生成式人工智能社会实验的伦理问题及治理

科学学研究 ›› 2024, Vol. 42 ›› Issue (1): 3-9.

• 专稿:生成式人工智能治理 • 上一篇下一篇

生成式人工智能社会实验的伦理问题及治理

俞鼎¹^,²(), 李正风³^,⁴

¹ 浙江大学马克思主义学院, 浙江杭州 310058
² 浙江大学马克思主义理论创新与传播研究中心, 浙江杭州 310058
³ 清华大学社会科学学院, 北京 100084
⁴ 中国科学院学部-清华大学科学与社会协同发展研究中心, 北京 100084

收稿日期:2023-08-14 修回日期:2023-10-09 出版日期:2024-01-15 发布日期:2024-01-15
作者简介:俞鼎(1988-),男,讲师,博士,E-mail:am_yuding999@163.com。
李正风(1963-),男,教授、博士生导师。
基金资助:
浙江省社科规划青年课题“生成式人工智能(AIGC)社会实验的伦理问题及其对策研究”(24NDQN069YB)

Ethical issues and governance in generative artificial intelligence social experiment

YU Ding¹^,²(), LI Zheng-feng³^,⁴

¹ School of Marxism, Zhejiang University, Hangzhou 310058, China
² Research Center for Marxism Theory Innovation and Communication, Zhejiang University, Hangzhou 310058, China
³ School of Social Sciences, Tsinghua University, Beijing 100084, China
⁴ CASAD-Tsinghua University Research Center for Synergetic-development of Science and Society, Beijing 100084, China

Received:2023-08-14 Revised:2023-10-09 Online:2024-01-15 Published:2024-01-15

摘要/Abstract

摘要：

2022年年末以来,以ChatGPT为代表的生成式人工智能(AIGC)驱动了大语言模型(LLM)从“判别范式的弱人工智能”转向“生成范式的强人工智能”的范式变革,并将“强拟人化”特性介入了人工智能社会实验场景。这种特性源于实验者长期以反映真实世界对话场景的大型文本数据语料库来训练LLM的结果,所具有的类人心智功能表现出涌现、创制、泛化等特点并衍生了“不诚实拟人化”的新伦理特征,这些又反过来对AIGC社会实验造成了负面影响。本文通过揭示并解析该实验场景内的拟人化认知倾向滥用、理智德性消减、多元化偏见风险叠加与“实验者效应”非道德强化等拟人化伦理问题,提出应基于“制度-原则-策略-引导”四维视角,来前瞻性构建AIGC社会实验的伦理治理机制。

关键词: 生成式人工智能, 社会实验, 强拟人化, 拟人化伦理

Abstract:

Since the end of 2022, the generative artificial intelligence (AIGC), represented by ChatGPT, has quietly driven the paradigm revolution of the Large Language Model(LLM)-from ‘weak-AI to strong-AI’, and integrated the ‘powerful anthropomorphism’ feature into the social experimental context of artificial intelligence.This characteristic stems from the long-term training of LLM by experimenters with large corpus of text data reflecting real-world conversation scenes.Its humanoid mental function shows the characteristics of emergence, creation and generalization, and derives a new ethical feature of ‘dishonest anthropomorphism’, which in turn has a negative impact on AIGC social experiments.By revealing and analyzing the anthropomorphic ethical problems such as the abuse of anthropomorphic cognitive tendency, the reduction of rational virtue, the superposition of diversified prejudice risks and the immoral enhancement of ‘experimenter effect’ in this experimental context, this paper puts forward that the ethical governance mechanism of AIGC social experiment should be established prospectively based on the four-dimensional perspective of ‘regulation-principle-strategy-guidance’.

Key words: generative artificial intelligence, social experiment, powerful anthropomorphism, anthropomorphic ethics

中图分类号:

俞鼎, 李正风. 生成式人工智能社会实验的伦理问题及治理[J]. 科学学研究, 2024, 42(1): 3-9.

YU Ding, LI Zheng-feng. Ethical issues and governance in generative artificial intelligence social experiment[J]. Studies in Science of Science, 2024, 42(1): 3-9.

参考文献 13

[1]	简圣宇. GPT语言模型:作为“类人型”人工智能的技术准备[J]. 广州大学学报(社会科学版), 2023, 22(4):19-24.
	Jian S Y. GPT language model: Technical preparation for humanoid artificial intelligence[J]. Journal of Guangzhou University(Social Sience Edition), 2023, 22(4):19-24.
[2]	李坤海, 徐来. 人工智能对侵权责任构成要件的挑战及应对[J]. 重庆社会科学, 2019(2):55-65.
	Li K H, Xu L. The challenge of artificial intelligence to the constitutive elements of tort liability and its countermeasures[J]. Chongqing Social Sciences, 2019(2):55-65.
[3]	Kosinski M. Theory of mind may have spontaneously emerged in large language models[J]. arXiv preprint arXiv:2302.02083, 2023.
[4]	Moghaddam S R, Honey C J. Boosting theory-of-mind performance in large language models via prompting[J]. arXiv preprint arXiv:2304.11490, 2023.
[5]	陈巍. 拟人化人工智能:神经基础与伦理承诺[N]. 中国社会科学报,2022-12-13(007).
	Chen W. Anthropomorphic artificial intelligence: Neural basis and ethical commitment[N]. Chinese Social Sciences Today,2022-12-13(007).
[6]	Epley N, Waytz A, Cacioppo J T. On seeing human:A three-factor theory of anthropomorphism[J]. Chinese Social Sciences Today,Psychological Review, 2007, 114(4):864-886.
[7]	Kaminski M, Rueben M, Grimm C. Averting robot eyes[J]. Maryland Law Review, 2017, 76:983.
[8]	Darling K. Extending legal protection to social robots: The effects of anthropomorphism, empathy, and violent behavior towards robotic objects[A]. Calo F K. Robot Law[C]. Edward Elgar, 2016:1-25.
[9]	Mourey J A, Olson J G, Yoon C. Products as pals:Engaging with anthropomorphic products mitigates the effects of social exclusion[J]. Journal of Consumer Research, 2017, 44(2): 414-431.
[10]	AI21 Labs. AI21 Labs concludes largest Turing Test experiment to date[EB/OL]. https://www.ai21.com/blog/human-or-not-results?utm_source=talkingdev.2023-04-13/2023-07-08.
[11]	喻英豪. 人工智能拟人化对个人不道德行为的影响研究[D].中南财经政法大学, 2022:16.
	Yu Y H. The effect of artificial intelligence anthropomorphism on individual unethical behavior[D]. Zhongnan University of Economics and Law, 2022:16.
[12]	European Commission. Proposal for a regulation of the European Parliament and of the Council laying down harmonised rules on Artificial Intelligence (Artificial Intelligence Act) and amending certain union legisilative acts, COM(2016)206 final[EB/OL]. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52021PC0206.2021-04-21/2023-7-27.
[13]	Boston Gov. City of boston interim guidelines for using generative AI[EB/OL]. https://www.boston.gov/sites/default/files/file/2023/05/Guidelines-for-Using-Generative-AI-2023.pdf.2022-05-18/2023-07-12.

生成式人工智能社会实验的伦理问题及治理

Ethical issues and governance in generative artificial intelligence social experiment

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

参考文献 13

相关文章 4

编辑推荐

Metrics

本文评价

[1]	李子浩, 王迎春. 科技脱钩下中国生成式AI安全风险识别与应对[J]. 科学学研究, 2026, 44(6): 1121-1128.
[2]	刘雨盼陈芹芹张胜. 用户知识的生产机制研究[J]. 科学学研究, 2026, 44(5): 972-981.
[3]	郭小东. 大模型时代全球人工智能治理的挑战与中国方案[J]. 科学学研究, 2025, 43(4): 694-702.
[4]	陈升, 刘子俊, 张楠. 数字时代生成式人工智能影响及治理政策导向[J]. 科学学研究, 2024, 42(1): 10-20.