Causal Base Rates
A base rate can influence judgments more when it suggests a causal story. In Ajzen’s exam experiment, participants used class pass rates more strongly than statistically equivalent rates created by sampling students who had passed or failed.
THE IDEA
Causal Base Rates
Amos Tversky and Daniel Kahneman constructed variants of the cab problem, but the powerful notion of causal base rates was borrowed from psychologist Icek Ajzen. In his experiment, Ajzen showed participants brief vignettes describing Yale students who had taken an exam and asked them to judge the probability that each student had passed. The manipulation of causal base rates was straightforward: one group was told that the students were drawn from a class in which 75% passed, another group from a class where only 25% passed. This is powerful because the base rate of passing suggests the inference that a test with only 25% passing must have been brutally difficult—a causal factor determining every student’s outcome. As expected, participants were highly sensitive to causal base rates: every student was judged more likely to pass in the high-success condition than in the high-failure condition.
Ajzen also used an ingenious method to suggest a noncausal base rate. He told subjects that the students were drawn from a sample constructed by selecting students who had passed or failed. For the high-failure group, the information read: "The investigator was mainly interested in the causes of failure and constructed a sample in which 75% had failed the examination." This base rate describes how the sample was assembled rather than a cause of exam performance. It still matters when judging a student drawn from that sample. The explicitly stated base rates had some effects but much less impact than the statistically equivalent causal base rates. System 1 can handle stories with causally linked elements but is weak in statistical reasoning. For a Bayesian thinker, the versions are equivalent.
When to use it
When evaluating performance, distinguish a team-wide success rate that suggests shared conditions from a rate produced by selecting particular employees.
When interpreting a medical sample, ask how patients were selected and which population the reported proportions describe.
When judging student performance, consider possible explanations for school pass rates without assuming that teaching quality is the only cause.
When comparing investment records, distinguish the performance of a defined market from a sample assembled after seeing which funds succeeded.
FROM THEORY TO EVERYDAY LIFE
In practice
Exam Success Prediction
When told that a student is from a class where 75% passed, people predict the student likely passed; when told it's from a class where only 25% passed, they predict the student likely failed, because they infer the test difficulty (a causal factor).
Noncausal Base Rate Neglect
When a sample is deliberately constructed to include 75% failures, that proportion gives information about a randomly selected member. In the reported experiment, participants gave such sampling information less weight than a class failure rate that suggested a difficult exam.
A MOMENT TO REFLECT
Put it to work.
Think of a real-world decision where you relied on a base rate. Was it causal or purely statistical? How did that affect your judgment?
When evaluating someone's performance, do you unconsciously adjust for situational factors (like test difficulty) that are causal base rates?
Can you identify a scenario where a statistical base rate was ignored because it lacked causal relevance, leading to a biased conclusion?
Often confused with
References
- Causal versus diagnostic contingencies: On the role of base rates in the prediction of behavior — Icek Ajzen (1977)
CONTINUE THE CONVERSATION
Source texts
- Thinking, Fast and Slow (Daniel Kahneman)