A Tale of Two Identities: An Ethical Audit of Human and AI-Crafted Personas

2025-05-07

AI Summary

This study addresses representational bias against racial minorities in synthetic personas generated by large language models (LLMs)—including GPT-4o, Gemini 1.5 Pro, and DeepSeek-V2.5—in data-scarce, high-stakes domains such as health and privacy. Auditing 1,512 synthetic personas, it focuses on race-related representational harms. Methodologically, it employs a mixed-methods approach: close reading, lexical frequency analysis, parametric creativity assessment, and integrates human-grounded benchmarks with community-centered validation protocols. The work introduces the novel concept of “algorithmic othering”—a paradox wherein models over-assign racial labels while eroding identity authenticity—and proposes a narrative-aware evaluation framework. Findings reveal systemic socio-technical harms, including stereotyping, exoticization, benevolent bias, and silencing. Collectively, the study advances both theoretical foundations and practical guidelines for accountable, culturally responsive synthetic identity generation.

Technology Category

Natural Language Processing: Ethics — Bias, Fairness, Transparency & PrivacyPhilosophy and Ethics of AI: Bias, Fairness & EquityMachine Learning: Ethics, Bias, and Fairness

Application Category

Social Networks and Social Media: Generative AI / large language models and their impact on social systemsUser Modeling, Personalization and Recommendation: Fairness-aware retrieval and rankingEconomics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments

Abstract

As LLMs (large language models) are increasingly used to generate synthetic personas particularly in data-limited domains such as health, privacy, and HCI, it becomes necessary to understand how these narratives represent identity, especially that of minority communities. In this paper, we audit synthetic personas generated by 3 LLMs (GPT4o, Gemini 1.5 Pro, Deepseek 2.5) through the lens of representational harm, focusing specifically on racial identity. Using a mixed methods approach combining close reading, lexical analysis, and a parameterized creativity framework, we compare 1512 LLM generated personas to human-authored responses. Our findings reveal that LLMs disproportionately foreground racial markers, overproduce culturally coded language, and construct personas that are syntactically elaborate yet narratively reductive. These patterns result in a range of sociotechnical harms, including stereotyping, exoticism, erasure, and benevolent bias, that are often obfuscated by superficially positive narrations. We formalize this phenomenon as algorithmic othering, where minoritized identities are rendered hypervisible but less authentic. Based on these findings, we offer design recommendations for narrative-aware evaluation metrics and community-centered validation protocols for synthetic identity generation.

Problem

• Audit ethical impacts of AI-generated personas on minority identities
• Compare racial representation in AI vs human-crafted personas
• Address algorithmic othering and stereotyping in synthetic identity generation

Innovation

• Audit synthetic personas using mixed methods
• Compare LLM and human-authored identity narratives
• Propose narrative-aware evaluation metrics