English

Auditing for Human Expertise

Machine Learning 2024-11-26 v3 Computers and Society Machine Learning

Abstract

High-stakes prediction tasks (e.g., patient diagnosis) are often handled by trained human experts. A common source of concern about automation in these settings is that experts may exercise intuition that is difficult to model and/or have access to information (e.g., conversations with a patient) that is simply unavailable to a would-be algorithm. This raises a natural question whether human experts add value which could not be captured by an algorithmic predictor. We develop a statistical framework under which we can pose this question as a natural hypothesis test. Indeed, as our framework highlights, detecting human expertise is more subtle than simply comparing the accuracy of expert predictions to those made by a particular learning algorithm. Instead, we propose a simple procedure which tests whether expert predictions are statistically independent from the outcomes of interest after conditioning on the available inputs (`features'). A rejection of our test thus suggests that human experts may add value to any algorithm trained on the available data, and has direct implications for whether human-AI `complementarity' is achievable in a given prediction task. We highlight the utility of our procedure using admissions data collected from the emergency department of a large academic hospital system, where we show that physicians' admit/discharge decisions for patients with acute gastrointestinal bleeding (AGIB) appear to be incorporating information that is not available to a standard algorithmic screening tool. This is despite the fact that the screening tool is arguably more accurate than physicians' discretionary decisions, highlighting that -- even absent normative concerns about accountability or interpretability -- accuracy is insufficient to justify algorithmic automation.

Keywords

Cite

@article{arxiv.2306.01646,
  title  = {Auditing for Human Expertise},
  author = {Rohan Alur and Loren Laine and Darrick K. Li and Manish Raghavan and Devavrat Shah and Dennis Shung},
  journal= {arXiv preprint arXiv:2306.01646},
  year   = {2024}
}

Comments

30 pages, 10 figures. Appeared in the proceedings of the 37th Conference on Neural Information Processing Systems (NeurIPS 2023). 11/2024 replacement fixes typo in the definition of $\tau_k$, as pointed out by Liuquan Nie

R2 v1 2026-06-28T10:54:44.610Z