Course 4: Heuristics and cognitive biases
Econs (= homo economicus)
(Rational) preferences, unbiased beliefs and expectations
Make optimal choices bases on these beliefs and preferences
Infinite cognitive abilities and willpower
Primary motivation is self- interest, limited altruism
Heuristics
= mental shortcuts or rules of thumb that people use to make decisions quickly
and efficiently, especially under uncertainty or limited information. Simplify
complex problems; good decisions w/out much time or effort
Cognitive biases
=systematic patterns of deviation from rational judgment that result from the use
of heuristics, as well as from emotional, motivational or social -influences.
Heuristics
1) Representativeness
= The degree to which an event is similar to the population
Our brain relies on similarities w/ previous experience when evaluating a
situation or probabilities
Evaluating according to their important, representative characteristics
2) Availability
=People estimate the likelihood of an event based on the ease with which they
can recall instances of that event
Things that come rapidly to mind
Overweight very recent, highly visible or vivid events
3) Anchoring and adjustment
= An initial value pulls judgements; subsequent adjustment is insufficient
Cognitive economy (mental laziness)
➔ Starting from a number you already have saves effort
Coherence seeking
➔ Compromise between anchor and model feels reasonable
Social stickiness
➔ Public commitments create reputational anchors
➔ Link to disposition effect: Price becomes anchor + loss aversion
1
, Biases in probability judgement
1) Law of small numbers (= sample size neglect)
=People expect small numbers to mimic the characteristics of the whole
population (or long-run frequencies). We over-interpret short streaks or a
few observations as if they were representative.
The fallacy: Samples represent the population, even small samples must look
balanced.
Mistaking noice for signal
Expecting randomness to self-correct immediately
Underestimating how random sequences accur
Related patterns
The Gambler’s fallacy — After a streak (e.g., HHHH), people expect
reversal (T “is due”), as if outcomes were negatively autocorrelated in the
short run.
➔ US refugee judges 3.3% more likely to reject a case if they approved the
previous one
➔ Loan applications: approval (per loan officer) 9% less likely after previous
application approved.
The Hot Hand fallacy — Short success streaks are read as persistent
skill, so continuation is overestimated.
Neglect of regression to the mean — Extreme outcomes (in small samples)
are expected to persist rather than revert toward average. Role of skill vs. luck
2) Partition dependence
= Judged probabilities depend on how the outcome space is partitioned or
described. A variant of the framing effect
Related patterns
Subadditivity: The sum of parts exceeds the judged probability of the
whole.
➔ which basketball teams would win the NBA? Estimated P probabilities for 8
remaining teams 218%.
Equipartition Heuristic — When unsure, people roughly split probability
across listed categories.
Design sensitivity: People’s choice depend on menu design. Change the
design; change the answer.
3) Conservatism
= insufficient belief revision, people move their posterior too slowly relative to
bayes when presented with new information
Cognitive friction; bayes is hard
Caution; prefer not to move fast
2
Econs (= homo economicus)
(Rational) preferences, unbiased beliefs and expectations
Make optimal choices bases on these beliefs and preferences
Infinite cognitive abilities and willpower
Primary motivation is self- interest, limited altruism
Heuristics
= mental shortcuts or rules of thumb that people use to make decisions quickly
and efficiently, especially under uncertainty or limited information. Simplify
complex problems; good decisions w/out much time or effort
Cognitive biases
=systematic patterns of deviation from rational judgment that result from the use
of heuristics, as well as from emotional, motivational or social -influences.
Heuristics
1) Representativeness
= The degree to which an event is similar to the population
Our brain relies on similarities w/ previous experience when evaluating a
situation or probabilities
Evaluating according to their important, representative characteristics
2) Availability
=People estimate the likelihood of an event based on the ease with which they
can recall instances of that event
Things that come rapidly to mind
Overweight very recent, highly visible or vivid events
3) Anchoring and adjustment
= An initial value pulls judgements; subsequent adjustment is insufficient
Cognitive economy (mental laziness)
➔ Starting from a number you already have saves effort
Coherence seeking
➔ Compromise between anchor and model feels reasonable
Social stickiness
➔ Public commitments create reputational anchors
➔ Link to disposition effect: Price becomes anchor + loss aversion
1
, Biases in probability judgement
1) Law of small numbers (= sample size neglect)
=People expect small numbers to mimic the characteristics of the whole
population (or long-run frequencies). We over-interpret short streaks or a
few observations as if they were representative.
The fallacy: Samples represent the population, even small samples must look
balanced.
Mistaking noice for signal
Expecting randomness to self-correct immediately
Underestimating how random sequences accur
Related patterns
The Gambler’s fallacy — After a streak (e.g., HHHH), people expect
reversal (T “is due”), as if outcomes were negatively autocorrelated in the
short run.
➔ US refugee judges 3.3% more likely to reject a case if they approved the
previous one
➔ Loan applications: approval (per loan officer) 9% less likely after previous
application approved.
The Hot Hand fallacy — Short success streaks are read as persistent
skill, so continuation is overestimated.
Neglect of regression to the mean — Extreme outcomes (in small samples)
are expected to persist rather than revert toward average. Role of skill vs. luck
2) Partition dependence
= Judged probabilities depend on how the outcome space is partitioned or
described. A variant of the framing effect
Related patterns
Subadditivity: The sum of parts exceeds the judged probability of the
whole.
➔ which basketball teams would win the NBA? Estimated P probabilities for 8
remaining teams 218%.
Equipartition Heuristic — When unsure, people roughly split probability
across listed categories.
Design sensitivity: People’s choice depend on menu design. Change the
design; change the answer.
3) Conservatism
= insufficient belief revision, people move their posterior too slowly relative to
bayes when presented with new information
Cognitive friction; bayes is hard
Caution; prefer not to move fast
2
