Governing the Machine: AI Ethical Frameworks and Governance Boards

Abstract

Artificial intelligence is being incorporated into high-stakes sectors of society such as healthcare, employment, and criminal justice at a rate that outpaces governance structures designed to mitigate the potential risks that come with the novel technology. To that end, despite its widespread usage, only 31% of organizations globally maintain formal AI governance policies, which has paved the way for a multitude of cases on bias concerns, privacy violations, and decreasing public trust. This study explores whether a mandatory, government-centered AI governance framework with Explainable AI (XAI) requirements is associated with stronger public trust outcomes than voluntary, industry-led approaches. Additionally, it aims to identify whether independent governance boards with genuine enforcement authority within organizational settings are associated with improved accountability. A structured narrative synthesis of six peer-reviewed sources was conducted in an effort to compare governance mechanisms and oversight models by drawing on survey-based, experimental, organizational, and theoretical samples. In order to evaluate the results of the studies across different sectors, effect sizes were documented and compared descriptively in their original reported metric, rather than converted to a common metric and statistically pooled. Two summary tables translated these data points into a structured comparison across sources. Across the six sources, XAI/transparency conditions were consistently associated with a small to moderate positive relationship with public trust, while a coercive, compliance-based mechanism examined in one organizational study was associated with the weakest outcome of any mechanism reviewed. Trust transfer dynamics further supported that human intermediaries can meaningfully drive AI trust given that expertise and credibility are maintained. Ultimately, these findings suggest that while Explainable AI requirements are associated with a small to moderate positive relationship with public trust, the maintenance and long-term strength of that trust may depend more on the presence of enforceable governance structures than on transparency alone. Because none of the six sources directly compares mandatory government-centered governance to voluntary industry-led governance, or directly tests the effect of independent board authority, these findings should be interpreted as associative evidence rather than as a direct causal test. These results tie together to inform policy recommendations aimed at proposing Maryland legislation encouraging XAI standards and independent AI governance boards, and also guiding outreach to small AI companies.