教育改革视域下生成式预训练模型的伦理风险与治理策略

Ethical Risks and Governance Strategies of Generative Pre-Training Models in the Perspective of Educational Reforms

生成式预训练模型以其高度个性化的学习体验和基于数据与反馈的动态学习方式重构了教育空间和教学方法,为教育体系带来了颠覆性变革。但随着生成式预训练模型普遍化,尤其是ChatGPT的崛起,一些新的伦理风险也逐渐凸显出来,其中包括技术风险、内容风险、数据风险和算法风险等。这些风险既有生成式预训练模型本身在训练过程中不自觉产生的偏见或误导性言论对人的侵犯,也有人类本身的因素如:开发者算法霸权、个体技术认知缺乏、教师数据素养欠缺等。针对这些伦理风险,可以从加强理论研究和知识普及、研发符合我国教育模式的生成式预训练模型、推动教育形态的结构性变革来进行伦理风险治理,以确保生成式预训练模型技术在教育领域的应用符合教育伦理规范。

Generative pre-training models have brought disruptive changes to the education system by recon-figuring the education space and teaching methods with their highly personalised learning experience and dy-namic learning based on data and feedback.However,with the generalisation of generative pre-training mod-els,especially the rise of ChatGPT,a number of new ethical risks have come to the forefront,including tech-nological risks,content risks,data risks and algorithmic risks.These risks range from the violation of hu-man beings by biased or misleading statements unconsciously produced by the generative pre-training models themselves during the training process,to human beings themselves such as:developers’algorithmic hege-mony,individuals’lack of technological awareness,and teachers’lack of data literacy.To address these ethical risks,ethical risk management can be carried out by strengthening theoretical research and knowl-edge popularisation,developing generative pre-training models in line with China’s education model,and promoting structural changes in education patterns to ensure that the application of generative pre-training model technology in education is in line with the ethical norms of education.