Evaluating Collaborative Explainable AI System Effectiveness

This document is a research report that evaluates the effectiveness of a Collaborative Explainable AI (CXAI) system in enhancing user understanding and satisfaction. The study investigates whether user-generated explanations improve users' mental models, performance, and satisfaction compared to traditional algorithmic explanations. Through two experiments involving undergraduate students, the report assesses the impact of CXAI on user knowledge and accuracy in answering questions about an AI image classifier system. The first study focused on comprehension and performance, revealing that users interacting with the CXAI system achieved higher accuracy than those using a visual browsing tool without explanations. The second study qualitatively assessed user satisfaction with the CXAI system. Results indicated that collaborative explanations significantly enhanced user understanding and satisfaction, suggesting that CXAI may provide advantages over conventional Explainable AI approaches. The findings contribute to the understanding of user interaction with AI systems and the potential benefits of collaborative explanatory platforms.