Perspectives on the mechanistic underpinnings of choice biases

Early foundational work in the decision sciences carefully balanced empirical observations and theoretical explanations. Dating back to Daniel Bernoulli, a handful of behavioral regularities observed in theoretical lotteries ignited the refinement of normative theories and the development of new descriptive frameworks of valuation and choice. However, more recent tendencies in behavioral economics and psychology place empirical observations on a pedestal: modern behavioral science has identified more behavioral biases than it has explained. Coupled with replication and reliability crises in experimental psychology, this has resulted in an explanatory gap in the field, in-between the descriptive and predictive levels. Here, we aim to close this explanatory gap by asking how choice biases can emerge from certain decision computations. We demonstrate that biased and irrational choice behavior may arise from multiple, equally viable mechanisms, such as relative value coding and selective information sampling. We posit that this “multiple realizability” problem highlights a broader issue: inferring mechanisms of complex behavior solely from behavioral measures is an underdetermined exercise. We propose that using time-resolved neural recordings to track how attention serially parses complex information during multiattribute, multialternative decisions can resolve this “multiple realizability” issue and arbitrate between competing mechanistic explanations of choice biases.