what is science? · 2013-06-10 · science: much deductive reasoning, much learning from authority,...
TRANSCRIPT
My Question
What makes science different from other forms of inquiry?
Especially respectable forms of inquiry…
Most especially respectable forms of inductive i.e. ampliative or non-deductive inquiry.
Some Forms of Inductive Inquiry
❖ Interpretation of sacred texts
❖ Other hermeneutic pursuits
❖ Systematic metaphysics
❖ Gathering folk wisdom
❖ Science
How Is Science Different?
❖ Better funded
❖ Special social organization
❖ Smarter personnel
❖ More empirical
How Is Science Different?
❖ Better funded
❖ Special social organization
❖ Smarter personnel
❖ More empirical
Empirical Exceptionalism
Science is more empirical than other knowledge-oriented pursuits
And because it is more empirical, it is in some ways better
(And maybe, in some ways, worse)
Three Ways To Be “Empirical”
1. Base your conclusions on observable phenomena
2. Test hypotheses with empirical content
3. Focus on a particular kind of relation between evidence and hypothesis: empirical confirmation
Confirmation
A piece of evidence confirms a hypothesis just in case it gives you more reason to believe the hypothesis than before
Used to be called: incremental confirmation
That a hypothesis is “confirmed” does not mean that it is likely to be true
Toward a Demarcation Criterion…
Science: Much deductive reasoning, much learning from authority, but all must be based in a substantial body of knowledge gained from empirical confirmation
Non-Science: Empirical confirmation, where it exists, is peripheral
Two Respectable Ways of Learning about Ravens
Empiricus:Go into the field, find ravens, observe their color, infer that all ravens are black
Two Respectable Ways of Learning about Ravens
Lector:Go to the library, read up on ornithology, find reputable authorities on raven color, infer that all ravens are black
Similarities
❖ Same hypothesis
❖ Supported inductively
❖ Evidence obtained through observation (raven color; sentences)
❖ A good way to learn
Inductive Support and Empirical Confirmation
A fact inductively supports a hypothesis whenever it increases your confidence that the hypothesis is true.
Empirical confirmation is a kind of inductive support, but not the only kind – not even the only good kind – though maybe the best.
Empirical Confirmation Is Essential to Science
There is much inductive support in science that is not empirical confirmation: books, journals, ornithological compendia…
But such things are “scientific” only insofar as the facts they convey are learned ultimately by empirical confirmation
What makes science special is that…
It is founded in a substantial body of empirical testing
Science Demarcated
Two Routes to a Theory of Empirical Confirmation
1. Build a self-standing theory of empirical inference
2. Given a theory of inductive support, specify a criterion for picking out the support relations that are empirical
Two Routes to a Theory of Empirical Confirmation
1. Build a self-standing theory of empirical inference
2. Given a theory of inductive support, specify a criterion for picking out the support relations that are empirical
More Exactly…
Evidence e empirically confirms hypothesis h just in case
1. The evidence inductively supports h
2. The evidence instance-confirms h, either directly or indirectly
Instance Confirmation Repurposed
❖ It is no longer an account of confirmation (that’s the job of the theory of support)
❖ It is an account of which support relations are empirical
Plan
1. A look at Hempel’s account of instance confirmation and its problems
2. A rival “causal” account that captures something intuitively empirical
3. Virtues of this kind of inductive support
Three Parts
1. Setup: the relata of the confirmation relation are sentences (hypothesis and evidence statement)
2. Direct confirmation
3. Indirect confirmation
Direct Confirmation
A hypothesis is confirmed by its positive instances and disconfirmed by its negative instances
(After Jean Nicod)
Instancehood
An evidence statement is a positive instance of All Fs are G just in case it entails that no object mentioned in the statement is a counterexample – an F that is not G.
☛ An F that is G
☛ A non-F
Indirect Confirmation
Consequence Principle: If e directly confirms (i.e., instantiates) one or more hypotheses, it indirectly confirms all logical consequences of those hypotheses.
Claims
Hempel confirmation is a prima facie a lot like empirical confirmation
What makes science special is the foundational role played by inductive support relations that are cases of instance confirmation
Four Problems
1. Only hypotheses about observables can be directly confirmed
2. Only hypotheses about observables can be indirectly confirmed
3. The white shoe
4. The Good problem
The Good Problem
World 1: 1000 objects, 901 ravens, 900 black
World 2: 1000 objects, 100 ravens, all black
The first thing you see is a black raven. Which world are you more likely to be in?
One Way Out of the Good Problem
Instance confirmation alone is not sufficient for empirical confirmation; you need inductive support as well.
The black raven does not empirically confirm the raven hypothesis
Nor does it empirically disconfirm it
Limited to: Causal Hypotheses
An In conditions Z, all Fs are G style hypothesis putatively true in virtue of a causal mechanism by which, in conditions Z, something about Fs causes G-hood
Target Mechanism
For a causal hypothesis In Z, all Fs are G:
☛ The actual mechanism whose consequences the hypothesis is supposed to describe
Associated Theory
For a causal hypothesis is the prevailing theory of the nature of the target mechanism
It’s what is taken to causally explain the hypothesis, if it’s true
Direct Confirmation
A hypothesis is confirmed by its positive instances and disconfirmed by its negative instances.
Instancehood
A state of affairs is an instance of If Z, all Fs are G just in case it is an F in conditions Z that is produced by the target mechanism
If it’s G, it’s a positive instance, if it’s not G, it’s a negative instance
Positive Instance
An instance that conforms to the hypothesis – a case where the hypothesis makes the correct prediction
A state of affairs is a positive instance of If Z, all Fs are G just in case it is an F in conditions Z that is produced by the target mechanism and that is G.
Instance Confirmation
A hypothesis is confirmed by its positive instances and disconfirmed by its negative instances
Indirect Confirmation
What is indirectly confirmed:
1. Associated theory
2. Hypotheses explained by associated theory
3. Initial conditions needed to explain this particular instance
4. G-hood of other Fs (in conditions Z)
Indirect Confirmation When e Instantiates h
1. The theory t that explains h
2. The other hypotheses explained by t
3. The initial conditions that explain e
4. The other things (instance predictions) explained by h
Example
Theory: Inflationary big bang cosmology
Hypothesis: The cosmic microwave background has such and such structure
Evidence: Observations of the microwave background
Theoretical Instances
Can All Fs are G be directly confirmed when F and G are theoretical, not observable, properties?
Yes, if you have evidence which confirms the existence of an F that also G – an instance.
Example
All electrons have charge 1.602 × 10–19 C
Instance-confirmed by an electron with such a charge
How to indirectly confirm such a state of affairs: find an effect that it helps to explain
Example
1. Observed motion of oil drop instantiates hypothesis about motion of drops
2. Indirectly confirm initial conditions used in explanation of motion
3. These include presence of electron with such-and-such charge
Four Problems in Hempel
1. Only hypotheses about observables can be directly confirmed
2. No upward flow of indirect confirmation from directly confirmed hypothesis
3. The white shoe
4. The Good problem
What makes science special is that…
It is founded in a substantial body of empirical confirmation
Science Demarcated
Reasons to Go Empirical?
Empirical confirmation is not dependent on background beliefs
Or at least, not so much as other forms of inductive reasoning
Dangers of Non-Empirical Reasoning
Metaphysical background beliefs:
Empiricus: What Aristotle says is typically true
Lector: What Aristotle says is almost always false
It Can’t Happen with Direct Confirmation
Facts about instancehood are totally determined by form of hypothesis and nature of target mechanism
It Can Happen with Indirect Confirmation
Look at confirmation of associated theories
Two theories: one posits p, other posits ¬p
As All Fs are G gets confirmed, opinions as to p or ¬p diverge
But…
If the theories, or p itself, are causal hypotheses…
☛ Use empirical testing to distinguish which of p and ¬p is true, or
☛ Use empirical testing to distinguish which of the two theories of the F/G connection is true
Two Complaints
1. Empirical reasoning – reasoning from effect to cause – ignores too much information of inductive importance
2. Empirical reasoning makes too many assumptions about the world’s causal structure
First Complaint
Important information ignored?
๏ That is the secret of science’s independence of background knowledge
๏ Tradeoff: Ignore relevant information for (perhaps) a longer-term gain
Reason, in order to be taught by nature, must approach nature with its
principles in one handKant, Critique of Pure Reason
Michael StrevensPhilosophy DepartmentNew York [email protected]