CoAX: Cognitive-Oriented Attribution eXplanation User Model of Human Understanding of AI Explanations
Researchers have developed a new cognitive model called CoAX to better understand how humans interpret explanations from AI systems, particularly when dealing with structured data. Through user studies and cognitive modeling, the team identified effective and ineffective reasoning strategies people use when trying to anticipate AI decisions. This model can help debug human understanding of AI explanations and guide the development of more usable and interpretable AI interfaces. AI
IMPACT Improves understanding of how users interpret AI explanations, potentially leading to more effective AI system design.