WHEN DO WE DECIDE TO STOP? AN INVESTIGATION OF HOW PEOPLE SOLVE THE OPTIMAL STOPPING PROBLEM

Abstract

Humans face sequential decision making tasks where they cannot return to a previous option on a daily basis. These types of problems are known as optimal stopping problems. We designed a simple “Card Stopping Task” to investigate how humans solve optimal stopping problems. In our task, participants were faced with the decision to stop on a known reward value or continue to an unknown future reward value under a variety of horizon conditions, where horizon conditions corresponded to the number of cards remaining. By modeling the behavior from this task, we were able to determine threshold values for stopping decrease and decision noise increases as the number of decisions remaining decreases. The qualitative nature of this resulting behavior was captured in a deep exploration model where a few (1-4) possible future decisions were simulated.