华东师范大学学报(教育科学版) ›› 2024, Vol. 42 ›› Issue (8): 64-75.doi: 10.16382/j.cnki.1000-5560.2024.08.006

• 人工智能与教育转型 • 上一篇    下一篇

通用大模型教育应用的潜在风险及其规避——基于技术伦理的视角

吴砥, 吴河江   

  1. 华中师范大学教育部教育信息化战略研究基地(华中),武汉 430079
  • 接受日期:2024-04-11 出版日期:2024-08-01 发布日期:2024-07-23
  • 基金资助:
    2022年度教育部哲学社会科学研究重大课题攻关项目“教育数字化转型的国际比较研究”(22JZD045)。

Potential Risks of the Application of General Large Model Education and Their Avoidance: From the Perspective of Technology Ethics

Di Wu, Hejiang Wu   

  1. Educational Informatization Strategy Research Base of the Ministry of Education (Central China), Central China Normal University, Wuhan 430079, China
  • Accepted:2024-04-11 Online:2024-08-01 Published:2024-07-23

摘要:

以ChatGPT为代表的通用大模型在教育行业快速普及,可能引发教育的系统性变革乃至生态重构。虽然技术应用于教育可能导致的伦理问题多年来一直受到关注,但从未像通用大模型的教育应用这样引发人们对技术伦理问题的担心。一方面,技术作用于教育活动的基本原理具有同质性,因此通用大模型教育应用的伦理风险类型较之以往具有趋同性,然而,因其智能化水平不同,在风险程度上将可能加剧传统伦理问题:在教育安全方面,对教育主体与组织的隐私窃取可能加剧教育过程安全危机、意识形态渗透可能加剧教育结果安全危机;在教育关系方面:可能加剧“师—机”关系中教师的技术恐惧与技术崇拜,以及“机—生”关系中学生的技术沉溺;在教育公平方面,可能加剧教育起点不公平与扩大教育结果不公平。另一方面,通用大模型有着特殊的技术基础与特征,为教育带来更为颠覆性的变革并导致新异化现象,可能产生新的伦理风险:在教育安全方面,大模型系统的对抗性攻击风险可能使学生面临算法规制而产生新的认知安全问题、模型能力的涌现性可能会让通用大模型在教育应用中失控;在教育关系方面,“人—机”协同育人模式可能阻碍师生间情感与思想交流;在教育公平方面,使用者可能在毫无察觉之下受到通用大模型的不公平对待,教育过程公平很可能难以保障。为规避通用大模型教育应用可能产生的伦理风险,应注重在制作层面确保通用大模型的道德化,在实践层面细化对通用大模型教育应用的指导,在管理层面建立通用大模型教育应用的监管体系,在研究层面加强对智能时代教育行业技术伦理问题的理论研究。

关键词: 通用大模型, 技术伦理, 教育安全, 教育关系, 教育公平

Abstract:

The rapid popularization of the general large model represented by ChatGPT in education industry may lead to systematic reform and even ecological reconstruction of education. While the ethical issues that can arise from the use of technology in education have been a concern for many years, never has the educational application of general large model raised such concerns about the ethics of technology. On the one hand, the basic principle of the effect of technology on educational activities is homogeneous, so the types of ethical risks in the application of general large model education are similar compared with the past. However, due to the different level of intelligence, the degree of risks may aggravate traditional ethical problems. In terms of educational security, the privacy theft of educational subjects and organizations may aggravate the security crisis of educational process, and ideological infiltration may aggravate the security crisis of educational outcome. In terms of educational relationship, it may aggravate the teachers’ fear and worship of technology in the teacher-machine relationship, and the students’ addiction to technology in the machine-student relationship. From the aspect of education equity, it may aggravate the inequity of education starting point and expand the inequity of education result. On the other hand, general large model has special technical basis and characteristics, which brings more subversive changes to education and leads to new alienation phenomenon, which may produce new ethical risks. From the aspect of education security, the adversarial attack risk of large model system may make students face algorithm regulation and produce new cognitive security problems, and the emergence of model ability may make generative AI out of control in education application. In terms of educational relationship, the “man-machine” cooperative education model may hinder the emotional and ideological exchange between teachers and students. In terms of educational equity, users may be treated unfairly by the general model without being aware of it, and the fairness of the educational process may be difficult to guarantee. In order to avoid the ethical risks that may arise from the application of general large model in education, attention should be paid to ensuring the moralization of general large model at production level, specifying the guidance for general large model in education application practice level, establishing the supervision system for general large model in education application at management level, and strengthening the theoretical research on the technical ethics of education industry in the intelligent era at the research level.

Key words: general large model, technical ethics, education security, educational relationships, equity in education