Greg Detre
Wednesday, February 12, 2003
nature vs
nurture
boesch �
most sophisticated in the wild
premack �
display cognitive skills they display under no other conditions
�the plural
of anecdote is not data�
physical
domain
object permanence � little effect
symbolic play � only anecdotal examples
object manipulation, tool use, categorisation �
some skill without contact, but more complex with exposure, observation
Andrea
talks of levels of indirection � is this kind of like Dennett�s zero-order
intentionality
Deb: being
able to take someone else�s visual perspective is a kind of prerequisition for
consciousness?
you can�t
just put anything in a human environment and expect it to �
what might
we gain from this for building curious machines?
an assumption that the
culture around you is a learning environment
social motivation to enjoy the triaed of sharing attention for its own
sake could underlie: declarative communication, joint attention, cooperation,
and teaching/learning
some basic similarity seems necessary for imitative learning and
scaffolding to work
could apes
differentiate accidents???
could you break theory of mind into purposes and beliefs, and just have
half???
kind of like the distinction between consequentialism and
intentions-based morality�
what�s a common accident for an ape???
does
everyone feel happy about what it means to say that something has theory of
mind???
set of capabilties/skills vs a kind of encapsuated module???
unitary??? sometimes demonstrate, sometimes not
passing the
false belief test needn�t require second order representation, just a large
representation of the world???
�while failure � isn�t necessarily informative about a child or animal�s conceptual abilities, success is� - Bloom and German, 2000
Deb: what would be the Occam�s razor hack for passing the false-belief test besides actually representing (second-order) someone else�s intentional beliefs?
Cynthia: well, high-functioning autistic
children seem to basically just learn a bunch of rules, like �if (the person is
in the room AND looking at the boxes) OR someone told the person�, rather than
actually representing the other person�s beliefs
I can�t
help but feel that imitation alone would not be sufficient to make the jump
towards assuming that what�s being imitated is a peer
Shettleworth:
�imitaiton is what is left when all other mechanisms of social learning have
been ruled out�
i.e. there�s something high/levelspecial about imitation � hmmm???
reference:
Call + Carpenter: Three Sources of information in social learning. In Imitation
in animals and artifacts, MIT Press 2002
primary question in defining imitation:
don�t understand + adopt goal � result doesn�t
occur � emulation
don�t understand and adopt goal � result occurs
� mimicry
understand and adopt goal � don�t get the
result right � goal emulation
understand + adopt goal � result occurs � imitation
social
learning
local enhancement � drawing your attention to the location/object
observational conditioning � when you already know a behaviour, learning
by watching a conspecific to use it under similar circumstances
mimicry
emulation
imitation
the first two get ignored, but might well be
crucial for boostrapping the last three
innate
supramodal body schema mapping movement-as-seen to movement-as-felt
this is very complex
the mapping is not only onto proprioceptive state, but also onto the motor plan or intention required to achieve it
really!!!???
it is this mapping which bridges the infant�s
external + internal worlds
predisposes the infant to think of people (�things that move like I do�)
as having internal mental states similar to its own
Meltzoff +
Moore. Imitation of facial and manual gestures by human neonates. Science, 198,
1977, pp 75-78
previous to this, Piaget had claimed that infants don�t do facial
imitation for at least 12 months
can imitate facial gestures as early as 42 minutes after birth
�I have difficulty learning new motor skills 42 minutes after getting
off a plane�� � Derek
�it doesn�t get much more innate than that� � Meltzoff + Gopnik
they were pretty careful to avoid just global
arousal � how???
could the experimenter being unwittingly
providing feedback to the infant biasing its behaviour?
second set of experiments � baby has pacifier � experimenter makes faces
� then waits with blank face � pacifier taken out 2 mins later � imitation
still occurs � this means that the imitation is from memory, and it�s not
reflexive
attack by definition � is this really �imitation�?
the infant doesn�t understand the goal
is
proprioception as innate too then???
What could
the mechanism be???
1997 paper
� 20 years later
3 key components for any explanatory mechanism to have
key observation:
after demonstration, observe quieting of general bodily movement
the infant isolates what to move before determining how to
move it
theory:
the infant has a small set of organs which it can recognise and map at
birth on the basis of their form � innate neural mechanism
support:
visual display of faces/hands leads to activation in monkeys
body babbling:
definition:
the infant�s movement of its limbs + facial organs in body play is analogous to vocal babbling
theory:
the process of body babbling allows the infant to learn mapping between motor plans and resulting organ relation end-states
this could even start to begin in
utero
iterative process of approximation
microanalysis of imiatating lateral tongue movement
might we see the progenitor of the innate supramodal body schema in apes?
Bard + Russell � duplicated the class Meltzoff + Moore study of facial imitation using 24-72 hour old chimp infants
imitative capabilities are found to be very similar to those of human subjects
this doesn�t quite fit with Call + Tomasello
but wouldn�t you expect it to lead further towards theory of mind than chimps get???
could we implement Meltzoff + Moore�s mechanistic handwaving?
would it provide the catalytic support for social learning?
the correspondence problem � extend this to dissimilar bodies?
dolphins can mimic human actions, high-level analogising rather than one-to-one body parts
infants can emulate both static + dynamic facial gestures
mirror neurons
motor cortex � if you see another primate do an action, they fire when you do it
presumably, the bit doing the recognition is the same as the system doing the generation
�pole-vaulting � the goal is to get over the bar without landing in hospital� � Derek
how could there be body babbling with facial movements???
how could there be an iterative process without feedback???
after all, deaf children don�t learn to speak�
do infants touch their faces while moving them???
you�ve got proprioception
I just don�t believe it could be hard-wired
can they do the imitation with non-facial bits??? e.g. if you hide their toes, can they wiggle them to imitate an experimenter
�understand and adopt goal� � what is �goal�?
hidden states?
surely the goal is the evaluation criteria??? too high level
how do these develop in perceptually-impaired infants?
why would the initial imitative capacity be there adaptively if it atrophies without use???
why is �emulation� not understanding/adopting the goal AND not copying the action successfully???
deaf babies babble too � and they follow a pretty standard progression in terms of the facial movements they make