University of Cambridge > > Artificial Intelligence Research Group Talks (Computer Laboratory) > Concept Embedding Models: Beyond the Accuracy-Explainability Trade-off

Concept Embedding Models: Beyond the Accuracy-Explainability Trade-off

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Mateja Jamnik.

Join us in Lecture Theatre 2 or on Zoom

Deploying AI-powered systems requires trustworthy models supporting effective human interactions, going beyond raw prediction accuracy. Concept bottleneck models promote trustworthiness by conditioning classification tasks on an intermediate level of human-like concepts. This enables human interventions which can correct mispredicted concepts to improve the model’s performance. However, existing concept bottleneck models are unable to find optimal compromises between high task accuracy, robust concept-based explanations, and effective interventions on concepts—particularly in real-world conditions where complete and accurate concept annotations are scarce. In this talk I will describe Concept Embedding Models, a novel family of concept bottleneck models which goes beyond the current accuracy-vs-interpretability trade-off by learning interpretable high-dimensional concept representations. Our experiments demonstrate that Concept Embedding Models (a) attain better or competitive task accuracy w.r.t. standard neural models without concepts, (b) provide concept representations capturing meaningful semantics including and beyond their ground truth labels, (c) support test-time concept interventions whose effect in test accuracy surpasses that in standard concept bottleneck models, and (d) scale to real-world conditions where complete concept supervisions are scarce.

This talk is part of the Artificial Intelligence Research Group Talks (Computer Laboratory) series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.


© 2006-2024, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity