Rational Perspectives on the Role of Stimulus Order in Human Cognition

Minimizing unwanted order effects is one of the most basic design strategies in studies of human behavior. While this practice provides the necessary simplification and control for laboratory studies, it has led to the view that sensitivity to stimulus order is a "bug" in human cognition, due to either constraints on cognitive resources or misconceptions about probability axioms. In this dissertation, I explore the possibility of adopting a rational perspective on the role of stimulus order---instead of being regarded as an extraneous variable, stimulus order may serve as an important statistical cue for inferring the true theory that accounts for a task environment. I present both computational and behavioral studies demonstrating that such a perspective not only applies to scenarios where the sequential order of stimuli is intuitively central to the task of interest, but also to other situations where stimulus order is a seemingly inconsequential factor. Overall, the current studies suggest that the effects of stimulus order on human cognition need not be interpreted exclusively via accounts focusing on how humans process information. Rather, certain aspects of human behavior can be understood as the result of inferring hypotheses that explain the order of stimuli in the input.