AI2030 Responsible AI Tools Marketplace

Key Takeaways

Making AI decisions understandable for time series data: The paper addresses a practical challenge in explainable AI by adapting explanation techniques to time series (data that changes over time), which is typically harder for humans to interpret than images or text. This is important because time series data is used in critical applications like medical monitoring and financial forecasting.
Two-level explanation approach improves transparency: The research proposes viewing AI explanations at both global (how the model generally behaves) and local (why it made a specific decision) levels. Users can seamlessly move between these perspectives to understand not just individual predictions but also broader model patterns and potential biases.
"What-if" analysis enables hypothesis testing: The system allows users to ask counterfactual questions—essentially "what would happen if we changed this part of the data?"—to verify whether the model's reasoning makes sense and to catch potential errors in both the model and the underlying data.
Expert feedback shaped the design: The researchers continuously incorporated feedback from domain experts throughout development, resulting in a system that integrates time series transformations directly into explanations rather than treating them separately. This user-centered approach suggests that effective XAI tools require collaboration between AI researchers and practitioners who understand the real-world context.
Practical validation across multiple use cases: The paper demonstrates the workflow successfully identifies feature relevance, detects model behavior issues, and uncovers data problems—suggesting this approach has concrete value for practitioners trying to audit and improve AI systems in practice.

Abstract

With the rising necessity of explainable artificial intelligence (XAI), we see an increase in task-dependent XAI methods on varying abstraction levels. XAI techniques on a global level explain model behavior and on a local level explain sample predictions. We propose a visual analytics workflow to support seamless transitions between global and local explanations, focusing on attributions and counterfactuals on time series classification. In particular, we adapt local XAI techniques (attributions) that are developed for traditional datasets (images, text) to analyze time series classification, a data type that is typically less intelligible to humans. To generate a global overview, we apply local attribution methods to the data, creating explanations for the whole dataset. These explanations are projected onto two dimensions, depicting model behavior trends, strategies, and decision boundaries. To further inspect the model decision-making as well as potential data errors, a what-if analysis facilitates hypothesis generation and verification on both the global and local levels. We constantly collected and incorporated expert user feedback, as well as insights based on their domain knowledge, resulting in a tailored analysis workflow and system that tightly integrates time series transformations into explanations. Lastly, we present three use cases, verifying that our technique enables users to (1)~explore data transformations and feature relevance, (2)~identify model behavior and decision boundaries, as well as, (3)~the reason for misclassifications.

14 pages, 6 figures, submitted to IEEE TVCG in December

Key Takeaways

Making AI decisions understandable for time series data: The paper addresses a practical challenge in explainable AI by adapting explanation techniques to time series (data that changes over time), which is typically harder for humans to interpret than images or text. This is important because time series data is used in critical applications like medical monitoring and financial forecasting.
Two-level explanation approach improves transparency: The research proposes viewing AI explanations at both global (how the model generally behaves) and local (why it made a specific decision) levels. Users can seamlessly move between these perspectives to understand not just individual predictions but also broader model patterns and potential biases.
"What-if" analysis enables hypothesis testing: The system allows users to ask counterfactual questions—essentially "what would happen if we changed this part of the data?"—to verify whether the model's reasoning makes sense and to catch potential errors in both the model and the underlying data.
Expert feedback shaped the design: The researchers continuously incorporated feedback from domain experts throughout development, resulting in a system that integrates time series transformations directly into explanations rather than treating them separately. This user-centered approach suggests that effective XAI tools require collaboration between AI researchers and practitioners who understand the real-world context.
Practical validation across multiple use cases: The paper demonstrates the workflow successfully identifies feature relevance, detects model behavior issues, and uncovers data problems—suggesting this approach has concrete value for practitioners trying to audit and improve AI systems in practice.

Abstract

14 pages, 6 figures, submitted to IEEE TVCG in December

Visual Explanations with Attributions and Counterfactuals on Time Series Classification

Key Takeaways

Abstract

Paper Info

Responsible AI Topics

Links

Visual Explanations with Attributions and Counterfactuals on Time Series Classification

Key Takeaways

Abstract

Paper Info

Responsible AI Topics

Links