ExplainabilityTransparencyAccountabilitycs.AI
ShapG: new feature importance method based on the Shapley value
Chi Zhao, Jing Liu, Elena Parilina
June 29, 2024
Engineering applications of artificial intelligence
29 citations
1 influential
Key Takeaways
-
ShapG makes AI explanations faster without sacrificing accuracy: The method reduces computational complexity compared to other Shapley value-based approaches while providing "more accurate explanations" on tested datasets, addressing a key barrier to implementing explainability in real-world AI systems.
-
The method works across different types of AI models: As a "model-agnostic global explanation method," ShapG can explain decisions from various AI systems regardless of their underlying architecture, making it broadly applicable across industries and use cases.
-
Feature relationships improve explanation quality: ShapG uniquely accounts for correlations between features by mapping them as a graph structure, which the authors argue leads to more accurate importance measurements than methods that treat features independently.
-
Transparency becomes more practical for complex systems: By demonstrating "wide range of applicability for explaining complex models" with improved efficiency, ShapG helps address the real-world challenge that many organizations face when trying to make their AI systems transparent to stakeholders and regulators.
Abstract
With wide application of Artificial Intelligence (AI), it has become particularly important to make decisions of AI systems explainable and transparent. In this paper, we proposed a new Explainable Artificial Intelligence (XAI) method called ShapG (Explanations based on Shapley value for Graphs) for measuring feature importance. ShapG is a model-agnostic global explanation method. At the first stage, it defines an undirected graph based on the dataset, where nodes represent features and edges are added based on calculation of correlation coefficients between features. At the second stage, it calculates an approximated Shapley value by sampling the data taking into account this graph structure. The sampling approach of ShapG allows to calculate the importance of features efficiently, i.e. to reduce computational complexity. Comparison of ShapG with other existing XAI methods shows that it provides more accurate explanations for two examined datasets. We also compared other XAI methods developed based on cooperative game theory with ShapG in running time, and the results show that ShapG exhibits obvious advantages in its running time, which further proves efficiency of ShapG. In addition, extensive experiments demonstrate a wide range of applicability of the ShapG method for explaining complex models. We find ShapG an important tool in improving explainability and transparency of AI systems and believe it can be widely used in various fields.