Research map / Large language models / AI and society
Auditing for Human Expertise
Rohan Alur, Loren Laine, Darrick K Li, Manish Raghavan, Devavrat Shah, Dennis Shung
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 GBS threshold > 7 instead of physician discretion | AGIB triage decisions · AGIB | accuracy: 0.55 ± 0.02 -> 0.79 ± 0.02 | Improved |
| Use GBS threshold > 7 instead of physician discretion | AGIB triage decisions · AGIB | sensitivity: 0.99 ± 0.00 -> 0.73 ± 0.02 | Hurt |
| Use GBS threshold > 7 instead of physician discretion | AGIB triage decisions · AGIB | specificity: 0.24 ± 0.02 -> 0.84 ± 0.02 | Improved |
| Run ExpertTest with L=500 matched pairs | Physician hospitalization decisions · AGIB | p-value (τ): 0.061 -> <.001 | Improved |
| Increase matched-pair count from L=500 to L=1000 | ExpertTest · AGIB | p-value (τ): <.001 -> <.001 | No clear effect |
| Use physician decisions as signals in addition to the nine GBS features | AGIB risk prediction · AGIB | conditional information about outcome | Improved |
Closest papers by meaning
- Human Expertise in Algorithmic Prediction (NeurIPS 2024 · similarity 0.88)
- Defining Expertise: Applications to Treatment Effect Estimation (ICLR 2024 · similarity 0.85)
- Bayesian Inference for Correlated Human Experts and Classifiers (ICML 2025 · similarity 0.83)
- Trustworthy and Practical AI for Healthcare: A Guided Deferral System with Large Language Models (AAAI 2025 · similarity 0.82)
- Doctor-R1: Mastering Clinical Inquiry with Experiential Agentic Reinforcement Learning (ICLR 2026 · similarity 0.81)
- Unlocking the Potential of Global Human Expertise (NeurIPS 2024 · similarity 0.81)
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.
