Cognitive Development 142 How do children read minds Theory of Mind and inferences about peoples beh

1
Cognitive Development (142) How do children read
minds? Theory of Mind and inferences about
peoples behavior

• Professor Leda Cosmides
• TA Andy Delton
• Office hours
• Cosmides Tuesdays, 1130-130pm, HSSB 1010
• Delton Mondays, 100-300 HSSB 1010
  (delton_at_psych.ucsb.edu)
• Course website http//mentor.lscf.ucsb.edu/course
  /fall/psyc142/
• E-res password clutch

2
Are instincts the opposite of reasoning?

• A common view
• Animals have instincts
• Humans do not
• Evolution erased our instincts
• Replaced them with a capacity for culture

3
But what, computationally, is a capacity for
culture?

• A single program that causes learning, using
  domain-general methods?
• Domain-general operates on information from any
  domain (any subject matter)
• Procedures are content-free doesnt matter what
  you are learning/reasoning about
• E.g., logical reasoning operant conditioning
• OR

4
But what, computationally, is a capacity for
culture?

• A set of programs, each specialized for learning
  about a different domain?
• Do we have reasoning instincts?
• Equipped with domain-specific procedures
• i.e., ones that generate useful inferences in one
  domain, even if they do not apply to another?
• Rocks versus humans
• What does cognitive development tell us?

5
Reasoning instincts

• Complexly specialized for solving an adaptive
  problem
• Reliably develop in all normal human beings
• Develop without any conscious effort
• Develop without any formal instruction
• Applied without awareness of their underlying
  logic
• Distinct from more general abilities to process
  information or behave intelligently
• after Pinker, 1994

6
Charlie task (Baron-Cohen, 1995)
7
Instinct blindness!

• As a species, we have been blind to the existence
  of these instincts
• Not because we lack reasoning instincts, but
  because they work so well
• Process information effortlessly automatically
• their operation is unnoticed, background
• Structure our thought so powerfully
• Cant imagine how things could be otherwise
• We take normal behavior for granted
• Dont realize it needs to be explained
• Mindreading is an example!

8
Our Folk Theory of Mind
Dualist ontology mental things different from
physical ones. (e.g., can touch cookie, but not
dream of cookie)
Belief-desire reasoning
9
Theory of mind / Mindreading

• Normal inferences that seem to require no
  explanation
• Folk psychology, intuitive psychology
• Explain behavior as resulting from beliefs,
  desires, intentions
• Dan Dennett intuitive modes of explanation
• The intentional stance
• The physical stance
• The design stance (tools)

10
Mindreading system

• Component parts
• ID Intentionality Detector
• EDD Eye-Direction Detector
• SAM Shared Attention Mechanism
• ToMM Theory of Mind Mechanism

11
Stimuli with self-propulsion direction
Eye-like stimuli
ID
EDD
Dyadic representations (desire, goal)
Dyadic represent-ations (see)
Triadic representations
SAM
Full range of mental state concepts, expressed in
M-representations
Knowledge of the mental, stored used as a theory
ToMM
12
Intentionality Detector (ID)

• Function
• to identify something as an AGENT, on the basis
  of perceptual input
• interprets data as indicating goals, desires
• Input data
• Source any modality (blind)
• Cues self-propelled motion nonrandom sounds,
  shape irrelevant
• Output Dyadic representations
• (e.g., Agent - wants - X Agent - has goal - X)

13
Evidence Intentionality Detector (ID)

• infants distinguish give and tease (change in
  adults goal)
• Heider Simmel (1944) triangles
• Cells that selectively respond to animal facing
  forward (even when in profile) goal attribution
  (superior temporal sulcus)
• Cells that fire selectively to tactile
  stimulation from agent other than self.
• Focal brain damage in humans lose specific
  ability to categorize things as animate vs.
  inanimate. (Dissociation from other parts of
  cognitive system)

14
Imitation computing intentions in infancy?

• Meltzoff, showed that newborn infants were
  capable of producing a range of responses to
  gestures modeled by an actor.

15
Representing intentions

• Older infants can also imitate intentions of
  actors, even when they dont see the completed
  action. They do not imitate machines.

16
Goals pursued rationally?

• Does the system that infers goals/intentions
  assume that goals are pursued in a rational
  manner?
• E.g., take the most direct path to a goal
• Gergely et al
• 12-14 month olds
• Habituation method

17
Habituated On
18
Tested On
Surprised
19
Tested On
Not Surprised
12-14 month olds
20
Eye-Direction Detector (EDD)

• Function
• To detect presence of eyes or eye-like stimuli
• If present, to compute whether those eyes are
  looking at me versus looking at not-me
• To infer that If another organisms eyes are
  directed at X, then it sees X
• Output representations
• Agent sees X.
• From mutual eye contact Agent sees me
• or
  I see agent

21
Eye direction detector (EDD)

• Input data
• Source Vision only
• Eyes in context of face
• white/dark, angle of dark
• Evidence
• 2-months look almost as long at face with eyes
  only as at whole face, but less at face with
  other parts, but no eyes
• when breast feeding, mother gazes at infant for
  very long durations (gt 30 secs)
• 6 months look 2x - 3x as long at a face looking
  at them than at a face looking away

22
Eye direction detector (EDD)

• Evidence
• 3 year olds computation of eye direction easy
  (Which photo is looking at you? probably
  during infancy too)
• Mutual eye contact (looking at me) increases
  GSR and brain stem activity triggers smiling in
  infants
• Infants can regulate amount of eye contact
  regulate arousal
• Peekaboo (game occluding/revealing eyes)
• Eye-direction detection known to be important in
  wide variety of animals, especially in assessing
  predation threat or aggression

23
(No Transcript)
24
Stimuli with self-propulsion direction
Eye-like stimuli
ID
EDD
Dyadic representations (desire, goal)
Dyadic represent-ations (see)
Triadic representations
SAM
Full range of mental state concepts, expressed in
M-representations
Knowledge of the mental, stored used as a theory
ToMM
25
Shared Attention Mechanism (SAM)

• Function
• To build tryadic representations. Relations
  among
• Agent
• Self
• (Third) Object
• I.e., makes you aware of shared universe
• Comparator, fusing dyadic representation of own
  perceptual state with dyadic representation of
  others perceptual state (You see that I see the
  bus)
• Makes output of EDD available to ID

26
Shared Attention Mechanism (SAM)

• Input data
• Representations ID produces as output
• Representations EDD produces as output
• (especially those from gaze monitoring)
• Requires information about another organisms
  perceptual state own perceptual state

27
Shared Attention Mechanism (SAM)

• Output
• Representations like this
• Agent - Relation - (Agent - Relation -
  Proposition)
• Mommy - sees - (I - see - the bus)
• Agent can be Self
• tryadic representation built from EDD data can be
  fed back into ID, then back into SAM
• Mommy sees that (I want the cookie)
• Mommy wants that (I look at the bus)
• Mommy sees that (I am refering to the bus)

28
Evidence Shared Attention Mechanism (SAM)

• 9 months (everywhere by 14 months) gaze
  monitoring (infant turns in same direction that
  another person is looking, then shows gaze
  alternation, checking back and forth a few times,
  as if to make sure that infant and other are
  looking at same thing.
• Same age Protodeclarative pointing gesture
  outstretched index finger at object, then gaze
  alternation between other and object pointed at.
• Showing game

29
Evidence Shared Attention Mechanism (SAM)

• 9-18 months Ambiguous action causes instant look
  to adults eyes unambiguous action does not.
• when goal of an action is uncertain, first place
  child looks for information to disambiguate is
  persons eyes. (Amb adult cups hands over
  childs when child doing something manual adult
  offers then withdraws as child reaches. Unamb
  give or present object to the child.)
• Toddlers bring objects into other persons line
  of regard
• Charlie experiment infering mental state from
  eye direction (arrow)

30
Charlie task (Baron-Cohen, 1995)
31
Stimuli with self-propulsion direction
Eye-like stimuli
ID
EDD
Dyadic representations (desire, goal)
Dyadic represent-ations (see)
Triadic representations
SAM
Full range of mental state concepts, expressed in
M-representations
Knowledge of the mental, stored used as a theory
ToMM
32
Theory of Mind Mechanism (ToMM)

• Function
• To represent epistemic mental states
• E.g., pretending, thinking, knowing, believing,
  imagining, dreaming, guessing, deceiving
• To tie together all mental state concepts --
  volitional (ID), perceptual (e.g., EDD), and
  epistemic into coherent understanding of how
  mental states and actions are related.
• Suspends semantic relations of truth, reference,
  and existence
• I.e., ToMM represents epistemic mental states and
  turns all the mentalistic knowledge into a useful
  theory

33
Theory of Mind Mechanism (ToMM)

• Input data
• Output of SAM
• Output from ToMM
• Meta-representations (M-representations), with 3
  slots
• Agent - Attitude - Proposition
• Mommy - is pretending - that the banana is a
  telephone
• Dad - believes - it is raining outside
• I - hope - we go to Disneyland

34
Theory of Mind Mechanism (ToMM)
An M-representation

• File folder in the mind
• Propositions decoupled from semantic memory

35
Evidence, ToMM

• 18-24 months begin to pretend, recognize when
  others are pretending
• 36-48 months evidence of understanding
  additional epistemic states, e.g., knowing,
  seeing leads to knowing.
• Universal pattern
• 3 years fail false belief task
• Performance limitations (inhibitory processing,
  etc)
• 4 years pass false belief task
• Autism (false belief, false photograph, charlie)
• Not performance limitations

36
Understanding pretend 18-24 months

• Mom believes it is a banana, but is acting as if
  it is a telephone
• Child is representing a representation in Moms
  head! (has a meta-representation)
• Alan Leslie, 1987 What does this imply about the
  mind?

37
M-representations

• Suspend truth relations
• 1The banana is a telephone versus
• 2 Mom is pretending that the banana is a
  telephone.
• 1 leads to inferences that 2 does not (e.g.,
  1 implies there are yellow edible phones, 2
  does not.
• Decouple the embedded proposition from semantic
  memory
• Dont store info about telephones as edible,
  soft, yellow
• Restrict the scope of application of inferences

38
Ability to understand pretend play at 18 months
implies the ability to understand beliefs and
form M-representations

• there appears to be a failure to understand an
  important aspect of mental life until about age
  4.

39
Belief-desire reasoning the false belief task
Where was the marble in the beginning? Where is
the marble now? Where will Sally look for her
marble?
4-year olds pass majority of 3-year-olds fail
40
People with autism also fail false belief task
Baron-Cohen, Leslie, Frith, 1985
41
Yet

• Children with autism pass the false photograph
  test
• Same task demands
• Must understand that photograph is out of dateit
  is a false representation of the world
• Understand physical representations
• Photographs, maps
• Do not understand mental representations
• Leslie Thaiss, 1992

42
Smarties task
When I first showed you the tube, before we took
the lid off, what did you think was in here?
Your friend Sam is outside, if I show him the
tube, what will he think is inside
4-year-olds pass majority of 3-year-olds fail
43
Appearance-reality task
shown object that looks like a brick
44
what does it look like? what is it really?
45
Why do 3 year olds fail?explaining the shift in
false belief reasoning

• Alternative theories
• They are still learning theory of mind concepts
  from the environment (exposure to siblings,
  etc.).
• Or
• Their theory of mind concepts do not change,
  only the ability to express the knowledge.
• Competence-performance distinction

46
1. Discontinuity in theory of mind concepts?
belief-desire theory
desire theory
2 years
3 years
4 years
47
2. Improvement in processing capacity?
belief tasks
desire tasks
2 years
3 years
4 years
48
What demands are there in the false belief task?
Must hold 2 representations at once (reality,
false belief) but inhibit the representation of
reality!
49
Reducing inhibitory demands the look first
procedure
Where was the marble in the beginning? Where is
the marble now? Where will Sally look FIRST for
her marble?
50
The look first procedure
51
Reducing performance demands the posting
procedure
52
The posting procedure
53
15 month olds?? (remember pretend?)


54
True belief taskBelief induction trials
55
False belief taskBelief induction trials
56
Results
57
Therefore what?

• Reducing inhibitory demands improves performance
  of 3-year-olds
• Increasing inhibitory demands decreases
  performance of 4 and 5 year olds!
• 2 year olds engage in pretend play
• How, without understanding mental states?
• 15 month olds pass an implicit false belief task
• Violation of expectation method
• Concept of belief present from 15 months, but
  cant display knowledge in the standard task
• Competence present performance limitations

58
Do people with autism fail the false belief task
for the same reason that 3-year-olds do?

• Reducing task demands (e.g., demands on
  inhibitory processing)
• Helps 3 year olds
• Does not help people with autism
• False photograph task
• 3 year olds fail it
• Same demands on inhibitory processing as false
  belief
• People with autism pass it
• Autism lack the concept of belief
• Competence difference, not performance

59
What is autism?
genes, brain damage
symptoms, diagnosis, incidence
60
Incidence

• initial estimates at about 4 in every 10,000 live
  births
• more recently incidence estimated to be much
  higher as many as 9 in every 1,000 (Wing, 1997).
• heightened awareness?
• more higher functioning cases diagnosed?
• older mothers? (clumpy distribution inconsistent
  with vaccines)
• more prevalent in boys rather than girls (31)
• For Aspergers syndrome the ratio is 101

61
Autism at the behavioral level?
62
Core features of autism triad of impairments
63
Features not universal to autism

• idiot savant abilities (calculation, drawing,
  music) in about 1 in 10 children with autism.
  More prevalent in autism than in population as a
  whole, but not present in most cases of autism.
• rocking, self-injury related behaviors, other
  motor stereotypies can be found in others with
  severe mental handicap

64
Diagnostic criteria
65
triad of impairments caused by failure in
theory of mind?
theory of mind
66
Representing epistemic states allows what?

• Pretend play
• Understanding that other people have beliefs and
  other knowledge states, can pretend, lie,
  deceive, be wrong
• w/o, how do you explain their behavior?
• w/o, how do you understand enough to interact
  socially?
• w/o, how do you communicate?
• Inference problem in communication

67
Theory of Mind Mechanism (ToMM)

• File folder in the mind
• Propositions decoupled from semantic memory

68
M-representations

• Suspend truth relations
• Decouple the embedded proposition from semantic
  memory
• Restrict the scope of application of inferences

69
Truth relations

• Propositions stand in certain relationships to
  one another, such as contradiction, equivalence,
  or mutual consistency. E.g.
• The Toblerone is in the box implies
• Existence There is a Toblerone.
• Truth The Toblerone is not outside the box.
• Reference A chocolote is in the box.
• Because Toblerone bars are chocolates

70
M-representations suspend truth relations

• Embedding a proposition within an
  M-representation takes it out of circulation
  by suspending these truth relations.
• Existence
• Truth
• Reference (substitutability)

71
M-representations suspend truth relations

• Nike believes the Toblerone is in the box
• can be a true statement even if, unbeknownst to
  Nike
• 1. Someone has already eaten the Toblerone
• Existence relation suspended
• 2. The Toblerone is not in the box
• Truth relation suspended
• 3. Moreover, because Nike may not realize that a
  Toblerone is a chocolate bar, Nike believes the
  Toblerone is in the box need not imply
• Nike believes a chocolate is in the box
• I.e., substituting chocolate for Toblerone is no
  longer truth preserving
• Reference relation suspended.

72
M-representations restrict the scope of inferences

• The cup is full.
• The empty cup is full.
• I pretend the empty cup is full.
• I pretend the cup is both empty and full.
• Why are 1 and 3 sensible, but 2 and 4 strange?
• Decoupling creates an extra level within the
  representation... Inference mechanisms respect
  the levels and apply to them one at a time. --
  Leslie Frith, 1990, p. 129.
• Upstairs
• Downstairs

73
M-representations restrict scope of inference

• 3. I pretend the empty cup is full.
• 4. I pretend the cup is both empty and full.
• Translation of 3
• I pretend of the empty cup it is full
• upstairs I pretend of the empty cup X (no
  contradiction detected)
• downstairs it is full (no contradiction
  detected)
• Translation of 4
• I pretend of the cup it is both empty and full.
• upstairs I pretend of the cup X (no
  contradiction detected)
• downstairs it is both empty and full.
  CONTRADICTION!

74
Decoupling propositions from semantic memory

• Semantic memory your mental encyclopedia of
  knowledge
• Decoupling a proposition keeping it in a mental
  file folder (e.g., the M-rep), separate from
  semantic memory
• When proposition is decoupled from semantic
  memory, inferences can be made about the content
  of an agents mental states without
• this content conflicting with info stored in
  semantic memory or
• This content being stored in semantic memory as
  true

75
Decoupling propositions from semantic memory

• Decoupling allows
• representation of false beliefs
• Counterfactual, suppositional reasoning
• Decoupling prevents data corruption,
  representation abuse
• Keep encyclopedia of knowledge, stored in
  semantic memory, (relatively) free of false
  information
• Inferences within semantic memory true enough,
  can mate promiscuously, creating further
  inferences

76
Within semantic memory

• Through inferences, propositions can mate
  promiscuously, producing new propositions in
  semantic memory
• Mercury is a poison
• Tuna has mercury in it
• Therefore Tuna has poison in it
• Deriving true propositions from true premises.
• Only useful if premises are true
• Design feature only retire a proposition to
  semantic memory if it is true enough
• Otherwise, store in an M-representation

77
Imagination and decoupling Impaired in autism?
78

• Draw an impossible house, a house that could not
  exist
• Draw an impossible man, a man that could not
  exist
• Fiona Scott Simon Baron-Cohen 1996
• Imagining real unreal things Evidence of a
  dissociation in autism J. of Cognitive
  Neuroscience 8(4) 371-382

79
(No Transcript)
80
Spontaneous versus Instructed drawing

• Spontaneous drawing
• Draw something frightening, really scary
• Instructed drawing
• Draw a spider / snake
• Draw a 2-headed monster
• Draw some big teeth
• Draw another monster head on body
• Draw 2 horns...

81
(No Transcript)
82
There is a close relationship between

• decoupling,
• source monitoring (who said it? who did it?
  Source of info)
• agents attitude,
• memory tags (source, time, place)
• High level cognitive abilities in humans
• When is decoupling from semantic memory
  necessary?
• When is source monitoring necessary?
• When is it necessary to store an agents
  attitude?
• When stored in memory, does the representation
  need a source tag? a time tag? a place tag?
• The answer to 1-4 is yes under a number of
  circumstances
• These circumstances include, but are not limited
  to, modeling of other peoples beliefs.

83
Decoupling, source monitoring, attitudes, and
memory tags (source, time, place) needed for
representations of

• Goals
• Plans
• Simulations of physical world
• Episodic memories
• What you experienced personally
• What someone else told you (credal value?)
• Own beliefs when these are yet confirmed
• Own beliefs when their truth is in question
• Other peoples beliefs
• Simulations of social interactions that have not
  (yet) happened (Is there a social working
  memory?)
• Fiction
• Dreams

84
Questions to think about throughout 142...

• What does the child know about the world?
• How does the child come to know what she knows?
• Is the childs mind different from the adults
  mind, or does the child just know less?
• Does the child come factory equipped with any
  knowledge of the world?

85
Questions to think about throughout 142...

• How does the environment affect development?
• How does maturation affect development?
• Why did scientists underestimate how much infants
  know?
• What is the competence/ performance distinction?
• Can one part of the brain know something that
  another part of the brain does not know?

86
Questions to think about throughout 142...

• What is the difference between studying natural
  competences and side-effects?
• What does learning mean?
• How many learning processes are there?
• Is instinct the opposite of learning?
• What is the design of the instinct that causes
  learning in a given domain?