Assessing Reliability of Symbol Detection in Concept Bottleneck Models

2026-06-15

AI Summary

Although Concept Bottleneck Models (CBMs) offer interpretability, their symbolic representations are prone to task shortcuts, leading to distorted explanations. This work presents the first systematic evaluation of symbol reliability in CBMs by swapping independently trained concept detectors and classification heads, and assessing performance degradation, concept-level metrics, and symbol uncertainty in an integrated framework. To mitigate reliance on unreliable symbols, the authors introduce a multi-head shared detector architecture coupled with a reliability-aware training mechanism that penalizes dependence on untrustworthy concepts. Experiments demonstrate near-lossless performance upon swapping on CUB-200-2011, and in weakly supervised synthetic tasks, the proposed approach nearly doubles swap accuracy while significantly suppressing symbol leakage.

Abstract

Concept Bottleneck Models (CBMs) are a relevant tool for explainable Artificial Intelligence because they make their predictions through human-interpretable symbols. However, high task accuracy does not guarantee that these symbols are detected faithfully: jointly trained CBMs may encode task-specific shortcuts in the bottleneck, making their explanations unreliable. In this paper, we study concept-detection reliability by swapping independently trained concept detectors and classification heads that share the same symbolic vocabulary. We use the resulting performance degradation, concept-level metrics, and symbol-wise uncertainty estimates to identify concepts that are especially prone to spurious firing. Finally, we propose a reliability-aware training strategy in which a shared concept detector is optimized with multiple classification heads and penalized for relying on globally or instance-wise unreliable symbols. On CUB-200-2011 with full concept supervision, detectors and heads are almost freely interchangeable (swap drop below one accuracy point, relative retention above $99\%$, and no concept detected below chance), whereas on a controlled synthetic task we show that, as the concept-supervision weight is reduced, models keep near-perfect task accuracy while swapped accuracy and agreement with the ground-truth concepts collapse to chance. Our reliability-aware training substantially mitigates this leakage, roughly doubling swap accuracy in the leaky regime.

Problem

• Concept Bottleneck Models
• symbol detection reliability
• explainable AI
• concept supervision
• spurious correlations

Innovation

• Concept Bottleneck Models
• Reliability Assessment
• Symbol Detection
• Explainable AI
• Reliability-aware Training