On Two XAI Cultures: A Case Study of Non-technical Explanations in Deployed AI System

The "explanation gap" is a critical problem in deployed AI systems: The paper identifies a significant disconnect between what AI experts consider "explained" versus what non-technical stakeholders actually understand. This gap means many deployed AI systems lack meaningful explanations for the people most affected by them, despite extensive XAI research.

Non-technical audiences are the primary stakeholders, not AI experts: The paper emphasizes that non-technical people—including end-users, regulators, and affected communities—are the major stakeholders in real-world AI deployments, yet XAI research has primarily focused on making AI explainable to technical audiences, creating a mismatch in priorities.

Two distinct "cultures" of XAI exist in practice: The research reveals that AI experts and non-technical stakeholders have fundamentally different expectations, goals, and definitions of what constitutes a good explanation. Understanding these cultural differences is essential for responsible AI deployment.

Successful AI deployment in regulated industries requires non-technical explanations: The case study demonstrates that a highly regulated industry achieved successful AI deployment by prioritizing clear, non-technical explanations to stakeholders, suggesting this approach is both feasible and necessary for responsible AI systems.

AI experts need practical guidance for communicating with non-technical audiences: The paper provides actionable suggestions for AI professionals to improve how they explain AI decisions to non-technical stakeholders, recognizing this as a learnable skill critical to responsible AI implementation.