The Human-in-the-Loop Interpretability Prior dataset is associated with research by Isaac Lage of Harvard University. It was used to develop an algorithm that minimizes user studies to find models that are both predictive and interpretable. The research demonstrates the approach on several datasets, showing trends towards different proxy notions of interpretability across tasks.
Use Cases
- Comparing the effectiveness of different interpretability proxies based on human subject results mentioned in the description
- Developing algorithms for human-in-the-loop model optimization as described in the research
- Benchmarking model performance on tasks where interpretability is a direct optimization goal
- Studying the relationship between model accuracy and human-assessed interpretability across different datasets
Strengths
- Dataset is associated with research from Harvard University
- Research methodology directly incorporates human subjects to evaluate interpretability
Limitations
- Row count is unknown, which may limit suitability assessment
- Column-level documentation is absent; field semantics must be inferred after download
- Last update date is unknown; freshness unverified
Provenance
- Source
- Harvard University
- Collection Method
- Associated with a research paper on optimizing models for interpretability using human-in-the-loop methods.