Studies in subjective probability l: Prediction of random events

When students are asked to predict the outcome of a random event, where all alternatives are equally probable (lotteries), they tend to choose central, "representative" values, and avoid extreme ones. In ten informal experiments, it is shown how this pattern of choices is influenced by various procedural and structural changes in the basic task. The results show that guessing behavior can be described as a kind of absolute judgment, subject to grouping, anchoring and context effects. Of the two general prediction heuristics originally proposed by Kahneman & Tversky (1973), "representativeness" applies better than "availability". In fact, a major strategy of guessing is apparently to eschew numbers with prominent, "non-random" properties, which at the same time are highly available to the subjects.