道德编程:ChatGPT对提问者道德心理的良性效应与潜在风险  

Moral Programming:Benign Effects and Potential Risks of ChatGPT on Questioners’Moral Psychology

作　　者：严玉 马玉洁 YAN Yu;MA Yujie(School of Law,Ningxia University,Yinchuan 750000,China;School of Ethnology and Sociology,Minzu University,Beijing 100081,China)

机构地区：[1]宁夏大学法学院,宁夏银川750000 [2]中央民族大学民族学与社会学学院,北京100081

出　　处：《燕山大学学报（哲学社会科学版）》2024年第6期89-96,共8页Journal of Yanshan University:Philosophy and Social Science

基　　金：国家社科基金重大项目“中华民族道德生活中的价值认同研究”(22&ZD039)。

摘　　要：ChatGPT及类似的大型语言人工智能在各领域扮演日益重要的角色,其与人类的互动已成不可避免的现实,人们对其信任、依赖,情感投入逐渐增加,这也导致了对人类道德心理的影响。虽然ChatGPT在回答问题时坚持遵循道德原则,但其回答模板存在明显不足,如缺乏人类情感理解能力、无法提供个性化建议,且容易在遭遇“越狱攻击”时产生道德漏洞。如果提问者将ChatGPT视为可信信息源,并根据其回答做出道德决策,则不道德或有偏见的回答可能导致错误的道德决策,需要采取相关防范措施以应对智能时代的道德困境,促进个体道德心理增强和社会稳定。ChatGPT and similar large-scale language artificial intelligence play increasingly important roles in various fields,and their interaction with humans has become an unavoidable reality.People’s trust,dependence,and emotional investment in them are gradually increasing,leading to an influence on human moral psychology.Although ChatGPT adheres to ethical principles inanswering questions,its response templates have remarkable shortcomings,such as a lack of human emotional understanding,inability to provide personalized advice,and susceptibility to moral loopholes inencountering“jailbreak attacks”.If questioners perceive ChatGPT as a credible source of information and make moral decisions based on its responses,unethical or biased responses may lead to erroneous moral decisions.Therefore,it is necessary to take relevant preventive measures to address the ethical dilemmas in the intelligent era,so as topromote the enhancement of individual moral psychology and ensure social stability.

关 键 词：ChatGPT 大语言模型 越狱攻击 道德心理 人机关系 

分 类 号：G23[文化科学]