Exploring Commonalities in Explanation Frameworks: A Multi-Domain Survey Analysis

People prioritize understanding over perfect accuracy: Across all three domains studied (medical, retail, and energy), both experts and non-experts consistently preferred AI systems that were easier to explain, even if it meant accepting slightly lower accuracy levels. This suggests that transparency and trustworthiness matter more to users than squeezing out marginal performance gains.

Feature importance and counterfactual explanations are universal needs: The research identified two explanation methods as critical across different industries—understanding which factors drove a decision (feature importance) and knowing what would need to change to get a different outcome (counterfactual explanations). These appear to be fundamental to how people want AI systems explained to them.

Domain expertise shapes explanation requirements: The study involved professionals from medical, retail, and energy sectors, revealing that while some explanation needs are universal, different industries have distinct priorities. This suggests that one-size-fits-all AI explanations may not work, and frameworks should be adaptable to specific professional contexts.

Both specialists and general audiences need clarity: The research included questionnaires from both experts and non-experts, finding common ground in explanation preferences. This indicates that responsible AI design should focus on making systems understandable to diverse audiences, not just technical specialists.

Interpretable algorithms deserve more attention: The study used genetic programming algorithms specifically because of their built-in interpretability, suggesting that choosing inherently explainable AI methods—rather than only trying to explain "black box" systems after the fact—is a practical path toward responsible AI.