Ch. 2 Flashcards
Social Judgement Theory
What does it do?
How is it derived?
What does it not consider?
What types of decisions is it used for?
How accurate it is?
- Tries to ID the factors that people use when making a particular type of judgment and how much weight is attached to these
- Purely statistical approach (ideally linear regression), doesn’t describe cognitive processes by which ppl reach a judgment
- applies to judgments, particularly in professional settings that are made repeatedly……
- Models can be used to predict future cases………
- outperform human judges
What are actuarial models?
linear equations for specifying how the relevant cues should be combined/weighted to make a judgment
What info do we need to create linear models through linear regression?
a. outcomes are known and b. info about cues used is available
linear regression can be used to derive equations to determine how much impact each cue had
linear model equation
Ye=the outcome, criterion in the environment about which a judgment must be made…. X = the cues that are used to predict… b = weight of the factor with regards to outcome… e=error…..
predictive model
everything except e, can be used to predict future judgments (no known outcome)….
Goldberg rule
what’s it for?
what factors does it involve?
how accurate it is?
- linear model for distinguishing between psychosis and neurosis……
- involves adding scores from three scales on MMPI (Minnesota Multiphasic Personality Inventory) and then subtracting scores from two other scales….
- If less than 45 then diagnose neurosis…..
- outperforms human judges – in 861 cases, 70% predicted right vs. humans predicted 50-67% right
Bootstrap model
When is it used?
how accurate is it?
predictive model (Ys) minus the error term…
used when info about true state of affairs unknown…
can be used to implement a judge’s policy in a consistent manner…
improper linear model because it doesn’t optimally weigh the cues, but still more accurate than the judge that it was derived from……
Ecological validity
the predictive value of a cue
Lens model
derived from Brunswik’s work…. Assumes that we don’t perceive the world directly but rather through a ‘lens’ consisting of various items of information (cues)…. Research done with representative design
Representative design
research methodology in which we randomly sample stimuli from the environment or create a stimuli in which environmental properties are preserved….
Brunswik used it for lens model research…..
thought it was more generalizable (e.g. externally valid) than standard experiments
Clinical prediction
What is it?
How accurate is it/under what conditions?
unaided human judgment…
better than actuarial in a changing environment……
but in stable environment even access to additional info/interviews doesn’t help
Actuarial models
What does it assume?
What role do people play in creating actuarial models?
people are less accurate than actuarial models…..
yet people are necessary to choose the variables used in the model and to code them to have clear directional relationships…..
assume a stable environment
Why are humans worse?
Name some of the reasons (9).
People are…. 1. inconsistent
- rely on irrelevant info
- ID too many cases as exceptions/special cases
- wrong weight of cues with criteria
- ppl can have skewed samples
- fundamental attribution error
- delayed/inaccurate feedback
- influenced by recent experiences
- fatigue, boredom
fundamental attribution error
underestimating the influence of situational factors in other people’s behaviour
improper linear models
Why are they improper?
Name some examples
do not optimally weigh the cues (e.g. judges)…
bootstrap models, random linear models (each cue weighted randomly and differently), equal weighting models (each cue weighed approx.. the same)
hungry judges (danzinger et al)
(danzinger, levav, avnaim-pesso 2011) – looked at judicial parole decisions in Israel….
Hearings are pseudo-random (depend on when lawyer shows up)…
judges take mid-morning meal/lunch brakes whenever they like and eat at the bench between cases…..
supposed to be consistent
favorable decisions peaked right after food
Why lack of impact of actuarial models in real world?
superiority of these is very robust finding… lack of impact probs because experts feel their expertise being questioned, ppl have selective memory, belief that every case is unique
Probabilistic mental models theory
What does it theorize about decision making?
What does it assume?
Gigerenzer & Goldstein……
when person is unable to distinguish between binary alternatives they initiate a sequential search for a cue that will provide a reason for choosing one alternative over another………
assumes judgment based on a single cue (vs. multiple ones in SJT/regression models)
One-reason decision making
PMM theory…… limited capacity processing of info
Fast and frugal
PMM theory…… less time and less info used to make decision
Heuristics
shortcuts used to make fast/frugal judgments
recognition heuristic
choose what you’ve hear of before
take the best heuristic
people take the most valid cue, then the next valid, etc. til find a discriminating one
take the last heuristic
select the cue that worked last time this judgment was made
minimalist heuristic
select a cue at random
accuracy of fast/frugal statistics
G/G simulations of these - found TTB = regression in accuracy (65.8% and 65.7%) using only 3 cues vs 10 cues, e.g. more efficient…
take the last used 2.6 cues and got 64.5% accuracy…
minimalist used 2.8 cues and got 64.7% accuracy
Empirical tests of recognition
G/G – German cities… 90% of population judgments were made in accordance with recognition…
criticisms focus on methodology, e.g. not controlling for prior knowledge, soccer cue insignificant to americans………….
Richter/Spath, Newell/Fernandez found that participants used knowledge together with recognition, and that increasing the validity of other environmental cues led to less reliance on recognition…..
appears that people use more than one reason
Empirical tests of take the best
ppl may not be learning cue validities well…
Strategy selection
What role does environment feedback play?
Risekamp/Otto show that environmental feedback affects learning of strategies….
Non-compensatory environment – sometimes one strategy was best, sometimes another…..
compensatory environment – one strategy was always best…..
across all studies they found that the Weight additive (WADD) strategy was preferred at start but then shift to the most profitable strategy right after
Non-compensatory environment
sometimes one DM strategy was best, sometimes another
compensatory environment
one DM strategy was always most appropriate
Broder and intelligence/DM
Did intelligent people adopt more complex decision making strategies?
What did this study show about frugal decision making?
the more intelligent people did not adopt more complex trategies, rather they used TTB in non-compensatory environment……
shows that a deliberate decision is made to ignore info and behave in frugal fashion
Select the successful heuristic
What determines the usefullness of a cue?
overall usefulness of a cue depends on its discrimination rate (whether it leads to a discrimination btwn choices) and its validity…..
however, _he (?) _found that success appeared overall to be a greater determinant of search than either validity or discrimination rate
Heuristics/Probabilistic Mental Models key point?
people use a single cue in order to make thier judgment
Social Judgment Theory key point?
actuarial models outperform clinical models, humans
