Description, Prediction and Origin of Human Decision-Making: A Multidisciplinary Approach Towards a Unified Choice Theory

In light of the descriptive inadequacy of rational choice models like Expected utility, the over-arching theme of my doctoral studies has been how real human beings make decisions under uncertainty, as opposed to what they should do. Specifically, we focused on developing operational models of individual decision-making in risky and ambiguous setups, representing the decision activity as a stochastic process and working at different levels of abstraction (axiomatic models, low-dimensional stochastic processes and computational models). Characterizing the psychological activity underlying our decisionmaking, we aim not only to describe observed behavior, but to unravel its “computational” origin. In the first part of the thesis, we investigate axiomatic models of choice. Relying on the formalism of quantum mechanics to describe human decision-making, we propose another parametrization of Quantum Decision Theory (QDT), based on Rank Dependent Utility Theory (RDU). We calibrate and compare it to several classical theories using experimental data on choices between pairs of lotteries, both at the aggregate and at the individual level. The results: i) confirm the inadequacy of the representative agent approach; ii) provide important quantitative evidence in favor of quantum decision theory; iii) contribute to our understanding on the stability of risk-preference and asset-integration attitude. In the second part, we develop a model where a stochastic process is assumed to capture the fuzzy deliberation human activity leading to a decision (evidence accumulator model). The key novelty, with respect to common expectation-like models, is the non-trivial interaction between probability judgment and value perception. Such nonseparability of tastes and beliefs offers an alternative explanation for the majority of human paradoxes, as opposed to all theories dealing with some sort of distorted mathematical expectation. Moreover, it can account for the effect of time pressure on risky choice with significantly less parameters than the popular Decision Field Theory. In the third part, we investigate discrete time quantum walks (DTQW) as an interesting formalism to model human decision making. The driving idea is to use DTQW to combine the quantum formalism of cognition (firt part) and evidence accumulator models (second part). We show how quantum walks predict observed probabilistic misperception like S-shaped subjective probability and conjunction fallacy. Our framework emphasizes the close relationship between response times and type of preferences and of responses. In the final part, we go from an effective representation approach (such as the one in the second part) to a more micro-founded description of the human stochastic deliberation process. In essence, the human brain is seen as a set of simplified neurons (Pott’s spins) embedded in a network, which interact via excitatory couplings to give a collective decision that we observe at the macro level. Our model is shown to represent a dual process of choice, capturing the widely believed idea that many phenomena in decision-making result from the interaction of 2 systems of reasoning. Moreover, it offers insights on why people exhibit “varying preferences”, in terms of how the perceived complexity of a given choice task impact the dual choice process