GAI时代批判信息学的价值迭代与多元批判体系构建OACHSSCD
Value Iteration and Construction of a Multi-Agent Critical System in Critical Informatics for the GAI Era
AI越强,批判信息学越有必要,因为这一"超级技术"带来的潜在风险正快速转换为现实威胁.GAI时代的批判信息学需以"双向赋能"为核心价值取向:一方面赋能AI技术发展,通过批判性观察促进人本主义设计创新;另一方面赋能用户与社会,提升公众对GAI风险的防范能力与参与水平.秉持双向赋能价值取向的批判信息学应当将四类微观对象纳入"批判性观察"范畴:智能算法的隐性漏洞(如"智能涌现"生成伪解释操纵认知)、服务设计的负面效应(如主动帮助削弱用户能力自尊)、微调过程的风险放大(如次级开发引发信息生态畸变)、用户行为的认知外包问题.为应对挑战,应构建涵盖AI自我批判、用户批判AI、政府与社会力量协同批判的多元批判体系,通过强化各类行动者的反思能力与治理参与,构建立体化风险防线.
The stronger AI becomes,the more necessary Critical Informatics is,as the potential risks brought by this"super-technology"are rapidly transitioning into tangible threats.In the era of GAI(Generative Artificial Intel-ligence),critical informatics should adopt"bidirectional empowerment"as its core value orientation:on one hand,empowering the development of AI technology by promoting humanistic design innovation through critical inquiry;on the other hand,empowering users and society by enhancing the public's ability to mitigate GAI risks and their level of participation.Thus Critical informatics could focuses on four categories of micro-level critical targets:hidden vulnerabilities in intelligent algorithms(such as the"emergence of intelligence"generating pseudo-explanations to manipulate cognition),negative effects of service design(such as proactive assistance undermining users' self-es-teem and competence),risk amplification during fine-tuning processes(such as secondary development triggering distortions in the information ecosystem),and issues of cognitive outsourcing in user behavior.To address these chal-lenges,a multi-faceted critical system should be established,encompassing AI self-critique,user critique of AI,and collaborative critique by governmental and societal forces.By strengthening the reflective capabilities and governance participation of various actors,a comprehensive risk defense framework can be constructed.
李桂华
四川大学公共管理学院,四川 成都 610065
社会科学
生成式人工智能批判信息学价值迭代多元批判体系
Generative Artificial Intelligencecritical Informaticsvalue iterationmulti-agent critical system
《成都大学学报(社会科学版)》 2026 (2)
2-6,5
评论