PowerPoint-Präsentation

Cognitive Psychology
Spring 2005
-Review SessionPascal Wallisch
University of Chicago
Overview
•Final logistics (5 minutes)
•Study tips (5 minutes)
•Some notes on the class as such (5 minutes)
•Overview and overarching issues (5 minutes)
•Review of the 10 topics (4 minutes each = 40 minutes)
The final
•Is on Wednesday, 06/08/2005, 1.30-3.30 pm
•Is 2 hours long (duh)
•Will be proctored by me (and other TAs)
•Is in TBA
The final
•Contains 12 essay questions, 11 of which have to be
done…
•Contains 20 MC questions, all of which have to be
done.
•Is comprehensive, with an emphasis on the things
that were not covered yet. Things from the TAs
lectures will ALSO be on the exam.
Exam-taking tips derived from this
• Be as concise as possible, no narratives please.
• “Essay question” is a trick term. No one expects you
to write an essay. In fact, don’t do that.
• One-or two-sentence answers to sub-questions (see
my “Sample Solutions” for the midterm for that).
Don’t take this as an encouragement to omit things.
• Make sure to answer the questions as focused and
relevant as possible.
• Save the BS/Shotgun-approach as a last resort.
• Don’t voluntarily give up your exam. Wait until I ask
you to hand it in.
Study tips
• Be sure to be able to understand all concepts that are
mentioned in the lecture notes. Be able to write a short
paragraph for each one of them.
• Try to identify underlying themes in the course
materials. There are some underlying threads running
through them. I’m pretty sure that she will ask for them.
• Go over the two midterms and make sure you
understand what was asked there. It’s not impossible
that lightning strikes twice
• Pay particular attention to Language, Problem Solving
and Reasoning, since these weren’t covered yet.
• Go over your QALMRIs. The papers are important.
How I would study
• Go through the whole book.
• Summarize each piece of information in terms of a question. (and an
answer)
• On Study-cards or on a PC.
• The book has 580 pages. This kind of information can usually be
compressed in a ratio of 1:5.
• You will end up with a ~100 page summary (on a PC).
• Learn that by heart. Questions and answers, until you know the answers
to all questions.
• If time is of the issue: Go through the lecture notes and do the above only
for topics that are on the lecture notes (from the book) and don’t do it for
things that were explicitly covered on the midterms.
• That’s how I would do it. It works. I developed this method studying for
the german undergraduate comprehensive exams. To get a degree.
• That were 4000 pages of information that I compressed on 1500 pages
(low factor because of dense information). I memorized that. And do it
since. It never failed.
• Most of the learning occurs in MAKING the summary because one does
deep processing (needs to understand it in order to summarize it), does
motor processing and puts it into the appropriate format.
Finally…
• Utilize the John Henry effect
(Google this)
Overarching issues
• The concept of a mental representation
• The reduction of computational complexity
• Cogpsy is more about methods rather than
issues
Mental representations
• Mediate between Stimulus and Behavior
• We don’t act on a stimulus, we act on our mental
representation of a stimulus.
• We don’t perceive a stimulus, we perceive a mental
representation of a stimulus.
• Mental representations have different properties in
Perception, Imagery, Memory, Categorization, Language,
Decision making, Reasoning, etc.
• Mental representations are crucial for success, e.g.
Problem solving, Creativity.
• Often issues in cognitive psychology revolve around the
nature and the properties of mental representations, e.g.
the Imagery debate.
Cognitive Complexity
• The world is complex. So are our theoretical options
to behave. We need some way to reduce this
complexity in order to behave at all.
• Selective mental representation of certain information
but not others is a way to reduce computational
complexity
• Many cognitive processes are implicitly or explicitly
aimed at the reduction of computational complexity.
• Attention and Categorization can be conceptualized at
dealing solely with complexity reduction. But in all
other areas, complexity also has to be reduced,
particularly Perception, Language, Reasoning,
Problem solving and Decision making.
Methods
• Cognitive Psychology doesn’t bring up many new
issues. There are already plenty of issues resulting
from Philosophy.
• Primarily, Cogpsy brings methods to study and
potentially resolve these issues to the table.
• Classical methods: Accuracy, Reaction times
• Lesioning methods (Ideally: Double dissociation)
• Psychophysiological methods: EEG, MEG
• Neuroimaging methods: PET, fMRI
• Electrophysiological methods: Single-, Multi
Unit Recording
Cognitive functions
1. Perception
2. Attention
3. Imagery
4. Short Term Memory
5. Long Term Memory
6. Categorization
7. Language
8. Reasoning
9. Problem-solving
10. Decision-making
Perception
Perception basics
*Perception is the formation of a mental
representation of the environment.
*This representation is NOT isomorphic, but
subject to many correspondence errors.
*To overcome the inherent ambiguity in sensory
data, the brain makes assumptions about the
world. (Learnt in phylogeny, ontogeny). Like
Gestalt rules.
* These assumptions can be uncovered by
research with visual illusions and brain damage.
Ambiguity
• Many different physical objects (the distal
stimulus) produce the same pattern on the
retina (the proximal stimulus).
• This is an inherent ambiguity in perception.
• The brain needs to overcome it in order to
infer the correct distal stimulus in order to
act properly.
Example: Laws of object
segmentation
• Gestalt laws:
• Law of similarity: Similar things group
• Law of proximity: Things that are near to each other
group
• Law of good continuation: Smooth angles are preferred
• Law of closure: We complete incomplete figures with
illusory contours
• Law of common fate: Things that are moving together
are grouped.
• Law of Pragnanz: Overall law, tying the others together.
Tendency to see the “Good figure”. Harmonious.
Evidence for the Modularity of
perception:
• Certain brain damages can cause visual agnosia,
the failure to interpret visual information, while
still seeing it.
• Certain brain damages disrupt the perception of
motion, but not color and vice versa.
• The face inversion effect (the inability of normal
observers to recognize inverted faces) and
prosopagnosia (the selective inability to
recognize faces after brain damage).
Attention
Attention
• Reduces cognitive complexity
• Only a few things are perceived, the rest not
(example: Change blindness)
• That way, one only acts on a subset of
stimuli in the real world.
• This ensures coordinated and goal-directed
actions. Enhances survival.
Attention theories
• Early selection: Attention is a filter that is
imposed before semantic cognitive
processing takes place (Broadbent)
• Late selection: Attention is a filter that is
imposed after semantic processing, but
before memory (Cocktail party effect,
Treisman)
Controlled versus automatic
• Controlled processing needs attentional
resources. Is done in new tasks, in
complications. No popout.
• Automatic processing is not limited by
attentional resources. Happens involuntarily.
For familiar tasks. Example: Stroop task.
Popout in visual search.
• Transformed by consistent mapping, impaired
by flexible mapping (Shiffrin & Schneider)
Imagery
Imagery
• Top-down activation of memory to form a
mental image. In the absence of a stimulus!
• Down to V1.
• Brain areas that are active in perception are
also active in imagery.
The imagery debates
• Pylyshyn: Mental images are represented in a
propositional format (language-like). They are
arbitrary and hence don’t preserve properties of the
actual percept.
• Kosslyn: Mental images are represented in an analog
format (perception-like). They are not arbitrary and
they do preserve properties of the actual perception.
Example: Mental rotation.
• Kosslyn “won” (not really). The issue was about
format, not content.
Short term memory
Short term memory
Coding, Capacity, Retention duration, etc.
Serial position effects (primacy, recency, use).
Mnemonic strategies: Chunking, rehearsal.
Working memory
Inferference (Proactive, retroactive)
Memory search (serial, exhaustive)
•Interference:
Proactive
1
vs.
2
Retroactive
1
•Working memory = structured STM
Visuospatial
sketchpad
Central
executive
Phonological
loop
2
Long term memory
Concepts to know
•Modal model of memory:
Sensory memory  Short term memory
Storage
Long term memory
Retrieval
Information
Response
•Encoding specificity
-Context effect
-State dependent
learning
-Cues!
Concepts to know
•Explicitness:
Explicit
vs.
Implicit
Bla
•Memory structure
LTM
Knowing how to...
Knowing that...
Declarative
Procedural
Implicit
Episodic
Vivid Recall
Semantic
Knowing
Explicit
Concepts to know
•Sins of memory
7
Long term memory
Coding, Capacity, Retention duration, etc.
Levels of processing theory
Forgetting: Decay, Interference, Overwriting
Encoding specificity: State-dependent learning,
Context effects, spacing, cues, mood dependent
learning.
Autobiographical memory
-Flashbulb memory (Vivid, yet not more accurate)
-Eyewitness testimony (Constructive, Post hoc)
-Repressed memories (Controversial, doubtful)
-Amnesia (Symptoms)
Memory for general knowledge
•Dichotomies:
Implicit vs. Explicit memory
Declarative vs. Procedural memory
Semantic vs. Episodic memory
•Models:
Hierarchical model
ACT model
Network models
Connectionist model
Feature comparison model
Scripts
Schemata
Highly inspired by Computer
Science, Linguistics
Memory (General)
• Memory functions:
• Allows to make inferences (based on few facts)
• Understand new events in terms of old knowledge
• Deliver knowledge when needed
• Andersons ACT model implements functions:
•
•
•
•
•
Working memory, declarative memory, procedural memory
Network model, nodes and relations between them
Spread of activation in network
Production rules: Goals, Conditions, Resulting Actions
Allows to explain goal directed behavior in terms of (working) memory
Categorization
Categorization
• Categories help to reduce cognitive complexity.
• By grouping similar or related things
• They enable higher cognitive functions like
language and reasoning.
• They improve the efficiency of lower functions
like perception and memory.
• The tradeoff is that they might be over inclusive
or biased (Stereotypes)
Major approaches
• Similarity based view
• Knowledge based view
• Be sure to be able to characterize and
contrast both of them
Similarity based views
• Classical view: Feature lists of necessary and
sufficient features. Dichotomous membership,
clear boundaries.
• Prototype view: Based on the typicality effect.
Categories as ideal members with all the
characteristic features. Fuzzy boundaries.
• Exemplar view: Based on actual instances of
objects that are represented. Fuzzy boundaries
Explanation-based view
• Schemata-theory: Concepts as hierarchically organized Schemata.
Explains Prejudice.
• Knowledge based view: Relationship between concepts and
instances is like theory and data. Things have an “essence”, a
nature. There are meaningful relationships between things. These
organize the classification into categories and the boundaries
between them.
• Similarity can’t explain everything. Everything is similar to
everything else in an infinite number of ways. What is relevant?
• Explanation/Knowledge-based view explains why we categorize
both a Great dane and a small dog as dogs, even though they don’t
appear similar. But we know about the nature of dogs. That’s why.
Language
Language
• Extremely rare in organisms. Unique to humans?
• 4 key properties:
• Arbitrarisness (No intrinsic relation between form
and meaning of symbols)
• Productivity (Allowing an infinite number of
expression from a finite set of symbols)
• Duality of patterning (Small numbers of speech
sounds are combined to form large numbers of
meaningful units)
• Discreteness (Elements are perceived categorically,
not continuously)
Structure of Language
• Phonology: Investigation of smallest sound
units in language. Mc-Gurk effect.
• Morphology: Investigation of smallest units of
meaning.
• Semantics: Investigation between words and
their meaning.
• Syntax: Investigation of the relation of the order
of words in a sentence (and it’s meaning).
• Pragmatics: Investigation of the effective use
of language (Language, Interaction between it’s
context and it’s meaning)
Pragmatic rules
Gricean maximes:
1. Quantity: Make your contribution just as
informative as necessary.
2. Quality: Try to be truthful
3. Relation: Try to be relevant
4. Manner: Try to avoid ambiguity, obscurity.
They are supposed to ensure effective communication
Top-down and bottom up
• Bottom-up: Language processing is bottom up, since it is
not influenced by context or higher level information
processing (Autonomy). It is functionally encapsulated
(Modularity).
• Evidence: Brain lesions (language spared, unless one
lesions certain areas that cause Aphasia). Involuntary
language processing (can’t NOT process language, read)
• Top-down: There IS an interaction between top-down and
bottom up processing. It is NOT modular or autonomous.
• Examples: Phoneme restoration effect (people hear things
that are not there, infer from context, cough). Speech
segmentation: Boundaries in spoken speech are NOT in the
physical stimulus, they are imposed top-down.
Reasoning
Types of reasoning
• Deductive reasoning: Propositional
reasoning. From the general law to the
specific instance. Secure, but does not gain
new information that is not already contained
in the propositions.
• Inductive reasoning: From specific
instances to general laws. Never fully secure,
but generates new information, new laws.
Probabilistic.
Hypothesis testing
• Classical: Wason selection task
• Shows that people suffer from confirmation bias:
People try to confirm their rule instead of trying to
falsify it in order to gain insight.
• Shows that people are poor at logical reasoning.
But not inherently. Putting the material into a
semantic context helps.
• Also: Baserate neglect. As a typical problem.
Problem Solving
Problems
• Well defined problems: Clear start state, clear goal
state, clear operations.
• Example: Chess, Tower of Hanoi, Games in general.
• Ill-defined, complex problems: Unclear start state,
unclear goal state, unclear operations.
• Example: Managing a company, waging a war,
leading a good life.
Strategies of problem solving
• Generate and Test: Come up with a lot of solutions and then test
them sequentially until success.
• Means-End Analysis: Comparing the starting state with the goal
state and generating intermediate steps that reduce the discrepancy.
• Working backward: Determine the last step before the goal step
and then generate steps from there on until one reaches the start
position. Useful if there is only possible solution.
• Backtracking: Making assumptions, deliberating as-if and
undoing them if it turns out that it is a dead-end.
• Using analogies: Realize the same underlying structure between a
familiar situation and a problem situation and act as if the familiar
situation would apply. Famous example: Tumor problem, the
gamma knife.
Detriments to problem solving
• Mental set: Being primed to do a task in a certain way, not
in others that might be more efficient. “Tradition”.
Example: Luchins (1942).
• Constrained problem space: Unnecessary constrains
make solutions impossible. Example: 9-dot problem.
• Functional fixedness: Thinking that objects have only
one, well-defined function.
• Inadequate mental representations: Representations
lacking the crucial features that would foster insight.
Example: Mutilated checkerboard.
• Lack of knowledge and expertise: Laypeople categorize
on a shallow level and don’t have memories of when they
solved similar problems before (like experts would).
Fostering Creativity
•
•
•
•
•
•
Practice (“10 year” rule)
Productivity (the more, the merrier)
High IQ
Good mental representations
Willingness to take risks
Personality factors (that increase the
former)
Decision making
Algorithms, Heuristics, Biases
• Algorithms yield optimal solutions (if they exist). They
provide certainty. But are often practically impossible to
apply. Example: Solving chess.
• Heuristics are rules of thumb. They yield a solution that is
good enough. But no certainty that it will work. Quick and
dirty. Example: Doing what everyone else does. They
reduce cognitive complexity
• Biases are cognitive illusions. Like perceptual illusions, it
is hoped that we can understand decision making by
investigating it’s illusions. Like in Perception. Unlike
Heuristics, they generally diminish the quality of the
decision by systematic distortions.
Heuristics and Biases
• Availability heuristic: Things that are overrepresented in the
mental representation are judged to be more likely.
• Representativeness heuristic: Things that are easier to represent
in the mental representation are judged to be more likely.
• Framing bias: Phrasing of a problem matters
• Anchoring bias: Starting point of a problem matters
• Sunk cost bias: Increased probability to continue an undertaking,
once investments have been made into it.
• Illusory correlation bias: Tendency to see structures and relations
where there are none.
• Hindsight bias: Tendency to think that one was right all along.
• Overconfidence bias: Tendency to be overly confident of the
quality of one’s own judgements.
• Confirmation bias: Tendency to seek out information that is
likely to confirm one’s hunches.
Expected Utility vs. Prospect theory
• Expected Utility theory: People calculate probability times
value of all possible outcomes of an option, sum it up and take
the option with the highest value. People are rational decision
makers. Classical Econ.
• Prospect theory: The former obviously only works for very
limited problems. In real life, people use heuristics to solve
problems. They also suffer from biases, particularly the framing
effect. This can be explained by different set-points and different
slopes of the value function for losses and gains. Hence, People
are not rational decision makers. They are irrational or boundedrational at best. (See the Kahneman & Tversky paper for details)
Best of luck!