Do Animals Teach? - University of Bristol

Lessons from Teaching
in Ants
Do Animals Teach?
Nigel R. Franks
Professor of Animal Behaviour
& Ecology
School of Biological Sciences
University of Bristol
Teaching is “to show by way of
information or instruction”
Does teaching always require a
theory of mind?
(Oxford English Dictionary).
For example, a human teacher
might think,
“I know what I know and I know
what you do not.
So, I will teach you to rectify this
discrepancy.”
Does teaching always require a
theory of mind?
NO!
TEACHING
An individual is a teacher if …
“An individual actor A can be said to teach if
it modifies its behavior only in the presence of a
naïve observer, B, at some cost or at least
without obtaining an immediate benefit for itself.
1) it modifies its behaviour in the presence
of a naive observer,
A’s behaviour thereby provides B with
experience, or sets an example for B. As a
result, B acquires knowledge or learns a skill
earlier in life or more rapidly or efficiently than it
might otherwise do, or that it would not learn at
all.”
3) in order to set an example
2) at some initial cost to itself,
4) so that the other individual can learn more
quickly.
Caro TM, Hauser MD (1992). Quarterly Review of Biology. 67(2):151-74.
Caro TM, Hauser MD (1992). Quarterly Review of Biology. 67(2):151-74.
Do Any Animals
Teach?
Teaching is one form of social
learning; another is
“eavesdropping”
Eavesdropping
In eavesdropping, the
individual being
overheard puts no effort
into being overheard.
In teaching the teacher
makes a direct and
costly investment in
communicating with a
pupil.
Do Ants Teach?
The role of
the one?
…and of the many?
An individual is a teacher if …
1) it modifies its behaviour in the presence
of a naive observer,
2) at some initial cost to itself,
3) in order to set an example
4) so that the other individual can learn more
quickly.
5) Teaching also involves bidirectional feedback
between teacher and pupil.
Teaching involves
communication
between
teacher & pupil
and
vice versa.
Teaching
is
dialogue
not
monologue!
Do Ants Teach?
On teaching,
“Recently, this already well-known behaviour, informing the naïve
nestmates by tandem running about a route to a new target area,
has been referred to as a case of ‘teaching’ in ants – the leader
ant ‘the teacher’ and the follower ant ‘the pupil’.
This might be a charming metaphor, but it adds little, if anything,
to our understanding of this fascinating recruitment behaviour.”
“The Superorganism” Hȍlldobler & Wilson 2009 p. 494
Compare tandem-running with carrying…
VIDEO CLIPS
Why is tandem-running so slow?
9.6
Leader
Follower
9.4
2mm Y grid squares.
9.2
9.0
8.8
8.6
8.4
8.2
14.2
14.4
14.6
14.8
Zones of Attraction & Repulsion
Max antennal range
Head-Gaster
distance (H-G)
40
LA= 10.50 -8.86(H-G dist)
N=541, r 2 = 0.30, P< 0.001
20
Constant speed
0
-20
L
CE
DE
EL
C
AC
RESPONSE: Leader acceleration (mm/s/s) (t+1)
2mm X grid squares.
40
FA = 4.79(H-G dist) – 5.35
N=547, r 2 = 0.14, P< 0.001
20
Constant speed
0
-20
EL
EL
C
AC
C
DE
RESPONSE: Follower aceleration (mm/s/s) (t+1)
-40
SEARCHING
-40
0
Average
antennal
reach
2
4
Stimulus strength
6
8
10
head-gaster distance (mm) (t)
S T ATIONARY
Average
0 antennal 2
4
reach Stimulus strength
6
8
10
head-gaster distance (mm) (t)
H-G < antennal range
H-G = antennal range
Leader accelerates
Follower decelerates
Constant velocity
antennal range < H-G
< 2*antennal range
Leader decelerates
Follower accelerates
12
An individual is a teacher if …
1) it modifies its behaviour in the presence
of a naive observer,
2) at some initial cost to itself,
3) in order to set an example
4) so that the other individual can learn more
quickly.
5) Teaching also involves bidirectional feedback
between teacher and pupil.
Time cost- tandem and previous individual return speed
18
16
Costs of Teaching
14
Speed (mm\s)
12
10
8
6
4
2
0
`
PREV
RETURN
TANDEM
PREV
RETURN
TANDEM
PREV
RETURN
TANDEM
PREV
TANDEM
RETURN
PREV
TANDEM
RETURN
Mann-Whitney test, p< 0.0001 in all 5 cases
Leader
Follower
Why are tandems
so slow?
Follower Searches
&
Leader Pauses
For 57% of the total
journey time the
leader is stationary.
Tortuous follower search
What is the
Follower Learning?
It is learning the route more quickly.
Boxplot of Discovery time (sec) vs Type
800
Discovery time (sec)
700
Why is tandem-running so slow?
Because the follower is pausing to
up-date its positional information!
600
500
It is learning landmarks and may
also use path integration (= dead
reckoning).
400
300
200
100
0
INDIVIDUAL
TANDEM
Type
Video Clip
Leaders improve
&
followers
frequently become
new leaders.
Tandem runs
get quicker as
they become
more direct.
Do Ants Teach?
From Pratt et al.
(2005) Anim. Beh.
70: 1023-1036
An individual is a teacher if …
1) it modifies its behaviour in the presence
of a naive observer,
Do Ants Teach?
2) at some initial cost to itself,
3) in order to set an example
4) so that the other individual can learn more
quickly.
Yes!
5) Teaching also involves bidirectional feedback
between teacher and pupil.
Is there evaluation
during teaching in
ants?
T. O. Richardson, P.A. Sleeman,
J. M. McNamara, A. I. Houston,
Nigel R. Franks (2007)
Teaching with Evaluation in Ants
Current Biology 17, 1520–1526
An individual is a teacher if …
1) it modifies its behaviour in the presence
of a naive observer,
2) at some initial cost to itself,
3) in order to set an example,
4) so that the other individual can learn more
quickly.
(from Caro & Hauser 1992)
5) Teaching also involves bidirectional feedback
between teacher and pupil.
(Franks & Richardson 2006)
Alex Thornton* & Katherine McAuliffe (2006)
Teaching in Wild Meerkats
SCIENCE 313: 228-229
Alex Thornton* and Katherine McAuliffe 2006 Teaching in
Wild Meerkats SCIENCE 313: 228-229.
“Despite the obvious benefits of directed mechanisms that
facilitate the efficient transfer of skills, there is little critical
evidence for teaching in nonhuman animals. …
“[To date, only one study provides firm evidence for teaching
(Franks & Richardson 2006).]”
…we show that wild meerkats (Suricata suricatta) teach pups
prey-handling skills by providing them with opportunities to
interact with live prey.
In response to changing pup begging calls, helpers alter their
prey-provisioning methods as pups grow older, thus
accelerating learning without the use of complex cognition.”
Teaching in Wild Meerkats
Alex Thornton* and Katherine McAuliffe
D
Nigel’s view
Young pups…
Teaching in Wild Meerkats is the provision of a learning
opportunity.
but play-back
calls of old
pups.
The format of the opportunity changes in response to
pup calls, which change with pup age.
E
Old pups …
but play-back
calls of young
pups.
Meerkat teachers use a “rule of thumb” rather than
evaluation.
They base the teaching opportunity they provide on
perceived pup age – but not on actual pup skill.
The alternative view:
True teaching involves
evaluation by the teacher…
“
“… a combination of observation,
judgment and intervention”
”
(Premack & Premack 1994)
The extremist view:
True teaching involves
evaluation…
based on theories of mind
…and is uniquely human.
Theories of mind
are thoughts about the
possible thoughts of others.
“Evaluation presupposes an image or mental
representation of a standard, a preferred action
Teaching with Evaluation
Pupil
or product, to which the teacher compares
Teacher
the action or product of the novice. There is no
evidence for evaluation – standards and
correction of a deficient novice – in nonhumans.
Teaching still looks to be unique to humans.”
Marc Hauser & David Premack
(2006 pers. comm.)
Correction
Standards
A teacher might….
1) Evaluate her own efforts.
Does teaching in ants
involve evaluation?
2) Evaluate the quality of the pupil.
3) Evaluate the worth of the lesson.
Teacher Giving-Up
Time
N = 648
(=TGUT)
Experiment 1
A teacher might….
Means with 95% confidence intervals.
1) Evaluate her own efforts.
100
90
2) Evaluate the quality of the pupil.
P< 0.001
TGUT (s)
80
70
62s
60
3) Evaluate the worth of the lesson.
50
40
30
30s
20
0
20
0
40
0
60
0
80
0
100
0
120
0
140
Interrupt time (s)
Patience…
How long should the
teacher wait?
“Constancy in Exertion or Effort”
(OED)
A teacher might….
A teacher might….
1) Evaluate her own efforts.
1) Evaluate her own efforts.
The more time she has invested the
more she is prepared to wait.
2) Evaluate the quality of the pupil.
3) Evaluate the worth of the lesson.
Experiment 2
P< 0.0001
P< 0.0001
Experiment 2
TGUT
(secs)
250
200
P< 0.001
150
0.29
0.29
0.17
0.17
100
50
33
22.5
0
Follow ers_w ith_1_a ntenna
Followers with 1 antenna
Follow ers_w ith_2_antennae
Followers with 2 antennae
Slower than…
Teachers are more patient with faster pupils.
A teacher might….
Experiment 2
Note In Expt. 2…
The follower was taken away at a set point not at a set time.
Pupils with both antennae would get to that set point faster
than pupils with one antenna.
Teachers are more patient with pupils with two antennae,
even though they will have invested less time in teaching them.
This strongly suggests that here teachers are evaluating the
quality of the pupil (rather than the time invested).
2) Evaluate the rate of progress of the
lesson.
She seems to be more “patient”
with “better” pupils.
A teacher might….
1) Evaluate her own efforts.
Experiment 3
2) Evaluate the quality of the pupil.
3) Evaluate the worth of the lesson.
Summary
• Teaching ants carry out three different types of evaluation.
“Evaluation presupposes an image or mental
representation of a standard...
Teaching still looks to be unique to humans.”
• Experiment 1 suggests that the teacher evaluates her investment
in the lesson and waits longer for her pupil the longer the lesson has
proceeded.
Marc Hauser & David Premack (2006 pers. comm.)
• Experiment 2 suggests that the teacher evaluates the rate of progress of
the lesson and waits longer for better pupils.
• Experiment 3 suggests that the teacher evaluates the worth of the lesson
and waits longer when the goal has a higher value.
“Evaluation presupposes an image or mental
representation of a standard...
Teaching still looks to be unique to humans.”
T. O. Richardson, P.A. Sleeman, J. M. McNamara,
A. I. Houston, Nigel R. Franks (2007)
Teaching with Evaluation in Ants
Current Biology 17, 1520–1526
Marc Hauser & David Premack (2006 pers. comm.)
E-mail to Tom Richardson from David Premack 13 June 2007.
No !
Evaluation can involve simple thresholds.
And ants do it.
“Marvelous article of great importance.
[What] I've written on the topic in the last 20 or 30 years…
seems to have worked--provoked you, irritated you
enough to do the work I'd not been able to get biologists
to do despite years of trying.”
Summary
• Evaluative teaching can occur without theories of mind?
Ant Search Strategies
after
Interrupted Tandem Runs
• Simple rules – sophisticated outcomes.
• The first case of teaching with direct evaluation in
non-human animals.
What happens
when lessons
break down?
Nigel R. Franks, Thomas O. Richardson, Samantha Keir,
Stephen J. Inge, Frederic Bartumeus & Ana B. Sendova-Franks
Lost Leaders
&
Interrupted Lessons
What do tandem leaders do when they give-up on a follower?
3%
5%
Leaders that have lost their follower
remain stationary and then proceed
to the new nest.
Leader entered new
nest
Leader entered old
nest
Indeterminate
behaviour
92%
92% of tandem leaders, when they give-up on a pupil,
go to the new nest.
(N=37)
What do
followers
that have
lost their
leader do?
What do followers,
that have lost their leader,
do?
A spiral
search
centred on
the point of
loss?
What do tandem followers do when the tandem run is interrupted?
Tandem 38
Int 1 New Nest
22
20
18
Tandem
16
Follower
14
12
10
22
24
26
28
30
32
34
36
38
40
Tandem 18, Interrupted at Point 2, Found New Nest.
A random walk explores over
short distances much more
quickly than over long distances,
and it will explore a given region
of space rather thoroughly.
Paraphrased from H.C. Berg (1983)
In 2D, a random walker will
always return to its starting
point……………………………
……………………………………
…………………….. eventually!
Drop an immortal ant on an
infinitely large table top and it
will always return to
the starting point, eventually.
Drop a spanner in outer-space
(3D) and you may never see it
again, even if you wait for ever!
The
“Wait for Mommy”
Algorithm.
Is it better for two
“people” to look for
one another or is it
better for one to stay
where they are?
Cumulative tandem pairs reconnected at time t (%)
Friends Re-United: quickly or hardly at all.
0
100
0.05
0.1
0.5
Does the follower’s behaviour depend on investment in the tandem run?
1
0
50
Old Nest
0
100
50
90
0
100
33
50
2
2
11
11
22
3
3
180
0
1
10
1
10
1
10
1
Time (flight number)
10
1
10
A random walk explores over
short distances much more
quickly than over long distances,
and it will explore a given region
of space rather thoroughly.
Arena for Experiment 3
Interruption Points
GUT
Super-diffusion
after the leader
is likely to have
left.
Random walk
diffusion before
the leader is
likely to have
left.
GUT
GUT
Paraphrased from H.C. Berg (1983)
Super-diffusion
•
What we know…
(d) Incomplete tandem runs are of value – anytime algorithms
Direction of tandems
before interruption
Direction of
former followers
Lévy Walks ?
•
What we know…
(d) Incomplete tandem runs are of value – anytime algorithms
Direction of tandems
before interruption
Direction of
former followers
So even a little
teaching can be
useful?
Drift + Diffusion
Small Societies
When does such
teaching occur?
- 400 individuals max
- Low forager density
- Information precious & losable
- 1-2-1 communicators
Swarm raiders
The Alternative:
Broadcasting
- Large societies
- 106 individuals
- High forager density
- Pheromone broadcasters
An individual is a teacher if …
1) it modifies its behaviour in the presence
of a naive observer,
2) at some initial cost to itself,
“Tell me, I forget…
Show me, I remember…
Involve me, I understand.”
3) in order to set an example
4) so that the other individual can learn more
quickly.
(Old Chinese Proverb)
5) Teaching also involves bidirectional feedback
between teacher and pupil.