Risk-Oriented Differentiated Governance for Ambiguous Information

The rapid advancement of generative artificial intelligence has led to an exponential increase in false information, especially ambiguous information. Hampered by resource constraints, conventional governance approaches fall short of enabling efficient and precise risk management. In response, this paper innovatively proposes a risk-oriented tiered governance framework tailored to ambiguous information. Specifically, we quantify epistemic uncertainty via multi-level semantic features, combine it with network propagation features to determine latent influence scores, and construct a risk function capturing their synergy. Critical and acceptable boundaries are introduced to categorize ambiguous information into discrete risk levels, tier-specific governance strategies are devised, and a propagation throttling (limiting each user's successful forwards per ambiguous information) strategy is innovatively proposed for medium-risk information. Finally, we simulate differentiated governance strategies across risk levels of ambiguous information, systematically assess governance effectiveness, and offer a comparative discussion grounded in fairness and ethics. This study offers an operational analytical framework for refined risk governance of ambiguous information, contributes methodological innovations in risk quantification modeling and tiered governance strategies, and provides theoretical support and practical guidance for risk management authorities in designing differentiated measures.