Speaker
Description
Our research in Human-Centered AI focuses on enabling domain experts to actively guide and interpret machine learning (ML) processes through interactive, knowledge-driven methods. We develop visual analytics (VA) techniques that support expert involvement in both the construction and interpretation of ML models, with the goal of improving transparency, trust, and alignment with human reasoning.
A key strand of our work involves supporting domain experts in transforming raw temporal and spatial data into semantically meaningful, episode-based representations. These enriched representations enable the training of ML models whose behavior can be more easily understood and validated. In parallel, we have explored methods for generating domain-aware synthetic instance neighborhoods to support localized model explanations, ensuring the plausibility and realism of instance-based reasoning.
The core of our current work centers on RuleSense, a visual analytics system designed to support expert-driven exploration of model logic through rule-based representations. RuleSense enables users to analyze and make sense of large rule sets extracted from trained models—such as random forest classifiers and regressors—by providing aggregated views of how feature values relate to predicted outcomes. The system allows experts to investigate whether a model’s internal logic is consistent with domain knowledge and to identify unexpected or potentially flawed decision patterns.
We demonstrate RuleSense across several use cases: from COVID-19 incidence prediction to movement pattern recognition in maritime data. In the life sciences domain, we apply it to a regression model predicting the medical potency of chemical compounds. By encoding extracted rules and applying topic modeling techniques, we help experts discover potentially interesting feature combinations associated with high predicted outcomes—an approach that supports hypothesis generation in early-stage drug discovery.
Together, these efforts contribute to the broader goal of designing AI systems that are not only technically effective but also interpretable, controllable, and meaningful to the human decision-makers who rely on them.
| Spotlight Presenter | Natalia Andrienko |
|---|
