Mattias Forsgren: Subjective value and preference

Abstract

Classical social science has often assumed "existing" preferences. These are informed by deeply held beliefs of subjective value and consequently relatively stable and logically coherent. In this view, people know what they like, and this is reflected in their preferences. Psychologists have challenged that account and instead suggested that many preferences are "constructed" in the moment. Such preferences are strongly influenced by the immediate context, making them unstable and potentially incoherent. In this view, people do not know what they like and need to figure it out on the spot. The purpose of this thesis is to critically examine the generality of the constructed preferences view by revisiting phenomena taken to support constructive theories. Study I tested the Decision by Sampling theory which postulates that subjective values of quantitative attributes (e.g. money) are constructed in the moment by ordinal comparisons between the target and a small sample of exemplars from long-term memory and recent experiences. If so, manipulating the distribution of recently experienced attribute values should shift people's subjective values, as revealed by pairwise choices. However, after addressing previous methodological limitations, we found evidence against that prediction. In Study II, we again found no effect of the same manipulation on pairwise choices, only on listwise choice. This effect was weaker for a group with substantial domain expertise. The effect direction is consistent with a previous explanation where the cognitive mechanism described by Decision by Sampling affects the interpretation of numerical information rather than subjective values per se. Study III investigated a criterion of logical coherence of preferences: "transitivity", which implies a mental ranking of options. Preferences for everyday objects like political parties and confectioneries, but also monetary gambles, were near-universally strongly stochastically transitive. That is, participants behaved as-if they knew what they liked but the subjective values were represented with noise. We only estimated a clearly non-zero prevalence of violations of transitivity for monetary gambles designed specifically to be difficult to choose between. To the extent that the scrutinised phenomena were taken to support constructed preferences, that may have been premature. Existing preferences may be more prevalent than commonly recognised.