iSee: Advancing Multi-Shot Explainable AI Using Case-based Recommendations

One explanation isn't enough: The research confirms that different users need different types of explanations to understand AI decisions. Even the same person may require multiple explanations depending on the situation, challenging the common practice of relying on a single explanation method.

Customized explanation strategies improve AI systems: The iSee platform demonstrates that AI designers can create tailored "explanation experiences" by combining multiple explanation approaches for specific users or groups, leading to better decision-making outcomes than one-size-fits-all solutions.

Sharing best practices accelerates responsible AI adoption: By using case-based reasoning to capture and reuse successful explanation strategies across different AI applications, the platform helps spread XAI best practices more widely, making it easier for organizations to implement explainable AI responsibly.

AI designers need better tools: The study with six design users across different expertise levels shows that providing AI system designers with practical tools to test and refine explanation strategies is essential for building trustworthy AI systems that actually meet user needs.

Trust and satisfaction depend on personalization: The research emphasizes that user trust in AI-assisted decisions increases when explanations are personalized to their specific needs and background, rather than providing generic explanations to all users.