Classification with Conceptual Safeguards

Hailey Joren, Charles Thomas Marx, Berk Ustun

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Keywords: classification, selective classification, concept bottleneck models, uncertainty quantification, selective classification

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

TL;DR: We introduce conceptual safeguards, a method for selective classification in concept bottleneck models that can reliably promote safety and interpretability in automation tasks

Abstract: We propose a new approach to promote safety in classification tasks with concept annotations. Our approach – called a *conceptual safeguard* – acts as a verification layer for models that predict a target outcome by first predicting the presence of intermediate concepts. Given this architecture, a safeguard ensures that a model meets a minimal level of accuracy by abstaining from uncertain predictions. In contrast to a standard selective classifier, a safeguard provides an avenue to improve coverage by allowing a human to confirm the presence of uncertain concepts on instances on which it abstains. We develop methods to build safeguards that maximize coverage without compromising safety, namely techniques to propagate the uncertainty in concept predictions and to flag salient concepts for human review. We benchmark our approach on a collection of real-world and synthetic datasets, showing that it can improve performance and coverage in deep learning tasks.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Primary Area: societal considerations including fairness, safety, privacy

Submission Number: 2858