Cognitive heuristics Flashcards
delving into the AI archives
Social psychology borrowed the heuristic term from AI
Computers, algorithms and the often-futile search for an optimal solution
Settle for approximations
We don’t always try to find perfect solutions
in general
“…the rules people use are fairly rational ones…But the rules are only useful if uncertainty exists or if too much effort is required to arrive at a more complete and accurate judgment. When people rely on heuristics in cases where they could well engage in more accurate types of analysis, and when uncertainty is reduced by the presence of useful data, then heuristics are a source of bias.” (Moskowitz, 2005)
what should we do?
Assess available alternatives for likelihood and worth of outcomes they promised (probability and value)
Calculate utility of each outcome (product of value and probability of each outcome)
Choose the option that maximises utility
In plain English…we make a decision that is most likely to deliver the benefits that we desire
we know this doesn’t always happen
We are not computers!
There may be too much information to sift through rationally
We rarely have the time
We can’t be sure of the outcome (we might still not be happy) – most life decisions do not come with a crystal ball or a guarantee
satisficers v optimisers
Most of the time we are ‘satisficers’ – making adequate inferences and decisions rather than
‘optimisers’ – drawing the best possible inferences and hence reaching the best possible decisions (March & Simon, 1958)
Kahneman and Tversky
Highlighted some of the ways we satisfice, relying on heuristics (using economic theory)
Shone a light on something we all do (examples)
system 1 and 2 thinking
System 1 allows you to: orient to the source of sudden sound; complete the phrase “bread and…”; Answer to 2 +2 (automatic)
System 2 allows you to: brace for the starter gun in a race; look for a woman with white hair; fill out a tax form (require attention and effort)
Not different from the dual process models previously mentioned
To do with resource availability
Good at spotting when system 2 thinking needs to happen
what happens when you pay attention to one thing?
You might not pay attention to something else
Inattentional blindness:
o We can be blind to the obvious
o We can be blind to our own blindness
applying the systems approach to our use of cognitive heuristics
see notes for long example
sample size
Sample size matters, but we often fail to take account of it
Statistics produce many observations that seem to beg for causal explanations but are simply due to chance
System 1 is the mode of thinking that leaps on causal connections
It runs ahead of the facts and jumps to (often wrong) conclusions
what other heuristics highlight system 1 thinking?
Representativeness
Availability
Adjustment and anchoring
representativeness example
see notes
What happened to taking the base rate information into account?
What about the fact that you were told the description was based on results of ‘uncertain validity’?
Would remembering to bear these things in mind cause you to revise your nerdiness characterization of Tom?
representativeness heuristic
a mental shortcut whereby instances are assigned to categories on the basis of how similar they are to the category in general
representativeness empirical illustration
Tversky and Kahneman (1974)
Estimate probability a man was either engineer or lawyer
Man sampled at random from group of engineers and lawyers
Some told: 70% engineers, 30% lawyers
Others: 30% engineers, 70% lawyers
Some got (useless) personality profile
Some didn’t
Tversky and Kahneman (1974) results
When people didn’t get a personality profile…estimate reflected base rates (if told 70% lawyers, estimated 70% chance he was)
When they did get profile (even though it was rubbish!), they became less rationale and ignored base rate information and essentially guessed (50/50 judgments)
why do we do it? - representativeness
Sometimes…there is some truth to stereotypes! But they can mislead us…does she have a college degree?
Base rate information
Predictive value – how credible is the source?
Sample size again – small samples less reliable
can we overcome the tendency to overlook base rate information?
It’s actually not that easy, but it can be done…
Instructing people to “Think like a statistician” enhanced use of base-rate information, but…” Think like a clinician” had the opposite effect! (Schwarz et al., 1991)
- Bias in way doctors make judgements
o Non-conscious stereotyping – racial and ethnic disparities in treatments
Doing task whilst frowning versus puffing your cheeks out – what happens?
o Difference in reliance on base rates
o Frowning makes us less confidence – uncertainty in the situation – reduces reliance on intuition
o Become aware of being uncertain
o Puffing cheeks is more neutral
Frowning increases vigilance; enhanced activation of System 2 – increased use of base rate info (Oppenheimer et al., 2007)
availability
How many famous celebrities have succumbed to plastic surgery in the past few years?
Would you be systematic? (conservative figure)
Or would you instead bring to mind (probably with very little trouble and a quick search on Google) the instances of celebrities who are well-known to have gone under the knife? (loads!)
If the latter, you’ve succumbed to the availability heuristic…
availability heuristic
a cognitive shortcut that allows us to draw upon information about how quickly information comes to mind about a particular event to deduce the frequency or likelihood of that event
As T & K say: “That associative bonds are strengthened by repetition is perhaps the oldest law of memory known to man. The availability heuristic exploits the inverse form of this law, that is, it uses strength of association as a basis for the judgment of frequency.”
o Quicker comes to mind = more available
It’s not always that simple…
biased estimates occur for 2 reasons
- It’s not only about frequency, what about familiarity and salience?
- Our own experiences with things happening frequently may be idiosyncratic
availability heuristic empirical illustration
Tversky and Kahneman (1973)
Memorise list of names – famous men and women
For some participants, men were more famous than women, for others, vice versa
Some asked to judge how many men and women in each list (numbers were equal)
o Only differed in famousness
Others asked to recall names
Tversky and Kahneman (1973) results
Gender with more famous names judged to be more frequent and participants recalled more (50% more) names of that gender
Fame made names salient => made them easier to recall => overestimation of frequency of the group
so why do we do it? - availability
We assume that volume of exemplars (content) correlates with its ease of retrieval (Schwarz et al., 1991; ELE).
Or… we simply feel that if information comes to mind easily (or not) then we it must say something about how frequent it actually is.
Schwarz et al. (1991) – pps recall six or twelve assertive behaviours, judge their own assertiveness.
Who do you think rated themselves as more assertive?
Those recalling only six assertive examples rated themselves as more assertive than those recalling twelve examples
Also works with unassertiveness
So, what is going on?
The feeling of difficulty/ease of retrieval (System 1) can matter as much as absolute numbers (content) (System 2).
other findings from Tversky and Kahneman (1973)
People believe that they use their bicycles less often after recalling many rather than fewer instances of doing so
People are less confident in a choice when asked to produce more arguments to support it
People are less impressed by a car after listing many of its advantages
anchoring and adjustment
When making judgments under uncertainty, you can reduce ambiguity by starting with an anchor
My conundrum: How many handouts should I print out for my next tutorial?
Last time 75% attendance
o Correct with the use of other information
But this next tutorial might contain exam tips…
Maybe more like 100% this time?
And we do this with people too… (correspondence bias is one example)
wheel of fortune
Participants stood in front of a Wheel of Fortune marked from 1-100, asked to spin the wheel
Wheel rigged to stop at either 10 or 65
Asked “What is your best guess of the percentage of African Nations in the UN?”
Did the arbitrary WOF number affect their guess?
Mean estimates of those who saw WOF stop at 10 = 25%
Mean estimates of those who saw WOF stop at 65 = 45%
Anchoring heuristic is one of the most robust and reliable results in experimental psychology
anchoring and adjustment heuristic
anchoring/adjustment heuristic a cognitive heuristic that makes us place weight upon initial standards/schemas (anchors) and as a result means we may not always adjust sufficiently far from these anchors to provide accurate judgments
o Initial anchor can constrain us
anchoring and adjustment empirical illustration
Englich et al. (2006)
Participating legal experts shown realistic case materials involving an alleged case of rape and asked to provide their sentencing decision.
Participants received one of three anchors: irrelevant source (a journalist); in a second condition they were told the anchor had been randomly chosen; and in a third condition, participants themselves randomly decided upon the anchor by the toss of dice.
In each case, the anchor was either high (three years) or low (one year).
In all three conditions, the anchor constrained sentencing decisions.
Stopped them being systematic system 2 thinkers
More inclined because they’re experts – they use shortcuts lots of the time
why do we do it? - anchoring and adjustment
Tversky and Kahneman did not agree on this!
Tversky: Traditional view…deliberate, if insufficient attempt to adjust from an irrelevant value that provides an anchor (Tversky & Kahneman, 1974): System 2
Kahneman: Anchoring that occurs by priming: System 1
Both are probably right, it sorts of depends…
adjustment is effortful - system 2
Remember the correspondence bias?
Same goes here – under cognitive load people adjust less (insufficient adjustment is a failure of a weak or insufficient System 2; Epley & Gliovich, 2006)
anchoring as a priming effect - system 1
Mussweiler & Strack (2000) – anchor primes associated concepts in memory
“Is the annual mean temperature in Germany higher or lower than 20 degrees C (or 5 degrees C?)”
Those shown 20C found it easier to recognize words like sun and beach
Those shown 5C found it easier to recognise words like frost and ski
