The sensitivity of probability assessments to time units and performer characteristics

This paper examines the sensitivity of probability judgments pertaining to real-world events (e.g., the chance of getting a speeding ticket) to the events' time horizon (e.g., 1, 2, or 8 weeks) and performer characteristics (e.g., 1, 2, and 8 drivers). Using the participants' multiple responses to different time horizons and number of performers, we assessed the degree of deviance from the normative prescription and potential explanations for such deviance. In Study 1, two questionnaires were administered to a sample of 120 students. In the first questionnaire, in violation of the normative disjunctive model, most participants did not adjust probabilities sufficiently for various time horizons. This pattern cannot be attributed merely to regression to the mean with a fixed error term (Erev et al.). In the second questionnaire, the same bias was replicated in scenarios featuring different numbers of individuals, indicating no time-specific contributing factors. In Study 2, a sample of 140 students responded to a computerized questionnaire in which the items preceding target questions were controlled. The direction of the deviance did not change as a function of previous items. The results are attributed to the ‘extension neglect’ tendency (Kahneman).