Concept bottlenecks change model sensitivity but do not guarantee robustness
Interpretable but Fragile? Robustness of Concept Bottlenecks under Geometric-Semantic Perturbations
Machine LearningArtificial Intelligence
Summary
Some computer models explain their decisions using clear intermediate ideas called concepts. This study looked at whether using these concepts makes models stronger against changes or noise in data. The authors found that while using concepts changes where the model is sensitive, it does not always make the model more reliable. How much the model’s robustness changes depends on the kind of change in data and the task itself. This means clear explanations in a model don’t always mean the model is harder to fool or break.
What this means in practice
- •For machine learning engineers: Design classifiers with interpretable intermediate steps while understanding how sensitivity shifts under different perturbations.
- •For security analysts: Assess vulnerabilities of concept-based models by testing sensitivity to geometric and semantic input changes to improve model safety.
Authors
Hanwei Zhang, Tianma Hu, Gaojie Jin, Xu Cheng, Ronghui Mu
Abstract
Concept Bottleneck Models (CBMs) are designed to provide interpretable intermediate representations, yet how such bottlenecks affect robustness remains unclear, with existing studies reporting mixed and sometimes contradictory findings. We argue that these discrepancies arise from conflating different robustness notions and perturbation regimes, rather than from fundamental disagreements about CBMs themselves. To disentangle these factors, we introduce a generator-based evaluation framework that enables controlled comparisons between standard classifiers and CBMs under two distinct perturbation types: continuous geometric perturbations in latent space and discrete semantic interventions in concept space. Within this framework, we evaluate robustness both empirically, via prediction and concept-level sensitivity metrics, and certifiably, using randomized smoothing in latent and concept spaces. Across experiments, we reconcile previously conflicting findings by clarifying when, and in what sense, concept bottlenecks do or do not improve robustness. By further analyzing robustness under varying task conditions, including class semantic similarity and concept vocabulary size, we show that interpretability does not inherently confer robustness. Instead, concept bottlenecks shift where and how sensitivity manifests, revealing a nuanced interpretability robustness trade off that depends critically on the perturbation regime and task structure. Together, our results show that interpretability and robustness are distinct objectives: interpretable intermediate representations do not uniformly improve robustness, but instead redistribute sensitivity across perturbation spaces and model families.