Template

1
Socializing Transfer
Daniel Schwartz School of Education Stanford
University
2
What I talked about last timeTransfer literature

• Detterman from Transfer on Trial.
• most studies fail to find transfer and those
  studies claiming transfer can only be said to
  have found transfer by the most generous of
  criteria and would not meet the classical
  definition of transfer.
• Classic stimulus generalization view --
  replication of old behavior in a new situation.
• Plays into educational and psychological
  literature emphases on efficiency.
• Faster and more accurate retrieval and
  application of previously learned behaviors.

3
Optimal Corridor of Instruction
Annoying Novice
Adaptive Expert
PFL
Routine Expert
SPS
Novice
Efficiency
4
Preparation for Future Learning Measures(SPS v.
PFL)
Innovation Activities
Efficiency Activities
Target Transfer Problem
5
The Effect on the PERC Audience?
6
Decided to give a different kind of talk.

• Poster session with Jose Mestre summarizes recent
  transfer findings.
• Continue theme of learning v. problem solving
  with two micro-findings that have just been
  waiting for a captive audience.
• In this case, not learning at test versus problem
  solving at test.
• Discuss cognitive science more broadly.

7
Learning is not the Same Thing as Problem Solving.

• Daniel L. Schwartz
• Stanford University

8
Some General Cognitive Science Contributions

• Multi-disciplinary
• Verbal protocols and discourse analyses
• Careful descriptions of problem solving
• Formal, executable models of cognitive process.

9
Some General Risks ofCognitive Science for
Education

• Separation of higher order cognition from
  important factors in learning.
• Motivation rarely appears in cognitive science
  journals.
• Use of problem solving models to explain
  learning.
• Search, schemas, mental models, goal
  decomposition, analogical reasoning, working
  memory, cognitive load come from problem solving
  research not learning research.
• The fact that I achieve a correct answer in
  problem solving (when I could not before), does
  not entail I have learned.
• Confusions between symbolic models of thought and
  thought itself.
• Transfer often described as an equivalence
  detection (X Y), instead of resonance (for
  example).
• All told, the picture that often gets transmitted
  to education emphasizes verbal (symbolic) problem
  solving, which may not do justice to learning.

10
Not true of all cognitive science.

• Separation of verbal problem solving from other
  mental systems, perception and affect, does not
  work so well for learning.
• The (problem-solving) talk may only reveal one
  part of what enables the (learning) walk.
• There are other styles of cognitive science.
• Describe two (utterly unrelated and
  self-indulgent) instances of basic, cognitive
  science research that may nevertheless address
  important educational issues.
• Non-verbal outcomes of symbolic, rule learning.
• Non-verbal processes regulate verbal learning.

11
Non-Verbal Outcomes of Verbal Rules (w/ Sashank
Varma)

• How do people build new knowledge on a foundation
  of old knowledge.
• Prevalent assumption people develop well-formed
  abstractions from perceptual representations.
• Embodied cognition
• P-prims
• Sensory motor ? Concrete Operations ? Formal
  Thought
• Enactive ? Iconic ? Abstract
• Propose that people can start with symbolically
  expressible rules that evolve into
  perception-like representations.

12
The case of mathematics

• Some relevance to physics, because mathematics
  play an important role in structuring physics
  understanding.
• Take a simpler case than physics.
• Building the integers on top of natural numbers.

13
What do we know about representations of natural
numbers?

• Peoples representations of natural numbers has a
  perception-like substrate.
• Evidence?
• The symbolic distance effect (SDE)
• Neuroscience correlations.

14
Symbolic Distance Effect

• Perceptual discrimination has a well-known curve.
• Near perceptual referents take longer to
  discriminate than far perceptual referents.
• Hearing
• Loud sound v. soft sound. (Fast)
• Softer sound v. soft sound. (Slow)
• Loud sound v. louder sound. (Slower)
• Effect is general it applies to nearly all
  perceptual magnitude judgments
• Saltiness, brightness, firmness, heat, weight

15
(No Transcript)
16
Ready
17

18
2 8
19
2 8
20
Correct
21
Natural numbers exhibit the same curve.

• Moyer and Landauer (1967)
• 2 v. 9 (Fast)
• 2 v. 3 (Slow)
• 8 v. 9 (Slower)
• This finding was a big deal.
• A purely symbolic task was processed
  perceptually.
• Cognitive-neuroscience has investigated this
  effect.

22
Neuroscience findings

• IPS activation correlated with SDE.
• IPS active for perceptual comparisons.
• IPS implicated in spatial reasoning (e.g.,
  rotation)
• IPS activates strongly for subtraction (not
  multiplication).

23

• Led to proposal that the representation of
  natural number uses a perception-like number
  line.
• So, what does the adult representation of
  integers look like?

24
Adults (n21)
Far
Near
Positive
1 v. 8
1 v. 3, 6 v. 8
Negative
-1 v. -8
-1 v. -3, - 6 v -8
Mixed
-2 v. 6
-2 v. 1

• Other factors for balancing the stimuli
• Larger number on right/left.
• Choose larger/smaller of numbers

25
Give it an introspective try.

• A short series of problems.
• See if you can detect the effect.
• Remember SDE predicts
• Near is slow.
• Far is fast.

26

27
1 9
28

29
2 3
30

31
3 2
32

33
-4 -2
34

35
-2 8
36

37
1 -2
38
Howd you do?

• Can you predict study outcomes?
• Hard to introspect because the difference between
  near and far is around 0.05 sec.
• Sometimes verbal reports of ones own problem
  solving do not work so well.

39
Distance Effect
40
Inverse Distance Effect
41
Summary of Results

• Within-in class comparisons show SDE
• Negatives increase by a constant (flip rule)
• Between-class comparisons show inverse SDE.
• What might explain this effect?

42
Perceptualizing Integers

• Integer comparison may have recruited another
  general perceptual mechanism.
• Categorical Comparison
• People are very fast at comparisons that fall on
  either side of a perceptual boundary.
• English speakers distinguish pa and ba faster
  than non-native speakers.
• Color comparisons across boundaries are faster
  than within boundary.
• Negative and positives have developed a boundary,
  and people are very fast at comparisons close to,
  and on either side of, the boundary.

43
Reflection not extension of the number line.
far 6 v. -2
near 1 v. -2
1
-2
44
Kids?

• Natural numbers.
• 7th-graders look like adults.
• Looked at rising 7th-graders.
• Integers introduced by 4th-grade.
• 2 years practice.
• Equal accuracy as adults.

Sekuler Mierkiewicz (1977)
45
Distance Effect
Adults
6th Graders
46
Distance Effect
Adults
6th Graders
47
Summary of Adults vs. 7th Graders

• 7th-graders same as adults for within-class
  comparisons.
• 7th-graders show no inverse distance effect
• They used rules to reason about negative and
  positives.
• If it has a negative sign, it is less.
• Ignore magnitudes.

48
Discussion

• Guiding hypothesis
• People use rules that become perceptualized.
• Needs more (cleverer) research.
• Still, a useful lesson.
• Negative numbers were taught with intuitive
  representations that made sense.
• At the same time, kids learned meaningful rules
  for managing integers.
• Over time (into adulthood) these rules changed
  underlying form.

49

• Relevant to physics.
• Intuitions and perceptual experiences are
  important for instruction.
• Also need mathematical rules, so it is possible
  to reason in more complex ways.
• It may take many years before these rules
  transform into immediately meaningful perceptual
  phenomena.
• One hypothesis is that the rules are necessary
  way station.
• Physics for non-majors seems fine, but they will
  never develop the base representations that make
  them see the world as physicists.
• When they see that ball rising in the air, they
  need to use rules to separate forces and
  velocity. You probably do not. You perceive
  the separation.

50
So, what do you think of cognitive science of
learning?
51
2. Non-verbal processes regulate verbal learning.

• Reversed the usual story we move to perception,
  not just from it.
• But, no learning process in this story.
• Need shorter-term learning to get process data.
• Switch to a completely different topic.
• More process data.
• More juicy cognition, too.

52
Social interactivity and Learning

• Value of social interaction for learning is often
  ascribed to information for improved problem
  solving
• Timely feedback.
• Introduction of alternative ideas.
• Increased elaboration.
• Perhaps there is something non-verbal,
  non-problem solving too.
• How to find out?
• Hold social interaction constant (no extra
  information).
• Manipulate peoples beliefs about socialness.

53
The mere belief of social.(w/ Sandra Okita and
Jeremy Bailenson)

• Participants read a passage about mechanisms of
  fever.
• Participants interacted with a computer character
  in VR.
• Told that character on other side was
• Agent Condition A computer agent
• Avatar Condition A person they just met
• Interaction held constant within and across both
  conditions.
• Read question about mechanisms of fever from
  monitor.
• Listened to canned responses.
• Received posttest on questions from VR and novel
  questions

Sandra Okita
54

• Avatar Condition
• Now you will be going into the VR environment
  to interact with Alyssa.
• You will be asking questions, and Alyssa
  will be answering from the other room. Please
  call to Alyssa first, and then read the question
  aloud.
• Agent Condition
• Now you will be going into the VR environment
  to interact with a VR character that is a
  computer program.
• You will be asking questions, and the computer
  program will be answering.
• Please call out computer program first, and
  then read the question aloud.
• In reality, it was always a computer program that
  simply played identical pre-recorded responses in
  both cases.

55
Flow of exchange.
56
Performance on a posttest administered outside VR
Quality of Answer Heard in VR.
57
Higher arousal if believed a person.
58
Arousal and Learning
59
Summary

• The mere belief of a social improves learning
• Even if no differences in information exchange.
• The mere belief of social increases arousal.
• Arousal is correlated with learning outcomes.

60
Study 2

• The mere belief of social
• Or, The mere belief of social action
• Just finished a second study.
• Replicate Avatar condition
• Replicate Agent condition
• New Silent-Avatar condition

61
Silent Avatar Condition
Read Question Silently
Alyssa is reading same question.
Remove social action.
62
(No Transcript)
63
(No Transcript)
64
(No Transcript)
65
(No Transcript)
66
(No Transcript)
67
Summary

• Have not analyzed arousal data yet. (Sorry.)
• One motivation for lectures over books is that
  there is a person speaking.
• Evidently, a talking head buys you factual
  knowledge, but not deep understanding.

68
Discussion

• Common argument that people do not learn well by
  being told.
• Many times, there is an appeal to constructivism
  you cannot just transmit knowledge into
  peoples heads.
• But constructivism is a general theory of
  knowledge growth that applies whether actively
  doing or quietly listening.
• A simpler explanation might be that people simply
  are not very aroused when listening to a lecture.
  There is no belief in the possibility of social
  action.
• Of course, there are other things too. But,
  remember... the non-verbal stuff is important.

69
Conclusion

• Cognitive science, when applied to education,
  often describes learning as a verbal process of
  managing declarative information to solve
  problems.
• The self-explanation effect Reducing cognitive
  load.
• Tried to show a side of cognitive science that
  does not exclusively focus on the verbal
  processes of problem solving.
• Talking the talk does not equal walking the walk.
• Other things going on under the hood are
  important for understanding and improving
  learning.

70

• Tried to demonstrate distinction between learning
  and problem solving using (ironically) verbal
  learning
• Children learn verbal rules for handling
  negatives, give same explanations as adults, but
  these rules evolved into perception-like
  representations underneath.
• Adults non-verbal responses (arousal) at learning
  are correlated with their subsequent verbal
  understanding.
• Problem solving protocols are hugely important,
  but we need to find ways to get beyond the
  assertion that good problem solving leads to good
  learning.
• One approach is to examine what people see (e.g.,
  Jose).
• Another approach is to document the concomitants
  of change.

71

• Need to consider models of learning.
• What is an example of a learning model?
• Not too many in cognitive science.
• Declarative ? Procedural.
• Connectionist Systems
• One of the appeals of brain science is that it
  brings learning back (i.e., plasticity).
• Amygdala, MTL, and Coordination.
• I hope there was at least one thing useful in the
  talk.
• Thank you.