ExplainabilityTransparencyAccountabilitycs.AI
Strategies to exploit XAI to improve classification systems
Andrea Apicella, Luca Di Lorenzo, Francesco Isgrò, Andrea Pollastro, Roberto Prevete
June 9, 2023
xAI
18 citations
Key Takeaways
-
Explanations can do more than just clarify decisions: While most XAI research focuses on helping users understand AI decisions, this paper demonstrates that explanations can also be used as a tool to actually improve how well AI systems perform their tasks—suggesting a dual benefit to investing in explainability.
-
Integrated Gradients shows promise for performance improvement: Among the XAI methods tested, Integrated Gradients—a technique that identifies which input features matter most for a decision—was found to effectively highlight features that could be used to enhance classification accuracy across multiple datasets (Fashion-MNIST, CIFAR10, and STL10).
-
There's an underexplored opportunity in XAI research: The paper identifies a gap in the field: most XAI literature focuses on the "how to explain" question rather than the "how to improve" question, suggesting that responsible AI development could benefit from more research on leveraging explanations for system enhancement.
-
Practical strategies exist for using explanations to improve models: The authors report and test two concrete strategies for converting explanations into performance improvements, indicating that the connection between interpretability and model improvement is not just theoretical but actionable.
Abstract
Explainable Artificial Intelligence (XAI) aims to provide insights into the decision-making process of AI models, allowing users to understand their results beyond their decisions. A significant goal of XAI is to improve the performance of AI models by providing explanations for their decision-making processes. However, most XAI literature focuses on how to explain an AI system, while less attention has been given to how XAI methods can be exploited to improve an AI system. In this work, a set of well-known XAI methods typically used with Machine Learning (ML) classification tasks are investigated to verify if they can be exploited, not just to provide explanations but also to improve the performance of the model itself. To this aim, two strategies to use the explanation to improve a classification system are reported and empirically evaluated on three datasets: Fashion-MNIST, CIFAR10, and STL10. Results suggest that explanations built by Integrated Gradients highlight input features that can be effectively used to improve classification performance.