Bias in Generative Artificial Intelligence:Manifestations,Mechanisms and Governance Paths
As generative AI demonstrates its transformative potential across diverse industries and fields,the issue of bias that comes with it has become increasingly severe.In the cultural domain,gener-ative AI reproduces a Western-centric cultural universality,ignoring the uniqueness of different cultures;in politics,generative AI exhibits obvious partisan,ideological,and geopolitical biases,reinforcing the unequal global political and economic power structures;in gender,generative AI amplifies inherent soci-etal gender biases through defamation,stereotypes,recognition biases,and underrepresentation.These biases arise from the mutual reinforcement,amplification,and reproduction of endogenous and exogenous biases in an ongoing cycle.The governance of bias in generative AI bias is not only a technical challenge but also a severe test of global ethics and social equity.It requires the construction of a multi-party collab-orative framework of"technology entity,user entity,regulatory entity".The technology entity should ad-here to the"alignment"principle to reduce systematic bias,the user entity should enhance their AI liter-acy and engage in bias governance practice,and the regulatory entity should establish a dynamic bias cor-rection and supervision system,to promote a balance between the development and security of generative AI,achieving its harmonious coexistence with human society.
万方数据
