Comparing value coding models of context-dependence in social choice

Abstract

Decision-makers consistently exhibit violations of rational choice theory when they choose among several alternatives in a set (e.g., failing to buy the best product in a set when it is presented alongside high-quality alternatives). Many of society’s most significant social decisions similarly involve the joint evaluation of multiple candidates. Are social decisions subject to the same violations, and if so, what account best characterizes the nature of the violations? Across five studies, we tested whether decision-makers exhibit context-dependent preferences in hiring scenarios and past U.S. congressional race outcomes and co¬mpared different models of value coding as sources of the hypothesized context-dependence. Studies 1a, 1b, and 1d revealed that a divisive normalization value coding scheme best characterized participants’ choices across a series of hiring decisions, and that participants exhibited context-dependent preferences. However, the distractor had the opposite effect of that predicted by divisive normalization once we accounted for the random effect of participant: as the value of the distractor increased, participants were more likely to hire the highest-valued candidate. In Study 2, we used a combination of archival electoral data and survey data to examine whether normalization models could explain the outcomes of congressional elections. Electoral outcomes were predicted by political candidates’ inferred competence, but this time in line with the divisive normalization account. Our findings offer mixed support for a formal, neurobiologically-derived account of when and how specific alternatives exert their effects on social evaluation and choice and highlight conditions under which high-value distractors increase versus decrease relative choice accuracy.

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 — 33 citations exist upstream.

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