【公开学术报告】When Does Unequal Representation Reflect Bias?

When Does Unequal Representation Reflect Bias? The Impact of Political Ideology on Judgments of Distributional Outcomes

Guest Speaker:   Jin Kim (Yale School of Management)

Date & Time:      9:30-11:00 (Beijing Time), Tue. 25th, Oct. 2022

Zoom Meeting:   848 547 50055（Password: 293870）

Click the Link:    https://us02web.zoom.us/j/84854750055

ABSTRACT

It is commonly asserted that an organization, a group, or a person is biased against some target group. Such judgments are often based on an observed imbalance in a distributional outcome, namely the under- or over-representation of a target group relative to some baseline (e.g., the population base-rate). Our research examines how individuals judge distributional outcomes as representing “bias” in decisions, such as hiring and admissions. Results across 27 studies (N=15,187) suggest that such judgments of bias depend on both the target’s characteristics (whether the target of bias is traditionally dominant, known, or ideologically relevant) and the observer’s political ideology (liberal vs. conservative). Specifically, when traditionally nondominant targets (e.g., women, Blacks, immigrants) face unfavorable distributional outcomes, conservatives have a higher threshold for judgment of bias than liberals (i.e., require greater imbalances against the target before recognizing “bias”). In contrast, when traditionally dominant targets (e.g., men, Whites, native-born citizens) face unfavorable distributional outcomes, now liberals have a higher threshold for judgment of bias than conservatives. Such relationships between political ideology and judgments of bias significantly weaken for unknown targets (e.g., “Type Js”) or ideologically irrelevant targets (e.g., animals). These robust results suggest that judgments of bias themselves may be biased—as the judgments vary with target’s characteristics, observer’s political ideology, and their interaction. Importantly, determining who is more biased depends on the standard of comparison. Our investigation highlights the methodological imperative of stimulus sampling and selecting the right control, and thus contributes to the ongoing debate on the ideological (a)symmetry hypotheses.

Significance Statement

There has been extensive debate about the prevalence of bias based on observed representation, with limited systematic investigation on factors affecting judgments of bias regarding distributional outcomes. Our research provides methodological and theoretical contributions within this context. Methodologically, we introduce an explicit measure of judgments of bias and demonstrate the importance of (i) stimulus sampling—because judgments of bias vary with target’s characteristics—and (ii) selecting the right control—because determining bias in judgments of “bias” depends on standards of comparison (e.g., bias relative to base-rates or unknown targets). Theoretically, our findings contribute to the ongoing debate about the (a)symmetry hypothesis by suggesting that we must consider what constitutes bias in judgment before we attempt to answer who is more biased.