Explaining Explainability: Recommendations for Effective Use of Concept Activation Vectors
Concept-based explanations translate the internal representations of deep learning models into a language that humans are familiar with: concepts. One popular method for finding concepts is Concept Activation Vectors (CAVs), which are learnt using a probe dataset of concept exemplars. In this work, we investigate three properties of CAVs: (1) inconsistency across layers, (2) entanglement with other concepts, and (3) spatial dependency. Each property provides both challenges and opportunities in interpreting models. We introduce tools designed to detect the presence of these properties, provide insight into how each property can lead to misleading explanations, and provide recommendations to mitigate their impact. To demonstrate practical applications, we apply our recommendations to a melanoma classification task, showing how entanglement can lead to uninterpretable results and that the choice of negative probe set can have a substantial impact on the meaning of a CAV. Further, we show that understanding these properties can be used to our advantage. For example, we introduce spatially dependent CAVs to test if a model is translation invariant with respect to a specific concept and class. Our experiments are performed on natural images (ImageNet), skin lesions (ISIC 2019), and a new synthetic dataset, Elements. Elements is designed to capture a known ground truth relationship between concepts and classes. We release this dataset to facilitate further research in understanding and evaluating interpretability methods.
Code (1)
Methods 이 논문이 사용한 방법론
Similar Papers 제목 키워드 기반
The emergence of Explainability of Intelligent Systems: Delivering Explainable and Personalised Recommendations for Energy Efficiency
The recent advances in artificial intelligence namely in machine learning and deep learning, have boosted the performance of intelligent systems in several ways. This gave rise to human expectations, but also created the…
Decision MakingRecommendation SystemsExplaining Image Similarity with Automatically Extracted Concept Activation Vectors
Image similarity underlies many computer vision applications, yet it is often unclear why two images receive a high or low similarity score. Existing explainability methods often rely on gradient-based attribution maps t…
Leveraging counterfactual concepts for debugging and improving CNN model performance
Counterfactual explanation methods have recently received significant attention for explaining CNN-based image classifiers due to their ability to provide easily understandable explanations that align more closely with h…
counterfactualCounterfactual ExplanationCounterfactual Reasoningimage-classification+1Context-aware explainable recommendations over knowledge graphs
Knowledge graphs contain rich semantic relationships related to items and incorporating such semantic relationships into recommender systems helps to explore the latent connections of items, thus improving the accuracy o…
Knowledge GraphsRecommendation SystemsFastCAV: Efficient Computation of Concept Activation Vectors for Explaining Deep Neural Networks
Concepts such as objects, patterns, and shapes are how humans understand the world. Building on this intuition, concept-based explainability methods aim to study representations learned by deep neural networks in relatio…