Noisy retrieval models of over- and undersensitivity to rare events

Research output: Contribution to journalJournal articleResearchpeer-review

Abstract

Review of previous research highlights 2 pairs of inconsistent reactions to rare events: (a) studies of probability judgment reveal conservatism (that implies overestimation of rare events) and overconfidence (that implies underestimation of rare events); (b) studies of choice behavior reveal overweighting of rare events in 1-shot decisions under risk, and underweighting of rare events in repeated decisions from experience. The current analysis shows that the 4 biases are not inconsistent, but that they can be the product of 2 distinct sources of errors that affect human judgment and decision making: reliance on small samples of past experiences, and overgeneralization, that is, the tendency to confound previously encountered tasks. The former is a sufficient condition for overconfidence and underweighting of rare events, whereas the latter is sufficient for conservatism and overweighting of rare events. Importantly, this "2-error" explanation leads to a novel prediction: It implies that overconfident judgment triggers overweighting of rare events in 1-shot decisions under risk, in which the relevant objective probabilities are known to decision makers. This prediction is based on the assumption that the overgeneralization that affects choice behavior is a reflection of the perceived similarity between decisions under risk and decisions with estimated risk, in which decision makers can only rely on (overconfident) subjective estimates of the relevant probabilities. The value of this analysis and of a computational abstraction of the "2-error" hypothesis is demonstrated in 3 experimental studies.

Original languageEnglish
JournalDecision
Volume2
Issue number2
Pages (from-to)82-106
ISSN2325-9965
DOIs
Publication statusPublished - 2015

Keywords

  • Black swan
  • Experience-description gap
  • Prospect theory
  • Truth plus error
  • Won't happen to me

Fingerprint

Dive into the research topics of 'Noisy retrieval models of over- and undersensitivity to rare events'. Together they form a unique fingerprint.

Cite this