Research map / Large language models / AI and society
Moral Change or Noise? On Problems of Aligning AI with Temporally Unstable Human Feedback
Vijay Keswani, Cyrus Cousins, Breanna Nguyen, Vincent Conitzer, Hoda Heidari, Jana Schaich Borg, Walter Sinnott-Armstrong
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 |
|---|---|---|---|
| scenarios with more feature tradeoffs (U to V/W) | participant responses · Repeated kidney scenarios | response stability: H(5)=222.23, p<0.001 | Hurt |
| higher model entropy (using more features) | participant responses · Repeated kidney scenarios | response stability: -0.117 | Hurt |
| longer reaction time | participant responses · Repeated kidney scenarios | response stability: -0.081 | Hurt |
| longer time gap between sessions | session-wise logistic models · Kidney allocation | model stability: rho = -0.12 | Hurt |
| participants being response and model unstable (C4 vs C1) | BT-NN · Kidney allocation held-out choices | error rate: +0.16 | Hurt |
| participants being response and model unstable (C4 vs C1) | GPT-FT · Kidney allocation held-out choices | error rate: +0.05 | Hurt |
| more queries since training | BT-NN for C4 participants · Kidney allocation | error rate: significant positive slope in 3 of 4 training sessions | Hurt |
| population-level fine-tuning (GPT-FT) instead of participant-specific models | preference model · Kidney allocation | error rate | Hurt |
Closest papers by meaning
- Position: Towards Bidirectional Human-AI Alignment (NeurIPS 2025 · similarity 0.89)
- ProgressGym: Alignment with a Millennium of Moral Progress (NeurIPS 2024 · similarity 0.89)
- AI Alignment with Changing and Influenceable Reward Functions (ICML 2024 · similarity 0.88)
- Strong Preferences Affect the Robustness of Preference Models and Value Alignment (ICLR 2025 · similarity 0.86)
- Language Model Alignment in Multilingual Trolley Problems (ICLR 2025 · similarity 0.86)
- Quantifying Misalignment Between Agents: Towards a Sociotechnical Understanding of Alignment (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.
