Thinking Flashcards
THINKING
INDICTIVE REASONING
- predicting future via past data/making judgements
- stat generalisations/probabilistic judgements/predictions
- hypothesis testing/rule induction
DEDUCTIVE REASONING
- solving logical/mathematical problems w/right answers
- based on grounds/given facts/generating valid conclusions/evaluating ones validity
PROBLEM SOLVING
- how to get from A-B
- numerous solutions/varying constraint degrees
JUDGEMENT/DECISION MAKING
- choosing among options
CREATIVE THINKING
- daydreaming/imagining
THINKING RESEARCH
- focused primarily where there’s:
- right answer/way of evaluating answer rationality/efficiency of getting to answer
- asks HOW people think (processes/representations)
- human imperfection emphasis (why irrationality/efficiency limits relative to ideal)
- practical focus motivation:
- practical false medical/legal/military importance
- improvements via training/IT support
- attempts changing behaviour
GENERAL DUAL-PROCESS THEORY: SYSTEM 1
- intuitive/automatic/unconscious/quick & dirty/approximate BUT domain-specific
- procedures/schemas/heuristics are:
- adaptive/effective when applied in appropriate domain (otherwise error)
- only approximate w/some built in biases
GENERAL DUAL-PROCESS THEORY: SYSTEM 2
- slow/sequential/effortful BUT…
- logical/rational/conscious reasoning system
- allocates attention to demanding effortful mental activities
- constrained via limited WM capacity/basic cognitive machinery limits
SYSTEMS 1 + 2 COMBINED
- effortful S2 = depleted; self-control/cognitive effort = mental work
- if cognitively busy, S1 = ^ beh influence/^ temptation following (ie. selfish choices/superficial social judgements)
INDUCTIVE REASONING
- illustrate basic cognitive machinery properties
- limit optimal reasoning strategies
- introduce biases:
- difficulty attending relevant info w/salient/irrelevant info available
- limited WM capacity
- LTM retrieval properties
- shifting mental set/perspective difficulties
DEDUCTIVE REASONING
- logical reasoning w/quantifiers (some/not/all) = reasoning via imagining concrete examples (mental models) rather than abstract/general logical operations
- illustrated WM limit impacts
- Wason’s 4 card deductive reasoning test w/if-then propositions = explains performance effects on problem content/characteristic error nature in domain-specific heuristics
PROBABILITY/FREQUENCY JUDGEMENTS
- some frequency facts = told to us/searchable (ie. lifetime morbid schizophrenia risk = 0.7%)
- BUT many based on experience (ie. will it rain today?)
MEMORY AVAILABILITY
TVERSKY & KAHNEMAN (1973)
- availability heuristic = judge as ^ probable/frequent events/objects ^ readily available in memory/environment
- works as generally easier to retrieve event/object memory that’s frequent
- retrievability also affected via: recency/salience/current case similarity
- over-estimate event probability where examples = easily retrievable AKA…
- availability bias = recent/personally salient/presently similar examples
AVAILABILITY BIAS EXAMPLES
- Cape Cod “Jaws” screening = drop in California coast swimming
- seeing accident/police = drivers slow (for a while)
SLOVIC et al (1980) - people overestimate dying of rare BUT dramatic/reported causes (ie. floods/tornadoes/measles)
- underestimate dying of common causes (ie. strokes/cancers/diabetes)
- aka. fear of British kids being run over > ran over
REPRESENTATIVENESS BIAS/BASE RATE NEGLECT
- when evaluating particular cases:
- tend to ignore important info source aka. base knowledge rate = particular event classes overall frequencies (ie. is someone = X features, we think they have X’s standard properties)
- possible best guess basis for category member w/o other info BUT biases prototypical property attribution even w/other info
REPRESENTATIVENESS BIAS EXAMPLE
TVERSKY & KAHNEMAN (1973)
- 100 descriptions (70 = lawyers; 30 = engineers; ie. kids/hobbies/age/motivations); pp asked engineer prob
- > 90% went w/older/conservatism/maths hobbies etc.
REPRESENTATIVENESS BIAS X SEQUENTIAL EXENTS
- head prob x11 in row = same as any as coin = no memory; expecting non-representativeness redressing = gambler’s fallacy
- difficulty ignoring sequence representativeness/unusualness or focusing on known individual event probs
FUNCTIONAL FIXEDNESS
DUNCKER et al (1945)
- classic Gestalt psychologists exps
- pps asked to support lighted candle on vertical wooden wall w/nails/matches etc
- < successful than pps w/same prob BUT pins tipped out of box
CONSERVATISM X CONFIRMATION BIAS IN INDUCTIVE REASONING
- IRL/research we want rules/principles describing experienced instances/testing hypothesised rules VS further observations
WASON (1960) - 2-4-6 seq generated rule; pp tries to guess via new 3n sequences w/positive/negative feedback (ie. fits rule/doesn’t then declares rule hypothesis)
- tend to offer over-specific hypotheses (ie. n+2)/conservatively reluctant to abandon them/seek confirmatory VS disconfirmatory evidence
- scientists equally prone
PROBLEM SOLVING
- studies = situations w/start/goal states; fast goal achievement via available operators subject to certain constraints (ie. missionaries/cannibals/Hanoi’s tower)
LUCHIN’S WATER JUG - start w/full 8 pint jug/empty 5 pint jug/empty 3 pint jug
- end w/4 pints of water in largest jug
THE “PROBLEM SPACE”
NEWELL & SIMON (1972)
- soluble problem = at least 1 path through state space between start/goal
- problem solver must search w/o knowing advanced optimal path/traversable intermediate states for operators that:
- move them through intermediate states towards goal
- avoid dead-end backing up/going in circles
- minimise path
WM CAPACITY/HEURISTICS X PROBLEM SOLVING
- huge workplace + time = exhaustively enumerate all possible legal moves/pick shortest path (ie. Big Blue playing chess) BUT illegitimate WM capacity SO…
- recognise familiar patterns/retrieve previous effective moves via LTM (ie. Kasparov’s chess)
- hunt through initial/goal states in small steps via heuristics (ie. mean-end analysis/don’t repeat moves)
MEAN-END ANALYSIS - pick general means of goal reaching; unavailable = create sub-goal of achieving availability means till 1 generates satisfiable sub-goal via available operator
- requires WM goal stack maintenance
DESIGN LIMITS INTRINSIC TO COGNITIVE MACHINERY
- design limits (in cog caps (memory retrieval properties/limited WM/relevant info attendance difficulty/cognitive set shift difficulty/sequential reasoning effortfulness difficulty) =
- heuristic reliance (approx rules of thumb)
- intrinsic biases when heuristics applied
MENTAL MODELS X SYLLOGISTIC REASONING
JOHNSON-LAIRD et al
- we DON’T reason w/formal logic mental version
- given premises we imagine +1 possible concrete worlds where premise = true (mental models)
- then generate conclusion/determine if offered conclusion = valid via mental model examination
- errors arise via: failure of generating all possible premise mental models/WM capacity lack for multiple model maintenance
- only 1st model construction matches conclusion = think inference valid BUT untrue
- 2nd model also describes affairs consistent w/premises where conclusion = false
ILLUSORY CORRELATION
IMAGINE…
- doc investigates disease presenting certain symptoms BUT present w/other diseases
- particular S in 80%; concludes S = good D predictor
- WRONG! must compare D frequency w/S VS w/o S
- 40/50 w/o S = D = confirmation?
- WRONG! must consider all cases w/S+D VS w/o (aka S+D/S-D/D-S/neither)
- 4/5 patients develop D regardless of S or w/o
- wrong conclusion = availability bias; patients w/o S/D = < striking than those who do
ABSTRACT VS CONCRETE REASONING
- concrete scenarios imagined assisting reasoning/mental models BUT…
- ability limited via WM representational capacity
- may fail all possible scenario consideration
- more available instances = easier to access
TROUBLE W/CONDITIONAL PROPOSITIONS
WASON’S 4-CARD PROBLEM
- all cards w/letter on 1 side; number on other
- RULE = if card had vowel, then odd number on other side; which cards flipped to check if true?
- pps choose A (as should) BUT 1 VS 2 too
- A+2 correct why? logical = if P then Q, only combo inconsistent w/rule = P+ NOT!Q so those (A+2) cards; BUT this says nothing about “if Q…” so no point checking 1
ARE PEOPLE JUST ILLOGICAL?
JOHNSON-LAIRD (1972)
- NO; changing problem content/context w/o changing formal structure = dramatic performance improvements
CHENG & HOLYOAK (1985)
- formally identical prob w/form w/transit/entering on 1 side then disease list on other; check rule observance: “if the form has “entering” on 1 side, then the other includes cholera amongst diseases”
- half pps given rationale (mere transit passengers don’t need cholera inoculation; visitors do); other told to check for tropical diseases
- w/o = poor 60% perf; w/ = well 90%
- SO if given/familiar w/social rules/permissions rule context, performance = good
DOMANI-SPECIFIC “IF-THEN” DEONTIC REASONING
- questions why performance ^ w/concrete contexts
CHENG & HOLYOAK - successful conditions engage familiar permission schema for social rules (what ought to happen aka. IF YOU WANT P, THEN YOU MUST Q)
- deontic “if-then” = same truth conditions as logical
DOMAIN-SPECIFIC “IF-THEN” CAUSAL REASONING
OAKSFORD & CHATER (1994)
- why characteristic error made (ie. Q alternative in abstract problem versions)?
- pps choices = rational under causal/probabilistic “if-then”; NOT same truth conditions as logical IT (ie. clouds cause rain = IF ran, THEN clouds)
- if proposition interpreted as causal/correlational claim = clouds imply rain prob
- SO IF clouds, reasonable to check for rain to collect info about relationship strength
SUMMARY: WHY IS THINKING ERROR-PRONE?
- “design limits” in cognitive caps (memory retrieval properties/limited WM/difficulty attending relevant info/difficulty shifting cognitive sets) =
- heuristic reliance (approx rules of thumb)
- intrinsic biases applying heuristics
- habit of reasoning w/concrete mental models coupled w/failure/inability to generate/represent all possible mental models
- “capture” of reasoning via relatively automatic domain-specific heuristics =
- adaptive in right contexts BUT…
- inappropriate for cases at hand
