Apparently Irrational Choice as Optimal Sequential Decision Making

In this paper, we propose a normative approach to modeling apparently human irrational decision making (cognitive biases) that makes use of inherently rational computational mechanisms. We view preferential choice tasks as sequential problems and formulate them Partially Observable Markov Decision Processes (POMDPs). The resulting model learns what information gather about which options, whether calculate option values or make comparisons between options when choice. apply the where context is known influence choice, an effect has been taken evidence cognition irrational. Our results show new approximates bounded optimal cognitive policy quantitative predictions correspond well Furthermore, uses help infer maximum expected value while taking into account cost limits. addition, it predicts when, explains why, people stop accumulation decision. argue provides apparent irrationalities are emergent consequences processes prefer higher (rational) policies.