works
Eliezer Yudkowsky Cognitive biases potentially affecting judgement of global risks incollection This article discusses cognitive biases that could potentially affect judgment of existential risks, drawn from research in the heuristics and biases program. It includes anchoring and adjustments biases, where subjects adjust from an initial, often arbitrary value, and underadjust, leading to overconfidence. The conjunction fallacy is discussed too, in which people overestimate the likelihood of events occurring together compared to either event occurring alone. Hindsight bias is also examined, whereby subjects overestimate the predictability of events after they have occurred. The paper then contrasts the availability heuristic, which estimates frequency by accessibility from memory, with the representativeness heuristic, where judgments of likelihood are based on how similar an event is to a prototype. Biases in judgments of risk are categorized into conjunction, disjunction, negativity and positivity effects. It also investigates confirmation bias, where people overvalue evidence that confirms their beliefs and undervalue evidence that contradicts their beliefs. Furthermore, the paper introduces the affect heuristic, where judgments are affected by emotional reaction to a given risk. The article concludes that knowing about these biases can help people make better judgments under uncertainty – AI-generated abstract.

Cognitive biases potentially affecting judgement of global risks

Eliezer Yudkowsky

In Nick Bostrom and Milan M. Ćirković (eds.) Global catastrophic risks, Oxford, 2008, pp. 91–119

Abstract

This article discusses cognitive biases that could potentially affect judgment of existential risks, drawn from research in the heuristics and biases program. It includes anchoring and adjustments biases, where subjects adjust from an initial, often arbitrary value, and underadjust, leading to overconfidence. The conjunction fallacy is discussed too, in which people overestimate the likelihood of events occurring together compared to either event occurring alone. Hindsight bias is also examined, whereby subjects overestimate the predictability of events after they have occurred. The paper then contrasts the availability heuristic, which estimates frequency by accessibility from memory, with the representativeness heuristic, where judgments of likelihood are based on how similar an event is to a prototype. Biases in judgments of risk are categorized into conjunction, disjunction, negativity and positivity effects. It also investigates confirmation bias, where people overvalue evidence that confirms their beliefs and undervalue evidence that contradicts their beliefs. Furthermore, the paper introduces the affect heuristic, where judgments are affected by emotional reaction to a given risk. The article concludes that knowing about these biases can help people make better judgments under uncertainty – AI-generated abstract.

PDF

First page of PDF