Shapelet-Based Counterfactual Explanations for Multivariate Time Series

Counterfactual explanations make AI decisions actionable: Rather than just explaining why a model made a decision, this method shows what specific changes would need to happen to get a different outcome—for example, in solar flare prediction, it could show which measurements would need to shift to change a "flare predicted" decision to "no flare predicted."

Time series data has been largely neglected in explainability research: The paper identifies a significant gap in XAI methods, noting that "fewer methods have been developed for time series data, and even less when it comes to multivariate datasets," which is problematic since many real-world applications (medical monitoring, financial forecasting, climate prediction) rely on time-based data.

Shapelets offer a human-interpretable approach to complex models: By using shapelets—distinctive patterns within time series data—the method produces explanations that are "visually interpretable," making them more accessible to non-technical stakeholders who need to understand and trust AI decisions in high-stakes domains.

This approach works across different AI models: The method is "model agnostic," meaning it can explain decisions from various types of black-box machine learning and deep learning systems without requiring access to their internal workings, increasing its practical applicability across industries.

Quality matters in explanations: The paper demonstrates that explanations should be evaluated on multiple dimensions—proximity (how realistic the changes are), sparsity (how few changes are needed), and plausibility (how likely the scenario is)—not just whether they technically work.