9.916Communication and Teaching

Today’s Lecture

Other’s actions are more than spatiotemporal patterns; they reflect, and predict, internal states

Predict future actions Theory of Mind

Learn new facts and skills Communication & Teaching

Communication & Teaching

X&*

X&*Reference

Facts skills

Encoding Decoding

X&* X&*

X&*

Message channel

Sample Evidence

Standard Model:

Ammendment (1):

X&* X&*X&*

Figure by MIT OpenCourseWare.

Communication & Teaching

X&*

X&*Reference

Facts skills

Encoding Decoding

X&* X&*

Message channel

Arbitrary

Signal

Behaviour

Form = Function

Standard Model:

Ammendment (2):

X&*X&*

Figure by MIT OpenCourseWare.

Communication & TeachingFor the teacher/communicator:

In Common:Audience-design

Distinct:

• Exploit learning mechanisms • Exploit perceptual sensitivities • Immediate cost • Immediate benefit • Motive: altruistic donation • Motive: manipulationor long-term benefits

Required reading, Sperber & Wilson 1986

SignalX&*X&*

X&*X&*

Figures by MIT OpenCourseWare.

Communication & TeachingFor the learner/communicatee:

In Common:Attention to new facts / skills

Faster / more efficient / more accurate learning

Distinct:

• Inference about intent • Inference(?) about reality • Context-sensitive • Relatively involuntary(?)

Required reading, Sperber & Wilson 1986

SignalX&*X&*

X&*X&*

Figures by MIT OpenCourseWare.

Communication & Teaching

• Primate alarm calls (specific predators, but not for informing receiver) • Primate affiliative calls -grunts & girneys (?)

Prototypical examples

Required reading, Sperber & Wilson 1986, Silk 2000

SignalX&*X&*

X&*X&*

Figures by MIT OpenCourseWare.

Pedagogical Sampling

Lack of evidence = evidence of lack

Teachers: - expert (i.e. basically correct) - helpful (i.e. provide approximately ideal evidence)

Learner Teacher

If you can point to two more daxes, you can have a sticker.

Here are two more daxes.

Xu & Tenenbaum 2007

Figures removed due to copyright restriction.

Pedagogical Sampling

Lack of evidence = evidence of lack

Teachers: - expert (i.e. basically correct) - helpful (i.e. provide approximately ideal evidence)

Learner Teacher If you can point to two more daxes, you can have a sticker.

Here are two more daxes.

Xu & Tenenbaum 2007 Courtesy of Wiley-Blackwell. Used with permission.

Figures removed due to copyright restriction.

Communication & TeachingFor the learner/communicatee:

In Common:Attention to new facts / skills

Faster / more efficient / more accurate learningDistinct:

• Inference about intent • Inference(?) about reality • Context-sensitive • Relatively involuntary(?) • Generic information • Local / episodic information

Required reading, Sperber & Wilson 1986

SignalX&*X&*

X&*X&*

Figures by MIT OpenCourseWare.

What’s missing?• “Telling” vs “Teaching” • Generic information?

Is human teaching unique?

0

1

1

1

2

2

3

2

3

4

5

6

7

Time

Run

ning

spe

ed (

mm

/s)

LeaderFollower

A

B

3

F

N

Leadbetter et al 2006

Ant tandem runningexpertnovice

Teaching:- Increases learning efficiency or content- Local cost to teacher - Only in the presence of a learner

Courtesy Elsevier, Inc., http://www.sciencedirect.com. Used with permission.

young. Prey is killed by the motherin front of the cubs. Later, when thecubs begin accompanying her onhunting trips, the mother releasesprey in front of them, which thecubs attempt to catch, sometimesat the cost of losing the preyaltogether [6]. The cubs’ predatoryskills improve over this period,although it remains to be shownthat this results directly from suchpractice (the same applies ina study on domestic cats [7]). Otherpotential cases of teaching involvechimpanzees learning to use stonehammers and anvils, and ospreysteaching their offspring to snatchfish from the water [4,5], but as yetthese rely only upon weakanecdotal evidence.

In contrast, Franks andRichardson’s [8] well-controlledstudy on tandem-runningTemnothorax ants was carried outin a laboratory. The intimateinteraction between leader andfollower in a pair of tandemlyrunning ants at first sight bears allthe hallmarks of a parent teachinga child to ride a bicycle. Anexperienced ant will lead individual

naı̈ve nest mates to newlydiscovered food sources or nestingsites, stopping if the follower losesregular antennal contact [9]. Whenthe pair becomes separated, asoccurs when the follower makeslooping movements possiblysearching for landmarks, the leaderremains still, only continuingtowards the food when the followerhas completed her exploratorycircuit (Figure 1). Franks andRichardson [8] demonstrate thatthere are clear two-wayinteractions between thetandem-running ants. When thegap between them becomes toolarge, and antennal contactbetween the pair is lost, the leaderslows down and the followeraccelerates to catch up. Thisbidirectional feedback loopappears to maximise the speed atwhich the two can progress, whileallowing the follower to memorisethe path and its surroundinglandmark features.

Such tandem-running meetsmost of the criteria for teaching setout in the definition given by Caroand Hauser [4]. When alone, the

leader does not incorporate thefrequent pauses which are used bythe follower to perform orientationloops. Hence the leader’sbehaviour is clearly modified in thepresence of a naı̈ve observer.Leaders incur a time cost: whenan experienced forager is notleading a follower, she travelsfaster to the food source anddoes not stop en route [8]. Asa result, the follower (pupil) findsthe target more quickly than shewould do if searching for it alone.While it appears likely thatfollowers learn route-specificinformation duringtandem-running, it remains tobe shown empirically preciselywhat information is obtained.Franks and Richardson [8] refine

Caro and Hauser’s [4] workingdefinition of teaching byintroducing an additional criterion:that feedback from the learner tothe experienced individual mustbe demonstrated. Suchfeedback clearly distinguishestandem-running from other formsof signalling in ants, such asscent-marking food sources, orreleasing alarm pheromones inthe presence of nest intruders [9].In these cases, both the signaland the response are largelyhard-wired; and there is no needto assume that learnt informationhas been transmitted, nor is therea need to invoke learning toexplain the receiver’s response.Most simple forms of signalling,such as use of pheromones, do notappear to meet several criteria laidout by Caro and Hauser’s teachingdefinition: such signals aredisplayed irrespective of thepresence of a naı̈ve receiver, anddo not lead to the long lastingchanges of receiver behaviourthat would qualify as learning[10,11]. So the additional criterionof feedback from the taughtindividual seems unnecessary.Responding to feedback frompupils makes for more efficientteaching, but teaching, albeitperhaps at a lower quality, can stilloccur in the absence of suchfeedback.In contrast, tandem-running in

ants, just like dancing inhoneybees, is a much moreadvanced form of communication.These behaviours specifically

Current Biology

Figure 1. Tandem-running by Temnothorax albipennis ants.

(A) Schematic view of path taken by a tandem-running pair of Temnothorax albipennisants from their nest (N) to a food source (F). (B) Running speed of leader (red line) andfollower (blue line) during the same tandem-run. Tandem leaders have experience ofthe food source, whilst followers are naı̈ve of its location. The leader proceeds towardsthe food source (red path) so long as the follower (blue path) maintains regular antennalcontact with the leader’s legs or abdomen. Progress of the tandem pair is slowed byfrequent periods when the leader remains still whilst the follower performs a loopedcircuit, possibly to memorise landmarks along the path (points 1 and 3) [8]. Oncethis exploratory circuit is complete, and the follower re-establishes antennal contact,the leader continues onwards towards the food. If contact between follower and leaderbecomes less frequent during a tandem-run, the leader will slow down to allow the fol-lower to catch up (point 2).

Current Biology Vol 16 No 9R324

Is human teaching unique?

Gergely Csibra

X&*X&*

The pedagogical stance:

18 month oldsHabituation:

Mmm! Yuck!

+ “ostensive cues”

Across people

+ Generalizable informatione.g. preference vs goodness

Figure by MIT OpenCourseWare.

Is human teaching unique? The pedagogical stance:

+ Generalizable information e.g. location: the A-not-B error

Figure removed due to copyright restrictions.

X&*X&*

1

0.8

0.6

0.4

0.2

0Ostensive-

communicativeNoncommunicative

A trails B trails

Proportion of correct searches (m

ean+SE)

Figure by MIT OpenCourseWare.

Figure by MIT OpenCourseWare.

Is human teaching unique? Teaching: - Increases learning efficiency or content - Local cost to teacher - Only in the presence of a learner - Generalizable skills or concepts

Meerkat feeding

Thomas et al 2006

0.00.10.20.30.40.50.60.70.80.9

Pup age0.00.10.20.30.40.50.60.70.80.9

Pup age0.00.10.20.30.40.50.60.70.80.9

Pup age

Intact Disabled Dead

ScorpionsNon-scorpions

Figure by MIT OpenCourseWare.

Is human teaching unique? Teaching: - Increases learning efficiency or content - Local cost to teacher - Only in the presence of a learner - Generalizable skills or concepts

Meerkat feeding

Thomas et al 2006

0.6

0.5

0.4

0.3

0.2

0.1

0.0

n = 10

n = 10

ExptControl

Proportion fed infact

Older pup calls

0.6

0.8

0.4

0.2

0.0n = 11

n = 11

ExptControl

Proportion fed dead

Younger pup calls

Figure by MIT OpenCourseWare.

What is unique about human communication?

Pedagogical Message encoding inferences and decoding

X&*X&*X&*X&* X&*X&* X&*X&*

Figures by MIT OpenCourseWare.

MIT OpenCourseWarehttp://ocw.mit.edu

9.916 Special Topics: Social Animals Fall 2009

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.