Research map / Large language models / AI and society
A Tale of Two Identities: An Ethical Audit of AI-Crafted Synthetic Personas
Pranav Narayanan Venkit, Jiayi Li, Yingfan Zhou, Sarah Rajtmajer, Shomir Wilson
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add more demographic detail to persona prompt (race only to full profile) (ablation) | LLM persona generation · Persona semantic diversity | diversity: 1.621 -> 1.713 (human 1.910) | Improved |
| add more demographic detail to persona prompt (ablation) | LLM persona generation · Persona semantic novelty | novelty: 0.453 -> 0.314 (human 0.105) | Improved |
| add more demographic detail to persona prompt (ablation) | LLM persona generation · Persona surprisal | surprisal: 1.165 -> 1.267 (human 1.276) | Improved |
| add more demographic detail to persona prompt (ablation) | LLM persona generation · Persona complexity | complexity: 1.720 -> 1.717 | No clear effect |
| use LLM generation instead of human self-description | Persona text · Semantic diversity | diversity: 1.910 -> 1.713 | Hurt |
| use LLM generation instead of human self-description | Persona text · Semantic diversity, minoritized groups (race only) | diversity: White 1.684 vs Asian 1.110, AI/AN 1.101 | Hurt |
| use LLM generation instead of human self-description | Persona text · Semantic novelty | novelty: 0.105 -> 0.314 to 0.453 | Mixed |
| use LLM generation instead of human self-description | Persona text · Semantic complexity | complexity: 0.656 -> 1.717 | Mixed |
| use LLM generation instead of human self-description | Persona text · Surprisal | surprisal: 1.276 -> 1.267 (full profile) | Hurt |
Closest papers by meaning
- LLM Generated Persona is a Promise with a Catch (NeurIPS 2025 · similarity 0.87)
- Persona Generators: Generating Diverse Synthetic Personas for Arbitrary Contexts (NeurIPS 2026 · similarity 0.87)
- Bias Runs Deep: Implicit Reasoning Biases in Persona-Assigned LLMs (ICLR 2024 · similarity 0.87)
- More of the Same: Persistent Representational Harms Under Increased Representation (NeurIPS 2025 · similarity 0.87)
- StereoTales: A Multilingual Framework for Open-Ended Stereotype Discovery in LLMs (NeurIPS 2026 · similarity 0.86)
- A Synthetic Dataset for Personal Attribute Inference (NeurIPS 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
