Heuristics Flashcards
What is probabilistic reward learning?
Consider you’re playing a game where you are repeatedly given the choice between a blue and yellow option, on each trial
Choosing the blue option will, with a probability of 0.7 on each trial, lead to an incremental small reward
Choosing the yellow option will, with a probability of 0.3 on each trial, lead to an incremental small reward
What did Tversky and Edwards (1966) find about probability matching?
found that judges probability matched when they were asked to predict which of two lights was going to turn on next.
Interestingly, the results do not depend that much (after some time) on whether people are being made aware of these probabilities or have to learn them
Tversky and Edwards found that participants predominantly probability matched those frequencies
What was the optimal probability matching stragegy?
The optimal strategy would be consistently choosing the left light, which would lead to a success rate of 70%.
Did participants ever learn the optima strategy in the matching task?
Even after long exposure to the task, most participants do not learn the optimal strategy
The remarkable thing about this is that the asymptotic behavior of the individual, even after an indefinitely large amount of learning, is not the optimal behavior… We have here an experimental situation which is essentially of an economic nature in the sense of seeking to achieve a maximum expected reward, and yet the individual does not in fact, at any point, even in a limit, reach the optimal behavior.
K. J. Arrow (Econometrica, 1958, p. 14)
Which sequence is random?
BABABABABABABAA or ABBAAABAABBBBAA
How do humans interpret randomness?
If Ss are asked to write down a random sequence of numbers (or letters, or coin tosses) they tend to try and make the sequence look random at every point. Kahneman & Tversky (1972) called this local representativeness.
People exclude long runs e.g. 12133333312
People try to make each number more equifrequent than would be expected by chance.
What is the gamblers fallacy?
If each “coin toss” / event is independent from the previous ones, there is absolutely no predictability or enhanced likelihood of next coin flip being tail even if there has been a series of 1 million heads before (!!!!)
When is the gamblers fallacy different?
This is of course different if, as in (most) lotteries, balls are drawn from an urn with limited balls.
If a ‘6’ has been drawn without replacement and there is only one ‘6’, obviously there cannot be another ‘6’….
What is the definition of local representitiveness?
the belief that a series of independent trials with the same outcome will be followed by an opposite outcome sooner than expected by chance.
How did Gillovich, Vallone, & Tversky (1985) research randomness or the “hot hand”?
examined people perceptions of the “hot hand” (or lucky streaks) in basketball (the same thing applies to any game including poker).
They reported statistical analyses of lucky streaks for specific basketball players and reported that these were simply misperceptions. In truth successful shots during lucky streaks were no more likely than that players overall probability of a lucky streak…lucky streaks are an illusion.
Can humans distinguish random coincidences from systemic patterns?
not really, The human brain searches for patterns in everything (‘attribution’) and deserves explanations for the phenomena we encounter - often requiring a scapegoat
(think of conspiracy theories – “Covid is just a hoax”)
In particular, humans massively mis-interpret short sequences (local representativeness)
What does Kaheneman say about “steve, the shy and withdrawn man” is he more likely to be a farmer or a librarian?
Kahneman argues that Steve is more likely to be a farmer, as there are many more farmers than there are librarians (in the US) – a fact that people tend to forget or ignore when they have to make their judgement
What is a problem involving base rate ngelect?
Consider the following problem from Tversky & Kahneman (1982) :
A cab was involved in a hit and run accident at night. Two cab companies, the Green and the Orange, operate in the city. You are given the following data :
85% of the cabs in the city are green and 15% are orange.
A witness identified the cab as orange. The court tested the reliability of the witness under the same circumstances that existed on the night of the accident and concluded that the witness correctly identified each one of the two colours 80% of the time and failed 20% of the time.
What is the probability that the cab involved in the accident was orange rather than green?
What is the medical diagnosis problem?
Casscells et al (1978) asked medical students the following question:
If a test is to detect a disease whose prevalence is 1/1000 has a false positive rate of 5%, what is the chance that a person found to have a positive result actually has the disease? Assuming that you know nothing about the person’s symptoms or signs: __%
18% responded 2%, i.e. the correct Bayesian inference.
45% responded 95%, i.e. the response that ignores the base rate.
Thus even medical students ignore base rates for diagnosis problems. This is normally attributed to the representativeness heuristic.
How did Cosmides and Tooby present the medical diagnosis problem?
Cosmides & Tooby (1996) presented the problem in both probability and frequency formats:
1 out of every 1000 Americans has disease X. A test has been developed to detect when a person has disease X. Every time the test is given to a person who has the disease, the test comes out positive (i.e., the “true positive” rate is 100%). But sometimes the test also comes out positive when it is given to a person who is completely healthy. Specifically, out of every 1000 people who are perfectly healthy, 50 of them test positive for the disease (i.e., the “false positive” rate is 5%). Imagine that we have assembled a random sample of 1000 Americans. They were selected by a lottery. Those who conducted the lottery had no information about the health status of any of these people.
Given the information above: on average, how many people who test positive for the disease will actually have the disease? 1 out of 50 = 2 out of 100 or 2%
