The effect of Bayesian feedback on learning in an odds estimation task

Are People Programmed to Commit Fallacies? Further Thoughts about the Interpretation of Experimental Data on Probability Judgment

“…We shall see that the concept of personal probability, or its objectivist counterpart, has an important part to play. Similar results have been reported by several other investigators such as Peterson and Du Charme (19671, Messick and Campos (1972) and Donmell and Du Charme (1975), despite the fact that this allegedly excessive respect for prior probabilities seems difficult to reconcile with subjects' alleged tendency, reported by Kahneman and Tversky (1973), Lyon andSlovic (1976) andBar Hillel (1980), to ignore prior probabilities altogether. In certain easily reproduced circumstances, it is claimed, laymen d o not make big enough revisions of their probability estimates when presented with additional information.…”

“…So the investigators concluded that conservatism in the integration of new information was a pervasive effect, increasing as the diagnostic value of a single chip increased. Similar results have been reported by several other investigators such as Peterson and Du Charme (19671, Messick and Campos (1972) and Donmell and Du Charme (1975), despite the fact that this allegedly excessive respect for prior probabilities seems difficult to reconcile with subjects' alleged tendency, reported by Kahneman and Tversky (1973), Lyon and Slovic (1976) and Bar Hillel (1980), to ignore prior probabilities altogether.…”

Intelligence Preparation of the Battlefield: Critique and Recommendations

“…However, they found that the transfer of this learning to other tasks was moderate, at best. Donnell and DuCharme (1975) have found similar results. As a result, suggested that intelligence analysts' training in probability assessment be confined to the tactical intelligence context.…”

Errors in probabilistic reasoning and judgment biases

“…The group that received posterior-probability feedback underinferred by less than the group that received true-state feedback, but over 200 trials, there was no detectable learning in either group, except possibly very early on. Donnell and DuCharme (1975) found that telling experimental participants the correct posterior probabilities after each of 60 updating problems eliminated their underinference, with almost all of the learning occurring in the first 10 trials. However, when participants were then faced with a new updating problem for which naïve participants tend to infer correctly, they overinferred.…”