Methodological and theoretical aspects of context effects in decision making

Contextual influences on human choice have been well documented in the literature, suggesting that preference is constructed during the decision making process based on the set of alternatives humans are confronted with. However, these so-called context effects violate axiomatic choice principles of economic and discrete choice models. In this dissertation, I investigate methodological (first part) and theoretical (second part) aspects of these contextual violations of axiomatic principles. In the first part of my dissertation, I focus on a commonly used empirical measure of context effects (the relative choice share of the target; RST) and I show that the RST can and does establish false empirical conclusions, based on simulations and applications of the RST to published research results (first manuscript). For this reason, I propose an unbiased modification of the RST. In addition, I show that, generally, the RST is the appropriate empirical measure of some context effects (i.e., similarity and compromise effects) but not others (i.e., attraction effect), where the absolute choice share of the target (AST) and the competitor (ASC) should be used instead. In the second part of my dissertation, I focus on the attraction effect and investigate its boundary conditions through a research synthesis (second manuscript). Specifically, I conduct a meta-analysis with more than 30,000 participants and 300 effect sizes spanning four decades of research. I conclude an average (but small) underlying attraction effect and
I establish the role of three factors in strengthening the attraction effect: numeric attribute values, between-subjects experimental designs, and lower decoy choice share. In the third manuscript, I build and apply a new cognitive sequential sampling model of context effects in order to investigate how group level violations of axiomatic principles can stem from variability in (risk/attribute) preferences at the individual level (third manuscript). I apply the model to data from a new experiment with the attraction, similarity and compromise
effects across the preferential (risky choice) and perceptual (area discrimination) choice domain. The empirical results show strong similarity and reversed compromise effects in the preferential domain but no context effects in the perceptual domain. Furthermore, a model without a context-effect-specific mechanism at the individual level best describes group level context effects in preferential choice; thus providing a cognitive explanation of how violations of axiomatic principles can, sometimes, arise only at the group level.