技术演化与涌现风险:生成式人工智能的协同式敏捷治理体系研究  

Technological Evolution and Emergent Risks:A Study on Collaborative Agile Governance Systems for Generative Artificial Intelligence

作　　者：郭全中 GUO Quan-zhong(College of Journalism and Communication,Minzu University of China,Beijing 100081,China)

机构地区：[1]中央民族大学新闻与传播学院,北京100081

出　　处：《编辑之友》2025年第4期49-56,共8页Editorial Friend

基　　金：北京市社会科学基金规划重点项目“首都互联网平台企业社会责任与协同治理体系研究”。

摘　　要：生成式人工智能具有强大的学习能力和自适应能力,以及自进化等显著特征,但同时也带来了智能涌现风险。与传统风险不同,涌现风险表现出一定的主体性,具备自主决策能力。由于AI能够不断自我演化,其潜在风险可能被掩盖,尤其是在其黑箱式运作机制下,可能产生认知偏差甚至幻觉现象。这种现象不仅影响AI的可控性,还削弱了人机关系的可持续性。此外,AI的自适应性使得风险呈现出动态演化的特征。为应对这一挑战,需从技术底层进行优化,激发多方协同治理的潜力,并通过协同式的敏捷迭代治理模式,及时应对复杂变化的风险。Generative artificial intelligence exhibits formidable learning capabilities,self-adaptive capacities,and self-evolutionary traits,aligning with developers'original intentions while simultaneously introducing emergent risks.Distinct from traditional risks,emergent risks demonstrate a degree of subjective agency and autonomous decision-making capabilities.Due to AI's capacity for continuous self-evolution,latent risks may remain obscured,particularly under its black-box operational mechanisms,which can induce cognitive biases or even hallucinatory phenomena.Such issues not only compromise the controllability of AI but also undermine the sustainability of human-AI relationships.Furthermore,the adaptive nature of AI allows risks to manifest dynamic evolutionary characteristics.To address these challenges,it is imperative to optimize foundational technical architectures,harness the potential of multi-stakeholder collaborative governance,and adopt a collaborative agile iterative governance model to promptly respond to the complex and evolving nature of risks.

关 键 词：技术演化 涌现风险 AI治理 协同监管 敏捷治理 

分 类 号：G206[文化科学—传播学]