To Ask or Not to Ask: Learning to Require Human Feedback
This addresses the problem of enhancing decision-support systems for classification tasks by enabling more effective collaboration between humans and AI, though it is incremental as it builds on existing LtD methods.
The paper tackles the limitation of Learning to Defer (LtD) by proposing Learning to Ask (LtA), a framework that determines when and how to incorporate expert human feedback into ML models, resulting in a more flexible and powerful approach for human-AI collaboration as demonstrated in experiments with synthetic and real expert data.
Developing decision-support systems that complement human performance in classification tasks remains an open challenge. A popular approach, Learning to Defer (LtD), allows a Machine Learning (ML) model to pass difficult cases to a human expert. However, LtD treats humans and ML models as mutually exclusive decision-makers, restricting the expert contribution to mere predictions. To address this limitation, we propose Learning to Ask (LtA), a new framework that handles both when and how to incorporate expert input in an ML model. LtA is based on a two-part architecture: a standard ML model and an enriched model trained with additional expert human feedback, with a formally optimal strategy for selecting when to query the enriched model. We provide two practical implementations of LtA: a sequential approach, which trains the models in stages, and a joint approach, which optimises them simultaneously. For the latter, we design surrogate losses with realisable-consistency guarantees. Our experiments with synthetic and real expert data demonstrate that LtA provides a more flexible and powerful foundation for effective human-AI collaboration.