Reading focus
Syllabus: How Do We Judge and Decide?
Track what each participant could have learned from their sample. Compare choices relative to the objective distribution and relative to the information actually encountered.
Why you're reading it. Show why an account built around stated probabilities may need modification when probabilities must be learned.
A question to consider. Would changing sample size while holding the final observed frequency constant help distinguish missing information from the way experienced outcomes are weighted?
Why this reading
Receiving a description of possible outcomes differs from learning about them through a limited sample. This experiment makes that difference visible while holding the underlying options fixed. It creates a useful challenge for any theory that treats known probabilities and experienced frequencies as interchangeable inputs.
Before you read
A payoff distribution specifies possible outcomes and their probabilities. Sampling reveals only some outcomes from that distribution. A rare event may never appear in a small sample. Underweighting relative to the true distribution differs from assigning too little weight to an event that was actually observed. Free sampling here precedes a consequential choice.
Read for
- Compare the information available in the description and experience groups. Distinguish equal underlying options from equal knowledge about them.
- Trace sample size, encountered outcomes, and recent observations separately. Identify which explanations the design can distinguish and which are inferred from associations.
- Compare the observed choices with the predicted direction for each problem. Ask what would happen if both groups received exactly the same outcome information.
Consider afterwards
- A learner never encounters a rare failure while trying a device. How could you distinguish ignorance of that possibility from underweighting a known risk? Design a comparison that equalizes the information before choice.
Editorial context
The six-problem student experiment demonstrates a consequential information-source distinction. It does not show that experience invariably produces underweighting, or that perceived probabilities alone explain the choices. Hertwig and Erev's later review compares several experience paradigms and discusses sampling and psychological explanations. The paper's opening health example is an illustration, not evidence about actual patient or clinician decisions.