Regulating eXplainable Artificial Intelligence (XAI) May Harm Consumers

Mandatory XAI regulation may not improve outcomes: The paper's game theoretic model shows that requiring firms to provide fully transparent AI explanations could actually make both firms and consumers worse off, challenging the assumption that more transparency always increases social welfare.

Market competition may naturally incentivize explanations: The research suggests that XAI regulation could be redundant because firms in competitive markets may voluntarily provide explanations to build consumer trust and gain competitive advantage without government mandates.

A tradeoff exists between transparency and welfare: Rather than transparency being universally beneficial, the paper identifies a fundamental tension where maximizing social welfare and maximizing access to explainable AI outputs may require different policy approaches.

"XAI fairness" is a distinct and difficult problem: The authors introduce the concept of XAI fairness—ensuring explanations are fair across different groups—and find that this may be impossible to guarantee even when XAI is mandatory, suggesting transparency alone doesn't solve fairness concerns.

One-size-fits-all regulation may be counterproductive: The findings suggest policymakers should carefully consider whether mandatory XAI requirements are the most effective approach, and may need to tailor regulations based on specific market conditions and consumer needs rather than implementing blanket transparency mandates.