【摘要】 大模型的巨大应用潜能与伴随而来的伦理问题引发了教育领域的广泛关注。伦理问题虽不是大模型所独有，但目前的技术架构增加了AI教育应用伦理的复杂性。大模型导致的伦理问题与自身的技术特征息息相关，脱离技术探讨AI伦理问题无助于理解其教育应用的真实困境。引发大模型教育应用伦理困境的主要技术诱因，包括预训练、交互、交付方式对数据隐私保护的隐患，以及偏见、有毒文本与幻觉的风险行为对生成内容管控的挑战。为应对这些问题，要形成“人在回路中”的政策监管，推进大模型应用民主化，以及突出基于嵌入式伦理的大模型普及教育。本研究强调了技术伦理审视在确保教育公平、有效和有益的重要作用，为教育工作者和技术开发者了解和应对大模型的伦理风险提供全方位的视角。更多还原

【Abstract】 Large language models(LLMs) are increasingly capturing the attention of the educational sector, showcasing vast potential alongside rising ethical concerns. The ethical challenges presented by LLMs in educational settings are closely tied to their technical nuances. Disregarding the interplay between ethics and these technical aspects will fail to fully capture the intricacies involved. This study investigates the key technological factors leading to ethical conundrums with LLMs in education. It addresses concerns like data privacy risks inherent in the pre-training, interaction, and deployment of LLMs; and challenges in regulating content due to biases, generation of toxic text, and hallucinatory outputs. To navigate these challenges, the study advocates for integrated strategies, including policy frameworks with “human-in-the-loop” oversight, widespread and equitable access to LLM technology, and a focus on incorporating ethics directly into LLM-focused education. This study emphasizes the importance of rigorous ethical scrutiny to ensure fairness, effectiveness, and positive impacts in educational applications, and it serves as a guide for educational practitioners, and developers in grappling with the ethical nuances of LLMs.更多还原