Research map / Large language models / Language, safety and interpretability
Fairness Perceptions of Large Language Models
Benjamin Cookson, Soroush Ebadian, Nisarg Shah
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 |
|---|---|---|---|
| ask LLM to be fair (default prompt) vs MNW baseline | GPT-4o, Claude-3.5-Sonnet, Gemini-1.5-Pro · Synthetic fair division | EF1 vs SW approximation: EF1 17-36% vs MNW 100%; SW 92-94% vs 86% | Hurt |
| increase utility variation lambda | LLMs (default prompt) · Synthetic fair division | EF1 approximation | Hurt |
| change allocation context to Sibling/Heirloom or Team/Machine (ablation) | LLMs · Synthetic fair division | EF1 approximation | No clear effect |
| present valuations grouped by goods (Object/Person) (ablation) | Claude and Gemini · Synthetic fair division | EF1 approximation / SW: lower EF1, higher SW | Mixed |
| present valuations grouped by goods (Object/Person) (ablation) | GPT-4o · Synthetic fair division | EF1 and SW approximation | No clear effect |
| prompt explicitly for EF1 fairness instead of 'fair' (ablation) | LLMs · Synthetic fair division | EF1 approximation | Improved |
| frame EF1 as purely combinatorial problem without fairness language (ablation) | GPT and Gemini · Synthetic fair division (varying lambda) | EF1 approximation | Improved |
| frame EF1 as purely combinatorial problem (ablation) | Claude · Synthetic fair division (varying lambda) | EF1 approximation | No clear effect |
| use real Spliddit instances instead of synthetic ones | LLMs · Spliddit.org | EF1 approximation | Improved |
Closest papers by meaning
- Distributive Fairness in Large Language Models: Evaluating Alignment with Human Values (NeurIPS 2025 · similarity 0.93)
- Prompting Fairness: Integrating Causality to Debias Large Language Models (ICLR 2025 · similarity 0.88)
- Is Your Model Fairly Certain? Uncertainty-Aware Fairness Evaluation for LLMs (ICML 2025 · similarity 0.88)
- Decision-Making Behavior Evaluation Framework for LLMs under Uncertain Context (NeurIPS 2024 · similarity 0.87)
- Guiding LLM Decision-Making with Fairness Reward Models (NeurIPS 2025 · similarity 0.87)
- LLM Agents Can Be Choice-Supportive Biased Evaluators: An Empirical Study (AAAI 2025 · similarity 0.86)
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.
