Research map / Large language models / AI and society
Towards Socially Responsible AI: Cognitive Bias-Aware Multi-Objective Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 bias-aware gender and race objectives instead of bias-agnostic L2 training | Emotion classification on EEC · EEC, None | accuracy: 0.8620 -> 0.9100 | Helped |
| Add the joint gender-and-race bias objectives instead of bias-agnostic L2 training | Emotion classification on EEC · EEC, None | gamma (Fear vs. Gender): 0.1384 -> 0.1609 | Helped |
| Add the joint gender-and-race bias objectives instead of bias-agnostic L2 training | Emotion classification on EEC · EEC, None | gamma (Fear vs. Race): 0.1517 -> 0.1867 | Helped |
| Add the joint gender-and-race bias objectives instead of bias-agnostic L2 training | Emotion classification on EEC · EEC, None | gamma (Anger vs. Gender): 0.1850 -> 0.1906 | Helped |
| Use gender bias objective only rather than gender and race objectives jointly (ablation) | Bias-Awr-Joint · EEC, None | gamma (Fear vs. Gender): 0.1609 -> 0.1986 | Helped |
| Use gender bias objective only rather than gender and race objectives jointly (ablation) | Bias-Awr-Joint · EEC, None | gamma (Fear vs. Race): 0.1867 -> 0.1320 | Hurt |
| Use gender bias objective only rather than gender and race objectives jointly (ablation) | Bias-Awr-Joint · EEC, None | accuracy: 0.9100 -> 0.9430 | Helped |
| Use race bias objective only rather than gender and race objectives jointly (ablation) | Bias-Awr-Joint · EEC, None | gamma (Fear vs. Race): 0.1867 -> 0.1873 | Helped |
| Use race bias objective only rather than gender and race objectives jointly (ablation) | Bias-Awr-Joint · EEC, None | gamma (Fear vs. Gender): 0.1609 -> 0.1403 | Hurt |
| Use gender-neutralized word embeddings instead of standard embeddings | Bias-agnostic emotion classifier · EEC, None | gamma (Fear vs. Gender): 0.1351 -> 0.0000 | Hurt |
| Add gender as an input feature to bias-agnostic training | Bias-agnostic emotion classifier · EEC, None | gamma (Fear vs. Gender): 0.1351 -> 0.0000 | Hurt |
| Add race as an input feature to bias-agnostic training | Bias-agnostic emotion classifier · EEC, None | gamma (Fear vs. Gender): 0.1351 -> 0.0000 | Hurt |
Closest papers by meaning
- Diverse Yet Biased: Towards Mitigating Biases in Generative AI (Student Abstract) (AAAI 2024 · similarity 0.86)
- MABR: Multilayer Adversarial Bias Removal Without Prior Bias Knowledge (AAAI 2025 · similarity 0.85)
- Interpretable Debiasing of Vision-Language Models for Social Fairness (CVPR 2026 · similarity 0.85)
- Beyond Right and Wrong: Evaluating Second-order Social Reasoning in Large Language Models (NeurIPS 2026 · similarity 0.84)
- The Devil is in the Neurons: Interpreting and Mitigating Social Biases in Language Models (ICLR 2024 · similarity 0.84)
- Mitigating Social Bias in Large Language Models: A Multi-Objective Approach Within a Multi-Agent Framework (AAAI 2025 · 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.