Designing an Interactive Interface for FACET: Personalized Explanations in XAI

Dion, Katharine; Pietrick, Alexander; Reiss, Jacob; Huang, Randy; Genin, Belisha

Student Work

Designing an Interactive Interface for FACET: Personalized Explanations in XAI

Pubblico Deposited

Recent years have witnessed a surge in the integration of machine learning into critical decision-making processes across various industries such as healthcare and recruitment, resulting in a growing reliance on automated systems. In response to this trend, explainable AI (XAI) methods have gained traction, aiming to enhance transparency and fairness by elucidating AI-generated decisions. One such XAI technique, counterfactual explanations, addresses queries through hypothetical scenarios — yet, current approaches often overlook user preferences and fail to produce actionable solutions. To overcome these limitations, a novel framework called the counterfactual region was developed, exemplified by FACET (Fast Actionable Counterfactuals for Ensembles of Trees). FACET offers personalized and robust counterfactual explanations, ensuring explanation robustness and providing users with actionable solutions tailored to their specific circumstances. To enhance user experience, we prototyped and developed an interactive application as a user interface for the FACET application, empowering users to adjust and prioritize constraints. We created and executed a comprehensive testing plan to ensure the effectiveness of FACET across different scenarios and datasets.

This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

Subject