Modeling choice paradoxes under risk: From prospect theories to sampling-based accounts

Important developments in the study of decision making have been based on the establishment and testing of choice paradoxes (e.g., Allais ') that reject different theories (e.g., Expected Utility Theory). One of the most popular and celebrated models in the literature, Cumulative Prospect Theory (CPT), has managed to retain its status despite a growing body of empirical evidence stemming from a collection of choice paradoxes that reject it. Two alternative models, Transfer of Attention Exchange (TAX) and an extension of Decision Field Theory (DFT e ), have been proposed as possible alternatives to CPT. To date, no study has directly compared these three models within the context of a large set of lottery problems that tests different choice paradoxes. The present study accomplishes this by using a large and diverse set of lottery problems, involving both potential gains and losses. Our results support the presence and robustness of a set of 'strong' choice paradoxes that reject CPT irrespective of its parametric form. Model comparison results show that DFT e provides the best account for the present set of lottery problems, as it is able to accommodate the choice data at large in a parsimonious fashion. The success of DFT e shows that many behavioral phenomena, including paradoxes that CPT cannot account for, can be successfully captured by a simple noisy-sampling process. Overall, our results suggest that researchers should move away from CPT, and focus their efforts on alternative models such as DFT e .