The 'catch all underestimation bias': availability hypothesis vs. category redefinition hypothesis

When branches of a fault are pruned, subjects do not fully transfer the probability of those branches to the all other category (cf. Fischhoff, Slovic, & Lichtenstein, 1978). The underestimation of the probability of the catch-all category has been explained by either an availability or a category redefinition mechanism. In this experiment, subjects received a full and a pruned fault tree and estimated the probabilities of branches related to a starting failure problem or an adult death causes problem. To increase the availability of the omissions, subjects in the filling-task condition used the same fault tree for both the frequency estimation task and the filling task. That is, they had to write down specific causes related to each branch of the tree before estimating the probability of the same branches. The type of omission was also manipulated. A high probability branch was omitted in the important omission condition whereas a low probability branch was omitted in the unimportant omission condition. The results showed that the bias did not disappear in each of the six experimental conditions. Furthermore, nearly three fourths of the subjects exhibited the bias. Finally, the bias was higher when a high probability branch was omitted than when a low probability one was omitted. The validity of the two explanations is analyzed.