Paper Number

ICIS2026-3466

Paper Type

Short

Abstract

Digital labor platforms make worker faces visible, allowing a profile photograph to convey categorical information about perceived racial presentation and evaluator-specific relational information through buyer-seller facial similarity. We examine how these cues jointly shape hiring on Fiverr. Study 1 analyzes observed-hiring and comparison dyads using two-way clustered linear probability models, residualized similarity, and seller-stratified conditional logit. Facial similarity is positively associated with hiring, but it does not uniformly narrow racial disparities. Black-presenting sellers face the clearest baseline disadvantage and show no compensatory similarity effect, whereas race-specific associations emerge for Middle Eastern- and White-presenting sellers. Building on the continuum model of impression formation, we theorize that evaluator-specific relational information receives less evaluative weight when judgments remain strongly anchored in categorical information. Study 2 uses GenAI as a planned causal probe of this weighting mechanism under race-preserving facial manipulation. The findings identify a boundary on relational updating under persistent categorical visibility.

Comments

06_Societal_impacts

Share

COinS
 
Dec 14th, 12:00 AM

Designing for Relational Similarity: Facial Similarity and Racial Disadvantage in Digital Labor Platforms

Digital labor platforms make worker faces visible, allowing a profile photograph to convey categorical information about perceived racial presentation and evaluator-specific relational information through buyer-seller facial similarity. We examine how these cues jointly shape hiring on Fiverr. Study 1 analyzes observed-hiring and comparison dyads using two-way clustered linear probability models, residualized similarity, and seller-stratified conditional logit. Facial similarity is positively associated with hiring, but it does not uniformly narrow racial disparities. Black-presenting sellers face the clearest baseline disadvantage and show no compensatory similarity effect, whereas race-specific associations emerge for Middle Eastern- and White-presenting sellers. Building on the continuum model of impression formation, we theorize that evaluator-specific relational information receives less evaluative weight when judgments remain strongly anchored in categorical information. Study 2 uses GenAI as a planned causal probe of this weighting mechanism under race-preserving facial manipulation. The findings identify a boundary on relational updating under persistent categorical visibility.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.