|dc.description.abstract||Many of society's most significant social decisions involve the joint evaluation of multiple candidates, and yet, we know that decision-makers consistently exhibit violations of rational choice theory when they choose among several alternatives in a set. Furthermore, context-dependence in social choice is well-documented but underexplored as a method for reducing decision inaccuracy or bias. Across three papers, I explored context-dependence in employment decisions.
In the first paper (Chang, Gershman, & Cikara, 2019), I evaluated two competing value coding models of context-dependence derived from computational neuroscience. Normalization, a canonical computation in various neural systems, refers to scaling inputs by other nearby inputs to reduce redundancy in signal processing. The first model - divisive normalization - proposes that the value of each option is scaled by the summed value of all options in the choice set. The second model - range normalization - proposes that the value of each option is scaled by the absolute difference of the highest and lowest value options in the choice set. Using a combination of archival electoral data and new survey data, I found mixed support for the divisive normalization model in predicting hypothetical hiring decisions and past U.S. congressional race outcomes based on the perceived facial competence of the candidates.
In the second paper (Chang & Cikara, 2018), I explored the efficacy of asymmetric dominance and compromise social decoys and tested whether choice set construction could be used to change preferences in hiring. I found that participants have systematically different preferences for the exact same candidate as a function of the other candidates in the choice set and the salience of the candidate attributes under consideration. Importantly, I demonstrated that I could mimic the social decoy effect in the absence of a third candidate by manipulating participants’ exposure to candidates’ attributes: balanced exposure to candidates’ warmth and competence information significantly reduced bias between the two candidates. While hiring decisions often involve joint evaluation of a set of candidates, as in the first two papers, other decisions such as promotion decisions often require separate evaluation of an individual candidate. In these kinds of decisions, it is often impossible to remove demographic information.
In the third paper, I examined whether committing to an importance ranking of evaluation criteria before seeing information about a candidate decreased gender bias, xenophobia, and anti-fat bias in separate evaluation decisions. I found that pre-commitment to evaluation criteria systematically increased decision accuracy: participants made better decisions by making more fine-grained distinctions between different levels (e.g., very experienced vs. somewhat experienced) of relevant attributes (e.g., previous experience). This resulted in less reliance on irrelevant demographic information. Furthermore, using choice-based conjoint experiments, I identified the mechanism of pre-commitment as increased focus on decision-relevant attributes. Leveraging what we know about context-dependence in social decisions can refine theorizing in social cognition and has practical implications for how we structure decision-making processes in consequential social domains for equitable outcomes.||