Research map / Learning theory and algorithms / Prediction and decision losses
Human-in-the-Loop Interpretability Prior
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 |
|---|---|---|---|
| Use human response-time feedback in model-based optimization rather than selecting models by an automatic interpretability proxy | Decision-tree models on Census and Mushroom · Census and Mushroom | mean response time | Helped |
| Use human response-time feedback to select models | Decision-tree models on Mushroom · Mushroom | proxy scores (number of nodes and mean path length) | Helped |
| Use human response-time feedback to select models | Decision-tree models on Census · Census | proxy score (number of nonzero features) | Helped |
| Optimize an automatic interpretability proxy that differs from the human-relevant measure | Candidate models across the evaluated datasets · Synthetic, Mushroom, Census, and Covertype | rank of the human-preferred model under the optimized proxy | Hurt |
| Compute the human-relevant proxy on a small sample of data points instead of accurately optimizing a mismatched proxy | Candidate models across the evaluated datasets · Synthetic, Mushroom, Census, and Covertype | rank of the human-preferred model | Helped |
| Use model-based optimization to choose models for evaluation rather than randomly sampling the same number of models | Candidate models across the evaluated datasets and proxies · Synthetic, Mushroom, Census, and Covertype | rank of the best model found | Helped |
| Compare a medium-mean-path-length model with a high-mean-path-length model in a between-subjects pilot | Pipeline-selected models in the Amazon Mechanical Turk pilot · Amazon Mechanical Turk pilot | mean response time: 31.62 -> 30.94 | No clear effect |
Closest papers by meaning
- Intrinsic User-Centric Interpretability through Global Mixture of Experts (ICLR 2025 · similarity 0.86)
- Local vs. Global Interpretability: A Computational Complexity Perspective (ICML 2024 · similarity 0.85)
- Measuring Per-Unit Interpretability at Scale Without Humans (NeurIPS 2024 · similarity 0.85)
- Enhancing Performance of Explainable AI Models with Constrained Concept Refinement (ICML 2025 · similarity 0.85)
- Enhancing Pre-trained Representation Classifiability can Boost its Interpretability (ICLR 2025 · similarity 0.84)
- Conformal Prediction Sets Improve Human Decision Making (ICML 2024 · 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.