基于LDA模型的人工智能伦理准则体系研究

Latent Dirichlet Allocation Model-based Research on System of Ethical Guidelines for Artificial Intelligence

摘要:

利用隐含狄利克雷分布（Latent Dirichlet Allocation，LDA）模型对全球人工智能伦理准则体系进行内容分析，通过对准则关键词的阶段差异对比、核心主题的聚合体系分析，总结出世界范围内人工智能治理共识的6大演变趋势，均表明实践导向已成为核心共识。但人工智能伦理准则存在“可操作性受限、考量要素不同质、伦理洗白、法律转化难”等实践适用性困境，亟须在技术、组织及制度层面弥补实践缺口。

关键词:科技伦理风险;人工智能伦理准则;演进趋势;实践困境

Abstract:

Based on the latent Dirichlet Allocation（LDA ）model， an analysis was made of the content of the global artificial intelligence ethical guidelines system.Six major trends of the worldwide consensus on artificial intelligence governance were summarized by comparing the stage differences of keywords and aggregating the core themes of the guidelines. Results indicate that practice orientation becomes the core consensus. However， the implement of the artificial intelligence ethical guidelines faces the dilemmas such as limited operability， heterogeneous considerations， ethical whitewashing， and difficult legal translation. Fianally， a discussion was made on the details to bridge the practical gap.

Key words:ethical risks in technology;artificial intelligence ethical guidelines;evolutionary trends;practical dilemmas