Research map / Large language models / Language, safety and interpretability
MoCa: Measuring Human-Language Model Alignment on Causal and Moral Judgment Tasks
Allen Nie, Yuhui Zhang, Atharva Amdekar, Christopher J Piech, Tatsunori Hashimoto, Tobias Gerstenberg
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 |
|---|---|---|---|
| scale up model size (GPT3-curie-v1 to GPT3.5-davinci-v2) | GPT-3 family · Causal Judgment | agreement (%): 36.8 -> 37.8 | Improved |
| scale up model size (GPT3-curie-v1 to GPT3.5-davinci-v2) | GPT-3 family · Moral Permissibility | agreement (%): 26.6 -> 32.3 | Improved |
| RLHF (davinci-v2 to davinci-v3) | GPT-3.5 · Causal Judgment | agreement (%): 37.8 -> 41.3 | Improved |
| RLHF (davinci-v2 to davinci-v3) | GPT-3.5 · Moral Permissibility | agreement (%): 32.3 -> 20.2 | Hurt |
| add social-simulacra personas, best persona | GPT3.5-davinci-v2 · Moral Permissibility | agreement (%): 32.3 -> 37.1 | Improved |
| add personas, averaged | GPT3.5-davinci-v2 · Moral Permissibility | agreement (%): 32.3 -> 28.9 | Hurt |
| persona choice (best vs worst) | GPT3.5-davinci-v2 · Moral Permissibility | agreement (%): 37.1 -> 18.5 | Mixed |
| add personas | GPT3.5-davinci-v3 · Causal Judgment | agreement (%): 41.3 -> 39.7 | No clear effect |
| Automatic Prompt Engineer prompt | GPT3.5-davinci-v2 · Moral Permissibility | agreement (%): 32.3 -> 40.4 | Improved |
| Automatic Prompt Engineer prompt | GPT3.5-davinci-v3 · Causal Judgment | agreement (%): 41.3 -> 40.6 | Hurt |
Closest papers by meaning
- Tracing Moral Foundations in Large Language Models (NeurIPS 2026 · similarity 0.90)
- Language Model Alignment in Multilingual Trolley Problems (ICLR 2025 · similarity 0.88)
- Unveiling the Bias Impact on Symmetric Moral Consistency of Large Language Models (NeurIPS 2024 · similarity 0.87)
- Training-Free Cultural Alignment of Large Language Models via Persona Disagreement (NeurIPS 2026 · similarity 0.87)
- Large Language Models Assume People are More Rational than We Really are (ICLR 2025 · similarity 0.87)
- ProgressGym: Alignment with a Millennium of Moral Progress (NeurIPS 2024 · similarity 0.87)
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.
