RACER improves how AI knows when to ask humans for help
RACER: Role-Aligned Competence Estimation for Human-AI Routing
Machine Learning
Summary
Many AI systems decide whether to handle a task alone or ask a human expert for help. The authors focus on improving this decision when new, unseen experts are involved, using a small sample of their behavior. They introduce RACER, a new way to estimate how competent an unseen expert is for each kind of task without relying on fixed labels. This approach better adapts to experts’ strengths in specific cases and works well on medical images and benchmark datasets.
What this means in practice
- •For medical imaging teams: Improve AI systems that decide when to ask radiologists for help based on observed expert performance on specific tasks.
- •For automated customer support providers: Develop systems that better route queries to human agents based on role-relative competence learned from limited prior interactions.
Authors
Joshua Strong, Emma Sun, Alexander Capstick, Pramit Saha, Cheng Ouyang, J. Alison Noble
Abstract
Learning to defer asks a predictive system when to act autonomously and when to defer to a human expert. Population-adaptive deferral extends this problem to unseen experts using a small context set of expert behavior. Neural context encoders such as L2D-Pop can be query-dependent, but may learn routing shortcuts tied to absolute class coordinates. Identity-Free Deferral (IFD) removes such shortcuts through role-indexed classwise competence profiles, but its estimates are constant within each class and cannot capture instance-level expert specialization. We propose RACER---Role-Aligned Competence Estimation for Routing---a role-relative framework for estimating an unseen expert's competence from context. RACER estimates the posterior-predictive probability that the expert is correct on a query under each candidate class role, then combines these estimates with the model posterior to obtain the Bayes-relevant expert-correctness probability. Nonparametric and neural kernel-pooling estimators use candidate-role relations, shared aggregation, and symmetric summaries, excluding absolute class-identity channels. We prove coherent class-relabelling invariance, derive a Bayes-aligned deferral surrogate, and give a plug-in regret bound relating routing regret to classifier and competence-estimation error. On controlled synthetic benchmarks, including a PathMNIST histopathology context-scaling study with simulated experts, RACER benefits from additional context under hidden subtype dependence and gives the strongest aggregate performance on a separately sampled unseen-expert split in the CIFAR-100 synthetic experiments. On the radiologist and human--AI chest-radiography benchmarks (VinDr-CXR and CheXpert), the RACER family is competitive or best in budget-swept deferral, with calibration results varying across metrics and datasets.