Deciding When to Stop: Cognitive Models of Sequential Decisions in Optimal Stopping Tasks

Many everyday decisions are sequential. We evaluate consecutive alternatives over time anddecide when to stop a search of options to make a conclusive decision. Research has shown thatpeople do not optimize the stopping point of the sequential search process, and instead there is abias from the optimal stopping point. Current psychological models rely on the use of a decision“threshold,” suggesting that humans adjust their aspirations throughout a sequence. Importantly,these models do not describe thecognitive processbehind stopping decisions. They are only appli-cable to tasks with specific characteristics and cannotgeneralizeto other tasks without significantmodifications.  They also assume that there is no learning from experience over repetitions ofthe task. Our research provides an integrated cognitive account of the process of stopping deci-sions in sequential tasks. We propose a model that relies on a generic inductive process by whichthese thresholds emerge, based on a known theory of decisions from experience:Instance-BasedLearning Theory (IBLT). The Instance-Based Learning (IBL) model makes decisions by learningthrough feedback from past decisions, considering the current alternative’s value and the numberof alternatives remaining in the sequence, without relying on the concept of a threshold. We simu-late the choices that individuals would make in two different optimal stopping tasks using the IBLmodel and compare the simulation results with empirical data and two existing cognitive models.We show that the IBL model is capable of makinga prioripredictions, in the absence of data,of human stopping decisions across two tasks. The IBL model stopping decisions are comparableto those produced by models that were developed to fit the human data separately in each task.Our results demonstrate that the IBL model describes human behavior and generates individu-alized thresholds similar to those generated by other models without assuming that people makedecisions using thresholds. In general, our approach provides an integrated cognitively plausibleprocess through which stopping decisions are made in sequential decision tasks, proposing thatthresholds emerge through learning from experience.

Erin Bugbee - 2nd Year Paper