Abstract:

Subjective probabilities are probabilities people express for uncertain events or outcomes. They are generated, or judged, by two major heuristics: 1. When outcomes are unique (e.g., the guilt of some defendant) or set in the future (e.g., the winner of the next election), the approach is "theoretical". People pull together whatever they know, or believe, to be relevant, and judge the probabilities of the possible outcomes by the closeness of the match between them and whatever "prediction model" they have built in their heads. This heuristic is called representativeness. 2. When outcomes are grouped in categories or by features (e.g., the percent of convictions for a given charge, or the percent of elections won by incumbents), the approach is "empirical": Let's sample what's out there and count. If the sampling is done in one's head, and the probabilities judged by the number of examples that come to mind, or by the ease – real or anticipated – with which they come to mind, the heuristic is that of availability. These heuristics have distinct signatures. They lead to predictable and systematic biases, among them: the extension fallacy, the base-rate fallacy, sample size neglect, regression neglect, the unpacking effect, overconfidence, hindsight bias and more.

Website