User Characteristics in Explainable AI: The Rabbit Hole of Personalization?

Most user characteristics don't significantly impact XAI effectiveness: In a study of 149 participants using an AI system that flagged inappropriate comments, researchers found that very few user characteristics actually mattered for how people interacted with and experienced AI explanations, suggesting the field may be overcomplicating personalization efforts.

Age and openness are the primary user factors that matter: Only two factors meaningfully influenced whether users actually understood the AI system—age and the personality trait of openness—indicating that XAI designers should focus on these characteristics rather than attempting to personalize for numerous demographic and psychological variables.

The pursuit of hyper-personalized XAI may be misguided: The paper questions whether the current research focus on tailoring explanations to fine-grained user characteristics represents a productive direction, suggesting the field should reconsider its approach to personalization in explainable AI systems.

Understanding and trust require different design approaches than previously assumed: The findings challenge assumptions that diverse users (data scientists, domain experts, end-users) need fundamentally different explanation strategies, potentially simplifying how organizations should design AI transparency features.