“Tongyi Qianwen,” an AI large language model launched by Alibaba on April 11 Photo: CFP

---

On March 25, Shandong University held a high-end forum on artificial intelligence and the modernization of national governance. Scholars from various disciplines, such as political science, public administration, and sociology, discussed the impact of AI on modernizing national governance, the role of AI in governance, and regulating AI for healthy development. They explored ways to promote the combination of “intelligence” and “governance” for better development in the new era.

Governance function

“The new development and practices of AI technology have profoundly influenced the modern governance system and mechanism,” said Ma Ben, associate dean of the School of Political Science and Public Administration at Shandong University. AI, as an information technology, can eliminate the influence of human and structural irrationality in decision-making and information transmission. For example, in terms of the management process, AI technology can replace the brainwork of content creation by formulating more precise and feasible plans and budgets based on large amounts of data and advanced predictive models. In addition, it can enhance managers’ information processing ability, supervise information collection and verification, and avoid corruption and collusion. 

AI displays a more prominent advantage in the governance of systemic risk events. “Addressing and defusing systemic risks is not an emergency management issue, but a holistic governance issue that demands the participation of multiple actors,” said Wei Jiuchang, executive dean of the School of Public Affairs at the University of Science and Technology of China. The key to applying AI technology to policy domains is to seek effective indicators to predict and identify risky situations, evaluate policy application scenarios, and locate problems in existing policies. AI technology can delineate the structural traits of policies, unfold risk problems in the way of seeking optimization and solutions, and demonstrate the evolutionary trends of policy application, providing intelligent policy instruments. AI technology serves as an effective tool to obtain risk information and identify the evolutionary tendency of policies, especially for decision makers not directly involved in systemic risks. An effective means to defuse risk is collaborative governance, and AI technology has the capability to identify the window of collaborative governance and confirm the number of actors to be coordinated through algorithms, establishing combinations of policy collaborative optimization.

Intelligent governance

 “The rapid development and wide application of AI makes intelligent governance a hot topic,” said Meng Tianguang, deputy dean of the School of Social Sciences at Tsinghua University. Intelligent governance chiefly covers AI-based and AI-oriented governance. Its general framework is composed of at least three basic elements: actors, resources, and mechanisms. Governance actors include the government, science and technology enterprises, and society; governance resources mainly refer to big data, computing power, and algorithms; and governance mechanisms consist of co-building, co-governance, and sharing. In the framework of intelligent governance, the predominant government leads the integration of AI and social governance, science and technology enterprises provide technical support, and social actors actively engage in and supervise the holistic process of governance. In the actual governance process, the three parties form a grid cooperation pattern, contributing governance wisdom and sharing governance fruits.

 “Social simulation can provide effective ideas and routes for the construction of intelligent social governance,” said Lyu Peng, a professor from the School of Public Administration at Central South University. The key to social simulation is to deal with the issue of “emergence.” At the stage of agent-based modeling and simulation, the core of emergence relates to the structures among agents. Theoretically, individual practices based on social common knowledge can be translated as a logical mode of social portrayal, social knowledge, simulation and deduction, social prediction, social intervention, and social optimization.

Strengthening regulations 

 “As booming AI is applied into multiple scenarios, it also raises ethical and governance challenges,” noted Xue Lan, dean of the Schwarzman College at Tsinghua University. Ethically, the governance bottleneck of AI lies in technics-out-of-control, misuse and abuse, the issue of AI as an emerging actor, as well as consequent short-term and long-term social impacts. In the long run, the governance of AI should move towards an agile governance model featuring multi-actor cooperation, calling for a trust governance relationship among the government, industries, and enterprises.

Large models can reduce systematic deviation, avoid overfitting of models, and make up for the deficiencies of small models in accurately depicting social systems. However, the computational reliability of judicial data under large models is controversial. According to Wang Fang, deputy director of the Data Science Institute at Shandong University, the advantage of frontier AI represented by ChatGPT is primarily reflected in the prediction ability, largely demonstrating the statistical correlation between variables rather than their causal relationship. The reliability of operational models, algorithms, data, and conclusions needs to be verified. In addition, large models encounter the challenges of particularity. 

When talking about the inability of judicial big data to accurately forecast hard cases with limited data and the insufficient integration of legal and computational logic, Wang suggested integrating the prediction ability of large models and the explainability of small models, combining the system identification algorithm and the thought of causal inference, and applying the non-linear recursive identification method to data analysis, so as to realize rational and scientific modeling.  

Regarding the ownership of AI copyrights, it is advisable to be vigilant about either overly conservative thinking without taking any protective measures or deliberately pursuing rule innovation and trying to create subversive legal rules, concluded Cong Lixian, dean of the Intellectual Property School at East China University of Political Science and Law.

Edited by YANG LANLAN