Work

Aligned but Blind: Alignment Increases Implicit Bias by Reducing Awareness of Race

Year: 2025

Type: conference-paper

Abstract: Although value-aligned language models (LMs) appear unbiased in explicit bias evaluations, they often exhibit stereotypes in implicit word association tasks, raising concerns about their fair usage.We investigate the mechanisms behind this discrepancy and find that alignment surprisingly amplifies implicit bias in model outputs.Specifically, we show that aligned LMs, unlike their unaligned counterparts, overlook racial concepts in early internal representations when the context is ambiguous.Not representing race likely fails to activate safety guardrails, leading to unintended biases.Inspired by this insight, we propose a new bias mitigation strategy that works by incentivizing the representation of racial concepts in the early model layers.In contrast to conventional mitigation methods of machine unlearning, our interventions find that steering the model to be more aware of racial concepts effectively mitigates implicit bias.Similar to race blindness in humans, ignoring racial nuances can inadvertently perpetuate subtle biases in LMs. 1

Source: Proceedings of the conference - Association for Computational Linguistics. Meeting

DOI: 10.18653/v1/2025.acl-long.1078

Authors: Lihao Sun, Chengzhi Mao, Valentin Hofmann, Xuechunzi Bai

Institutions: University of Chicago, Rutgers, The State University of New Jersey, University of Washington, Allen Institute for Artificial Intelligence

Language: English

Topic: Social and Intergroup Psychology

Subfield: Sociology and Political Science

Field: Social Sciences

Domain: Social Sciences

SDG: Reduced inequalities

Open Access status: bronze

Funder: University of Chicago