Choice evaluation and context effects

Abstract

Behavioral research has long documented that the choices an individual makes do not always follow the maximization of expected values. To describe the utility an individual maximizes through his or her choices, one class of models - static models - has been previously developed. These models are reviewed in Chapter 1. To assess the static models, a non-parametric method to reveal the utility of alternatives is developed in Chapter 2. The results show that the utility predicted from the static models deviates from the estimated utility. Utility, however, is relatively unstable across contexts determined by information presentation formats, choice set-sizes, the structures of alternatives, and the relationships between alternatives. This instability is a topic for Chapters 3, 4, and 5. Following Chapter 3, which examines effects of information presentation formats and choice set-sizes on risk-taking, Chapter 4 further investigates how the contexts impact on choice evaluation. Then, Chapter 5 examines process of choice evaluation by analyzing eye-movements during choices. The results from these three chapters indicate that choices are systematically altered with contexts, supporting instability of utility. The instability of utility conflicts with the principle of utility maximization, and Chapters 5 and 6 consider another class of models - dynamic models - which can accommodate utility instability. A dynamic model assumes that an individual iteratively and stochastically develops preferences for each alternative, until preference for one alternative reaches a choice criterion. The exact processes of preference development is investigated in Chapter 5, which suggests that a dynamic model should be based on single-attribute pair-wise comparisons. Following this suggestion, a new model - multi-alternative decision by sampling - is proposed in Chapter 6. Chapter 7 discusses overall implications of the results for the principle of utility maximization and model evaluation. I conclude that models should be assessed not only on their ability to predict choices but also on their ability to predict concurrent process measures, including eye-movements.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 0 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.