Research map / Bandits and online learning / Multi-armed bandits
Why so gloomy? A Bayesian explanation of human pessimism bias in the multi-armed bandit task
Dalin Guo, Angela J. Yu
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 |
|---|---|---|---|
| use dynamic belief model instead of fixed belief model | learning model · Human bandit choices | predictive accuracy: 0.3856 -> 0.4182 | Improved |
| use dynamic belief model instead of RL | learning model · Human bandit choices | predictive accuracy: 0.4038 -> 0.4182 | Improved |
| assume environment volatility (gamma<1) rather than stationarity | DBM vs FBM · Prior mean recovery | difference of recovered prior mean across environments: p = 0.4350 | Improved |
| vary environment abundance and variance | human participants · Reported reward estimates | average reported estimate: p = 0.1570 | No clear effect |
| high versus low abundance environment | DBM · Fitted parameters | softmax b and gamma: p < .01 | Mixed |
| high versus low variance environment | DBM · Fitted parameters | softmax b and gamma: p > .05 | No clear effect |
| fit parameters per participant and environment instead of shared (ablation) | DBM/FBM/RL fitting · 10-fold cross validation | held-out predictive accuracy | Improved |
| underestimate prior mean in FBM with softmax policy | FBM · Simulated high variance environments | earned reward per trial | Improved |
| underestimate prior mean even more in DBM | DBM · Simulated high variance environments | earned reward per trial | Improved |
Closest papers by meaning
- InfoRM: Mitigating Reward Hacking in RLHF via Information-Theoretic Reward Modeling (NeurIPS 2024 · similarity 0.86)
- Model Alignment as Prospect Theoretic Optimization (ICML 2024 · similarity 0.85)
- Post-hoc Reward Calibration: A Case Study on Length Bias (ICLR 2025 · similarity 0.85)
- A Unifying Normative Framework of Decision Confidence (NeurIPS 2024 · similarity 0.85)
- Hindsight PRIORs for Reward Learning from Human Preferences (ICLR 2024 · similarity 0.85)
- Robustly Improving Bandit Algorithms with Confounded and Selection Biased Offline Data: A Causal Approach (AAAI 2024 · similarity 0.85)
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.
