopulation is a serious problem in many areas of the world. Inadequate Impermissible This is not a question of the value of life. Inadequate Permissible If he can save the lives of five people, he has the moral right to do so. He

can save more people this way. Adequate

Permissible BTW, me and my friends were talking about this yesterday and the man is actually Oscar's friend in his version. And would it be more wrong for Oscar to turn the switch while the train is going towards his friend on while the train is going towards the main track. I think neither way is permissible but if I have to choose, saving more lives is more desirable.

Adequate

Permissible I think either action­­throwing the switch or not­­is permissible with respect to any "universal" moral law, that is, whatever Hank does, we must presume was the lesser of two evils in his mind. Therefore, Hank would have tried to do the right thing, as he understands it. Doing what one thinks is right seems to be the equivalent of doing one's best, which is all that could ever be expected of anyone, and therefore seems universally defensible

Adequate

Permissible The question is which is safer for the train; how many people are on it. Assuming the train would be fine and the only worry is the people, one is to lose the one to save the 5. For the most part, it seems down to number or morality [Too light to read well]

Adequate

Permissible The needs of the many outweigh the needs of the few, or the one. Adequate Permissible Instinctively yes because you’re going to save 5 people. Adequate Permissible Simple math: 1 man dead vs. 5 men. None of them should have been

crossing the tracks!! Adequate

Permissible Murder is not immoral if it prevents other murders. But the law would disagree. Were I placed in the role of Oscar, I might have to consider the option of not taking any action, and pretending not to be able to help.

Adequate

Permissible 1 < 5 Adequate Permissible It is morally permissible for Oscar to do either, but if you want to know

what I would do if I was Oscar, I would not throw the switch. Inadequate

Permissible When answering this question, I in no way use the argument that one Inadequate

118

death is better than five, but instead look at the possibility of having no deaths. If the train stays on its present course, 5 people will die. However, if the train is redirected, there is a chance, though a slight one, that everyone may live. I would redirect the train, doing everything I could to get the man's attention, and, God willing, he'd turn around and be able to escape the oncoming train.

Permissible On the one hand, nobody may act as God, and decide who lives and who dies. An alternative is only feasible if it does not implies the death of a human being. On the other hand, the real world is complex. Since a long time ago that that human lives are lost to save a culture, a nation, etc. Furthermore, I spent almost 10 minutes trying to decide and I am still undecided. So, it is reasonable that the man is also undecided, and any option will therefore be acceptable.

Inadequate

Permissible The man on the side track can hear train and thus try to jump off track. Thus saving 5 men and probably 6.

Inadequate

Permissible I don’t know if that’s what I would do, but I don’t think it’s immoral to do it.

Inadequate

119

9.5Experiment 5

9.5.1 Scenarios

1. Bystander Problem—Foreseeable Battery (Hank) Hank is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Hank sees what has happened: the driver of the train saw five men walking across the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the five men. It is moving so fast that they will not be able to get off the track in time. Fortunately, Hank is standing next to a switch, which he can throw, that will turn the train onto a side track, thereby preventing it from killing the men. Unfortunately, there is a man standing on the side track with his back turned. Hank can throw the switch, killing him; or he can refrain from doing this, letting the five die.

Is it morally permissible for Hank to throw the switch?

2. Modified Bystander Problem # 2 (Mark: Intentional Homicide) Mark is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Mark sees what has happened: the driver of the train saw five men walking across the tracks and slammed on the brakes, but the brakes failed, and the driver fainted. The train is now rushing toward the five men. It is moving so fast that they will not be able to get off the track in time. Fortunately, Mark is standing next to a switch, which he can throw, that will turn the train onto a side track, thereby preventing it from killing the men. Unfortunately, there is a man on the side track. Mark can throw the switch, killing him; or he can refrain from doing this, letting the men die.

Mark then recognizes that the man on the side track is someone who he hates with a passion. “I don’t give a damn about saving those five men,” Mark thinks to himself, “but this is my chance to kill that bastard.”

Is it morally permissible for Mark to throw the switch in order to kill the man on the side track?

3. Modified Bystander Problem #3 (Richard: Better Alternative) Richard is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Richard sees what has happened: the driver of the train saw five men walking across the tracks and slammed on the brakes, but the brakes failed, and the driver fainted. The train is now rushing toward the five men. It is moving so fast that they will not be able to get off the track in time. Fortunately, Richard is standing next to a switch, which he can throw, that will turn the train onto a side track, thereby preventing it from killing the men. Unfortunately, there is a man on the side track with his back turned. Richard can throw the switch, killing him; or he can refrain from doing this, letting the men die.

120

By pulling an emergency cord, Richard can also redirect the train to a third track, where no one is at risk. If Richard pulls the cord, no one will be killed. If Richard throws the switch, one person will be killed. If Richard does nothing, five people will be killed.

Is it morally permissible for Richard to throw the switch?

4. Modified Bystander Problem #4 (Steve: Disproportional Death) Steve is taking his daily walk near the train tracks when he notices that the train that is approaching is out of control. Steve sees what has happened: the driver of the train saw a man walking across the tracks and slammed on the brakes, but the brakes failed and the driver fainted. The train is now rushing toward the man. It is moving so fast that he will not be able to get off the track in time. Fortunately, Steve is standing next to a switch, which he can throw, that will turn the train onto a side track, thereby preventing it from killing the man. Unfortunately, there are five men standing on the side track with their backs turned. Steve can throw the switch, killing the five men; or he can refrain from doing this, letting the one man die.

Is it morally permissible for Steve to throw the switch?

9.5.2 Responses

A. Pair #1: Bystander Problem (Hank) and Intentional Homicide Problem (Mark) Problem Judgment Justification Adequacy Bystander Permissible

Intentional Homicide

Impermissible

1. Because I don't think there is a right and wrong answer and that there are multiple alternatives to this scenario, I think Hank can throw the switch and probably also save the man standing on the side track. I think Hank can warn the man to move while he is throwing the switch. Furthermore, I think that by not throwing the switch the train can cause more damage than what is now foreseeable.

2. While Mark would be saving the men walking across the tracks, it is not right for Mark to throw the switch primarily to kill the man on the side track.

Inadequate

Bystander Permissible

Intentional Homicide

Impermissible

Scenario totally implausible from a time/speed/distance/safety equipment perspective. Also, throwing the switch might derail the train, killing all six pedestrians & the engineer.

Inadequate

Bystander Permissible

Intentional Homicide

Impermissible

Although I'm not advocating taking a person's life in this situation I believe that the loss of one would be far greater an outcome than five. With that being said, I might not feel that way if the one person was someone I had personal ties with such as a father or brother. Question #2: I do not believe that malicious intent of any kind is an acceptable solution to causing anyone harm or loss of life even if I despised an individual. I'm a firm believer that what goes around comes around and I

Inadequate

121

believe that God can and will take care of the individual in His own time.

Bystander Permissible

Intentional Homicide

Impermissible

Mark's choice to kill is the amoral decision in the 2nd scenario.

Inadequate

Bystander Permissible

Intentional Homicide

Impermissible

Hank Scenario: Do what's best for the greater good as long as the decision is not impacted by personal feeling.

Mark Scenario: Personal feeling to make life­threatening decision is not morally correct.

Adequate

Bystander Impermissible

Intentional Homicide

Impermissible

I'm not sure what is meant by the phrase "morally permissible." I'm not sure if that is supposed to mean whether I think a particular act is justified or whether I think a majority of people would approve of a particular decision. Ultimately I decided to avoid taking affirmative steps to end another person's life. I have no idea if that is "moral," however.

Adequate

Bystander Permissible

Intentional Homicide

Permissible

The hypo made it sound like the choice was between certainly saving 5 people v. certainly saving 1. If the chances of saving those 5 were not as certain, I would feel differently; even 90% would change my answer.

Inadequate

B. Pair #2: Bystander Problem (Hank) and Better Alternative Problem (Richard) Problem Judgment Justification Adequacy Bystander Permissible

Better Alternative

Impermissible

1) While it is morally permissible, it is not morally advisable.

Inadequate

Bystander Permissible

Better Alternative

Impermissible

He should pull the cord and save everyone's lives. Inadequate

Bystander Permissible

Better Alternative

Impermissible

Easy questions. The questions about terminating life without dignity are harder. Questions about the death penalty are harder still.

Inadequate

Bystander Permissible

Better Alternative

Impermissible

Without knowing the individuals in the scenarios, I believe that, if it is possible to save 5 while harming only 1, then you should do so. However, if you have a third option which would harm no one, then that is the ideal choice.

Adequate

Bystander Permissible

Better Alternative

Impermissible

"First do no harm," and then within that context "do as little as possible." Since Richard has a third and vastly preferable choice, it is not morally permissible for him to take an active role in choosing what he considers the lesser of 2 evils. If he can "first do no harm" he should morally always choose that role.

Adequate

Bystander Impermissible Reason for first: the five are at fault and though it would Adequate

122

Better Alternative

Impermissible be nice to help them, I don't think it's moral to kill a completely innocent person to do this. If I would say yes, I feel like I would be making a statement that five guilty people's lives are more important than one innocent person's. I think that's wrong.

Reason for second: he has an alternative to save everyone and it's morally right to choose that one.

Bystander Impermissible

Better Alternative

Impermissible

I think it would be fine if he pulled the cord to save everyone, but it is not morally permissible to decide if 1 man or 5 has the right to live or die.

Adequate

Bystander Impermissible

Better Alternative

Permissible

First scenario: The five men on the tracks chose to be there on the tracks and know the risks as the train is approaching. The man on his own is on a different track unknowing . . . And I would be choosing to murder him to save 5 others . . . My liability. Second scenario: A chance to save all 6 . . . only issue is . . . what about the speed of the train? Is it a passenger train? By making the switch would it derail? Could I put more people at risk by attempting to make the switch and it not be successful?

Inadequate

123

C. Pair #3: Bystander Problem (Hank) and Disproportionate Death Problem (Steve) Problem Judgment Justification Adequacy Disproportional Death

Permissible

Bystander Impermissible

Life of 5 is greater than 1. Adequate

Bystander Permissible

Disproportional Death

Impermissible

While no innocent person's fate should be in the hands of another in an ideal world, in this case, death is inevitable, and it would be morally best to do the greatest good for the greatest number. Therefore, intervening and putting the train off of its course for a net of four lives saved is morally justifiable, if not ideal. In this case, my definition of moral is doing the right/rational thing.

Adequate

Bystander Permissible

Disproportional Death

Impermissible

They should both take the action that saves the most lives

Adequate

Disproportional Death

Permissible

Bystander Impermissible

Due to the preservation of life, it is better to lose one life than to lose five. My vote is do not throw the switch, but pray that the one that is on the track is able to get off safely.

Adequate

Bystander Permissible

Disproportional Death

Impermissible

These don't take into account: A) That Hank could yell to the people/the man on the side of the track (who ought have [illegible])

B) That his throwing of the switch could actually derail the train and kill/injure even more people.

Nevertheless, while 5 people is more than one person in the standpoint of potential casualties, the five are also 5 times more likely as a unit to notice the sound of the speeding train.

Hank is in a position to play God. Perhaps.

Inadequate

Disproportional Death

Permissible

Bystander Impermissible

Both scenarios provoke a lot of thought and human introspect. The fact that you (Hank) has the ability to change fate is somewhat scary. There are a lot of questions that should be answered prior to making a truly informed decision. My hope is that given the circumstances Hank would have the ability to warn all of the men to get them out of harm's way.

Inadequate

Disproportional Death

Impermissible

Bystander Impermissible

Who am I to say who should live or die. It is the Lord's will.

Adequate

Disproportional Death

Impermissible

Bystander Impermissible

Hank/Steve cannot be certain of the outcome of his actions or how his actions will affect the outcome.

Adequate

Bystander Impermissible

Disproportional Death

Impermissible

I felt that whether it was 5 men on the tracks or 1, it was partly their responsibility for being in harm's way. The bystanders shouldn't interfere and cause an innocent man or man's lives to be lost ­ basically. The

Adequate

124

men on the tracks should have been more careful ­ their lack of care should not result in another's death.

Bystander Impermissible

Disproportional Death

Impermissible

The main reason that I answered "No" to both questions is that the thought of making a value judgment regarding human life seems immoral (or at least uncomfortable) to me. I understand that from a utilitarian standpoint, 5 lives are worth more than 1, but that 1 person is still valued immensely by those who love him. It doesn't seem appropriate for me to intervene, thereby deciding whose life or lives are more precious.

Adequate

Bystander Impermissible

Disproportional Death

Impermissible

It sounds as if the men walking across the tracks see the train coming. They should have made sure the tracks were clear before crossing. It's not fair to kill some man (or men) whose back is turned and has no idea what is going on.

Adequate

Disproportional Death

Permissible

Bystander Permissible

Question 1: The dilemma involving Hank (1st Paragraph) asks the question if it is "morally permissible" for Hank to throw the switch. It is "permissible" using a Utilitarian moral system ­ the greatest good for the greatest number of people. Hence, there is at least one moral system where this act is permissible. The irony is that this Enlightenment position devalues life (in its totality) and it can also be reasoned that the man with his back turned, who will presumably be killed, may have a more valuable life than the others. Ultimately, then, either action would permit a "moral" result ­ depending on your cognitive framework. Question 2: Same answer as above, though culturally a harder argument to make. The position of modernity is to count human beings ­ and thereby assume their equivalence. This is seen, for example, in the economic presumption of ["homo economics"] ­ which economics (contemporary) is based upon. The ultimate equivalence of human beings, however, shares the same fate as this concept ­ in practice homo economics does not exist, though it must be assumed intellectually.

Inadequate

125

9.6 Experiment 6

9.6.1 Scenarios

1. Transplant Problem ­ Intentional Battery (Dr. Brown) Dr. Brown has five patients in the hospital who are dying. Each patient needs a new organ in order to survive. One patient needs a new heart. Two patients need a new kidney. And two more patients need a new lung. Dr. Brown can save all five patients if he takes a single healthy person and removes her heart, kidneys, and lungs to give to these five patients. Just such a healthy person is in Room 306. She is in the hospital for routine tests. Having seen her test results, Dr. Brown knows that she is perfectly healthy and of the right tissue compatibility. If Dr. Brown cuts up the person in Room 306 and gives her organs to the other five patients, he will save the other five patients, but kill the person in Room 306 in the process. If Dr. Brown does not cut up the person in Room 306, the other five patients will die.

Is it wrong for Dr. Brown to cut up the person in Room 306?

2. Trolley Problem—Foreseeable Battery (Charlie) Charlie is driving a train when the brakes fail. Ahead five people are working on the track with their backs turned. Fortunately Charlie can switch to a side track, if he acts at once. Unfortunately there is also someone on that track with his back turned. If Charlie switches his train to the side track, he will kill one person. If Charlie does not switch his train, he will kill five people.

Is it morally permissible for Charlie to switch his train to the side track? Y or N

9.6.2 Responses

A. Transplant Problem – Intentional Battery (Dr. Brown) Judgment Justification Adequacy Impermissible I believe that it would be wrong to cut this 306 person up without them

even knowing it. It would be different if Dr. Brown had asked this person if they would donate their organs and he had received their permission. That is why I would blame him if he took their life.

Adequate

Impermissible I think that Dr. Brown should ask the person in room 306 if they would like to be cut up to save the other peoples' lives.

Adequate

Impermissible I said no because it never said that she gave permission to kill him; to give away her body parts…I did not feel good about it because I would not like somebody to take my body parts.”

Adequate

Impermissible Okay, if given consent. Adequate Impermissible For question #1 I said yes, because the person in room 306 was not

informed she would die and neither was her family. But the 5 patients are prepared and their families were aware that they were close to dying.

Adequate

Impermissible Dr. Brown should ask the person in room 306 permission. Adequate

126

Impermissible It's not fair to kill a perfectly healthy person that has not done anything wrong.

Adequate

Permissible It depends on how old they are. Adequate Permissible I think it's ok because your saving 5 lives for one. Adequate

B. Trolley Problem – Foreseeable Battery (Charlie) Judgment Justification Adequacy Permissible It's better to kill one person than five ­ if you have no other choice. Less

people will die. Adequate

Permissible Because if you kill 5 people it is worse. Just killing 1 person is not as bad as killing 5 people. I would shout out in front of the people to get out so I wouldn't hit the person. Could turning work?

Adequate

Permissible I think Charlie should change tracks because then he saves at least 5 but one is killed unless he hears the train so all and all to change would be better.

Adequate

Permissible I have no comments or questions. I do not know what to say. Inadequate Permissible If he switches tracks, will he still be going to the same place?"

If he didn't switch he may be blamed more and probably fired. He would be doing a good thing if he switched to the side track.

Inadequate

Permissible Wouldn't they hear the train coming? Inadequate Impermissible Not blaming him because he had to make a choice either way. Inadequate

127

Notes 1 Although this paper was written by me, it summarizes several years of research conducted with my colleagues Professor Elizabeth Spelke of the Department of Psychology, Harvard University, and Dr. Cristina Sorrentino of the Department of Psychology, New York University. Hence I use plural forms throughout. I was principal investigator in these studies, which began while I was a Lecturer and Research Affiliate in the Department of Brain and Cognitive Sciences at the Massachusetts Institute of Technology and continued while I was a law student at Stanford Law School. Although our findings have not yet been published, they have been presented in posters at annual meetings the Cognitive Science Society and the Society for Research in Child Development, as well as in lectures, workshops, and graduate seminars at Cornell University, Harvard University, M.I.T., Stanford University, the University of Berlin, and the Peter Wall Institute for Advanced Studies at the University of British Columbia. 2 The distinction between descriptive and explanatory adequacy is potentially confusing because correct answers to both problems are both descriptive and explanatory in the usual sense. A solution to the problem of descriptive adequacy is a description of the mature individual’s moral competence; at the same time it is an explanation of her moral intuitions. Likewise, a solution to the problem of explanatory adequacy is a description of the initial state of the moral faculty; at the same time, it is an explanation both of the individual’s acquired moral competence and (at a deeper level) those same intuitions (Mikhail, 2000; Mikhail, in press). 3 The syntactic form of the action­descriptions located on the nodes of these act trees (Goldman, 1970) calls for comment. Drawing on Goldman (1970) and Ginet (1990), we take the central element of what we call a complex act­token representation (Mikhail, 2000) to be a gerundive nominal, whose grammatical subject is possessive. Following Katz (1972), we use the symbol ‘at t’ to denote some unspecified position on an assumed time dimension, and we use superscripts on occurrences of ‘t’ to refer to specific positions on this dimension. We assume that superscripts can be either variables or constants. We take ‘t’ with the superscript constant ‘0’, i.e., ‘t (0) ’, to function as an indexical element in the complex act­token representation, serving to orient the temporal relationships holding between it and other such representations. Superscript variables (‘n’, ‘m’, etc.) denote members of the set of natural numbers and appear in superscripts with prefixed signs ‘+’ and ‘­’, indicating an appropriate number of positive and negative units from the origin point (‘t (0) ’) of the time dimension. For example, the symbol ‘t (+n) ’ signifies ‘n units to the right of the origin,’ and the symbol ‘t (­n) ’ signifies ‘n units to the left of the origin’. We also use additional variables after the signs ‘+’ and ‘­’ in our superscripts whose interpretation proceeds in accord with the conventions for adding and subtracting in algebra. For example, the symbol ‘t (+n + ( ­ m)) ’ signifies ‘n ­ m units to the right of the origin,’ whereas the symbol ‘t (­n + (­ m) + (­o)) ’ signifies ‘n + m + o units to the left of the origin.’ Thus the ordered sequence of representations on the vertical line (i.e., the trunk) of the act tree in Figure 3 indicates that throwing the man occurs before causing the train to hit the man, which occurs before preventing the train from killing the men. As we discuss in §8.1, this notational system for representing temporal structure also adopts an important convention, which is to date an action from its time of completion. See Mikhail (2000) for further discussion. 4 The elements of intentional battery in tort law are more complex and vary among commentators and jurisdictions (e.g., Dobbs, 2000; Shapo, 2003). For the purposes of this study, we characterize battery as touching without consent, intentional battery as touching without consent used as a means to a given end, and foreseeable battery (a neologism) as touching without consent embedded in an agent’s action plan as a side effect (see §2.1). These are not meant to be legally adequate definitions, but they are adequate given our present purposes. 5 An alpha level of .05 was used for all statistical tests. 6 The exact breakdown of participants and scenario pairs is listed below. Each participant received a questionnaire with one of the following ten pairs, with the number of participants in each condition listed in brackets:

• Pair #1: Bystander Problem (Hank) and Footbridge Problem (Ian) [10] • Pair #2: Bystander Problem (Hank) and Modified Bystander Problem (Karl) [5] • Pair #3: Footbridge Problem (Ian) and Modified Footbridge Problem (Luke) [5] • Pair #4: Footbridge Problem (Ian) and Modified Bystander Problem (Karl) [5] • Pair #5: Bystander Problem (Hank) and Modified Footbridge Problem (Luke) [5] • Pair #6: Modified Bystander Problem (Karl) and Modified Footbridge Problem (Luke) [5] • Pair #7: Trolley Problem (Charlie) and Bystander Problem (Hank) [5] • Pair #8: Trolley Problem (Charlie) and Footbridge Problem (Ian) [5] • Pair #9: Trolley Problem (Charlie) and Modified Bystander Problem (Karl) [10] • Pair #10: Trolley Problem (Charlie) and Modified Footbridge Problem (Luke) [10]

7 Recall that two subjects in Experiment 2 did not provide information about their gender. Both were given a Consensual Contact scenario. One judged the action to be permissible and one judged it to be impermissible.

128

8 We thank Fei Xu and Yaoda Xu for their assistance in collecting this data. 9 For one notable effort in this direction, see Hauser, Cushman, Young, Jin & Mikhail (manuscript submitted for publication). 10 According to our hypothesis, this action is judged to be impermissible because its bad effects are perceived to be disproportional to its good ones. 11 Differences between the two scenarios are underlined and italicized to make them more noticeable. Participants were given questionnaires without these markings. The Ned­Oscar pair and its looped­track design were also inspired by a notable debate over the principle of double effect in the philosophical literature (see Thomson, 1985; Costa, 1987). 12 This is merely an implication of the temporal order of good effects, bad effects, and batteries in these representations, which it may be recalled is what led us to create the Ned­Oscar pair in the first place. 13 Differences between the two scenarios are underlined and italicized to make them more noticeable. Participants were given questionnaires without these markings. 14 To clarify, we assumed that respondents would represent Mark as intending the bad effect but not the good effect. A more complex scenario would be one in which the agent has both intentions and the opportunity to hit two birds with one stone, so to speak. We leave investigation of this alternative for another occasion. 15 Although we do not pursue the issue here, the breakdown of responses by gender did appear significant. 7 of 8 girls and all 7 of the boys in the Foreseeable Battery condition judged the action constituting foreseeable battery (“Charlie’s turning the train”) to be permissible. By contrast, 1 of 6 girls in the Intentional Battery condition judged the action constituting intentional battery (“Dr. Brown’s cutting up the patient”) to be permissible, but— surprisingly—a majority of boys (5 out of 9) disagreed. These results suggest, albeit provisionally, that boys may be slower than girls in arriving at the standard adult view that certain deontological violations are impermissible. More research is needed to clarify this issue. 16 The same holds, of course, for other cognitive domains, like vision, musical cognition, face recognition, and so on. In each of these domains, we take for granted—and can discover empirically—that experimental subjects are consistently incapable of articulating the operative principles on which their intuitive judgments are based. Moral cognition appears to be similar to other cognitive domains in this respect, that is, to be both principled and intuitive, involving patterns of “unconscious inference” (e.g., Helmholtz, 1867/1962). 17 So conceived, the presumption of good intentions appears related to the so­called “first principle of practical reason,” according to which “good is to be done and pursued and evil avoided” (Macnamara, 1990; Mikhail, 2000). 18 The “Self­Preservation Principle” holds that if an agent’s doing something to a moral patient necessitates killing her, then the moral patient would not consent to it (Mikhail, 2000). The principle may presumably be overridden in certain circumstances, but we do not address that topic here. 19 Like Petrinovich and colleagues, Greene et al. (2001) do not appear to investigate deontic knowledge as such, because instead of asking subjects to decide whether a given action is morally permissible, they ask whether the action is “appropriate.” See Greene et al. 293 (5537): 2105 Data Supplement—Supplemental Data at http://www.sciencemag.org/cgi/content/full/293/5537/2105/DC1 (last visited 9/25/2001). That this question appears inapposite can be seen by considering the analogous inquiry in the study of language: asking whether a linguistic expression is “appropriate” rather than “grammatical.” Chomsky (1957, p.15) emphasized the importance of distinguishing grammatical from closely related but distinct notions like significant or meaningful, and the same logic appears to apply here. Additionally, Greene et al. (2001) do not provide evidence that trolley intuitions are stable, systematic, or widely shared. Instead, they merely report that “most people” say that one ought to turn the trolley in the Trolley Problem but one ought not to push the man in the Footbridge Problem, and they then note in passing that “[p]articipants’ responses to versions of the trolley and footbridge dilemmas were consistent with the intuitions described above” (Greene et al., 2001, p. 2108, n. 11). Further, whether one ought to perform a given action is distinct from whether the action is morally permissible, and the authors conflate this crucial distinction. Finally, none of the intuitions studied by Greene and colleagues appear to qualify as “considered judgments” in Rawls’ (1971) sense—that is, as moral judgments “in which our moral capacities are most likely to be displayed without distortion” (Rawls, 1971, p. 47)—because all of their probes are phrased in the first­person (e.g., “You are on a footbridge over the tracks … Is it appropriate for you to push the stranger?”) (emphasis added). This not only contravenes Rawls’ sensible warning that the theorist’s data set should exclude those judgments which “are likely… to be influenced by an excessive attention to our own interests” or which are given “when we are upset or frightened, or when we stand to gain in one way or the other” (Rawls, 1971, p. 47). It also appears to stack the deck in favor of the authors’ hypothesis that “variations in emotional engagement” (Greene et al., 2001) are responsible for generating this particular class of moral intuitions.