Please join us for a UW Data Science Seminar featuring Boris Kovalerchuk from the Department of Computer Science at Central Washington University on Wednesday, May 21st from 4:30 to 5:20 p.m. PT. The seminar will be held in Electrical and Computer Engineering Building 125 – Campus Map.

“How to Identify and Avoid Quasi-Explanation in AI and Machine Learning”

Abstract: Low-quality explanations of predictive models prevent domain experts from trusting and deploying models in high-stakes domains like medicine, even when these black-box models are accurate. Explanations suitable for data scientists to debug models and improve accuracy differ from those required for domain experts to confidently risk deploying models in critical applications. Unfortunately, popular explanation methods often generate quasi-explanations that fail to meet domain experts’ needs. Crucially, these methods bring foreign assumptions to the domain on top of those already brought by black-box models, most of which remain hidden and are difficult for domain experts to assess. To address this, we advocate for explicitly revealing these assumptions, enabling domain experts to critically assess their validity. We will present methods that produce explanations free from such foreign assumptions, supported by open-source software. These include interactive Visual Knowledge Discovery (VKD) methods, grounded in a novel framework for lossless visualization of multidimensional data by General Line Coordinates (GLC). It complements other AI/ML techniques, bridging the gap between technical accuracy and domain-specific interpretability.

Biography: Dr. Boris Kovalerchuk is a professor of Computer Science at Central Washington University. His publication activities include five books published by Springer:”Data Mining in Finance (2000), “Visual and Spatial Analysis” (2005), “Visual Knowledge Discovery and Machine Learning” (2018), “Integrating Artificial Intelligence and Visualization for Visual Knowledge Discovery” (2022), and “Artificial Intelligence and Visualization: Advancing Visual Knowledge Discovery” (2024), chapters in the Data Mining/Machine Learning Handbooks (2006, 2010, 2023) with total over 200 publications. He coined the terms Visual Knowledge Discovery (VKD) and General Line Coordinates (GLC), which link AI and visualization frontiers. Dr. Kovalerchuk has been a principal investigator of research projects supported by the US Government agencies and a senior visiting scholar at the Air Force Research Lab. He delivered relevant tutorials at major conferences.

The 2024-2025 seminars will be held in person, and are free and open to the public.