Context Effects, Value Learning, and Individual Differences in Value-Based Decisions

Abstract

In this dissertation, we present two lines of research that investigate value-based decisions. The first focuses on value-based decisions in multi-attribute choices. Specifically, we investigate contextual preference reversals in multi-attribute decisions. These reversals occur when the choice preference between two options changes in the presence of a third unchosen decoy. We demonstrate for the first time that these reversals also occur in ethical dilemmas involving both quantitative and qualitative attributes. However, these reversals do not arise to the same extent across ethical dilemmas. We use a generative computational model to show that the variation of reversals across dilemmas can be partly explained by individual differences in rankings of ethical features. The second line of work focuses on value-based decisions in the Value Learning Task (VLT; Raymond & O’Brien, 2009), a paradigm where people learn values associated with options in win and loss conditions through trial-and-error while trying to maximize accumulated reward. The VLT has a symmetric outcome structure for wins and losses. However, people consistently learn wins better than losses (Lin et al., 2020). We investigate the nature of this asymmetry with a simple reinforcement learning model. The model predicts the learning asymmetry observed in empirical data regardless of whether the parameters are set to maximize empirical fit or total payoff in the task. This asymmetry arises as a result of the interaction between a neutral initial value estimate and a choice policy that exploits while exploring, leading to more poorly discriminated value estimates for loss stimuli. We also illustrate that the final value estimates produced by the model can provide a simple account of a post-learning explicit value categorization task. Lastly, we also investigate how differences in estimated individual learning rates help to explain individual differences in the observed win-loss asymmetries. Together, these two lines of research investigate some complicated aspects of value-based decisions such as value learning through experience and attribute-value integration and evaluation in multi-attribute choices. However, beyond the complex phenomena, our two lines of work integrate the same simple theory — the bounded rationality framework. In this dissertation, we also discuss how our research connect to the bounded rationality framework.