Issue Network Dynamics in AI Policymaking - An ERGM Analysis Based on the 2025 AI Action Plan Data

As artificial intelligence (AI) becomes increasingly central to public policy, traditional frameworks such as the Iron Triangle no longer suffice to explain decision making in this rapidly evolving and complex domain. This study examines how the issue network framework provides a more accurate and adaptable lens for understanding contemporary AI governance. Drawing on data from the 2025 U.S. AI Action Plan, including 5,166 policy recommendations submitted by 760 organizations, we apply an Exponential Random Graph Model (ERGM) to analyze patterns of overlapping recommendations across twenty distinct policy domains. Our analysis reveals that most organizations tend to collaborate across stakeholder types rather than within their own, indicating that shared concern over specific issues, rather than institutional similarity, is the primary driver of engagement. Notably, industry associations are an exception, exhibiting strong internal alignment. These findings offer empirical support for the issue network perspective and highlight the significance of inclusive and cross sectoral collaboration in shaping effective AI policy.