Irrational base-rate use in human contingency learning

Abstract

People often fail to use base-rate information appropriately in decision-making. This is evident in the inverse base-rate effect, a phenomenon in which people tend to predict a rare outcome for a new and ambiguous cue combination. Previous accounts of this effect have appealed to either learned attention biases or inferential reasoning strategies. Yet, questions still remain about the processes involved. This thesis therefore examined the psychological factors underlying the inverse base-rate effect. Chapter 3 investigated the parameters of the effect to characterise inferential strategies that could produce the rare outcome bias. Experiments showed that rare choice is not based on transfer trial novelty, and that a combination of relative trial and global outcome frequency differences is important for the development of rare outcome biases. However, there was no strong evidence for widespread use of a single reasoning strategy. Chapter 4 found that two prominent models of attention in learning, EXIT and Mackintosh, could reasonably account for these effects. Although derived from the same theoretical principles, these models made diverging predictions for relative levels of attention paid to cues during learning. Chapter 5 tested these predictions by measuring changes in cue associability and fixation to cues. Results showed greater attentional benefit for rare than common predictors, which appeared to be moderated by learning about the context. These results are more consistent with the predictions of the EXIT model than those of the Mackintosh model. Chapter 6 showed that the effect is more reliably observed in a discrete outcome choice test phase than in a continuous outcome rating test phase, highlighting differences in the sensitivity of these measures in observing learning effects, particularly on ambiguous test trials. Chapter 7 discusses implications for prominent explanations of the effect, and for theories of learning and decision-making more generally.