4月17日上午，联合教育学院二楼会议室举办了Gavin教授的人工智能与机器学习领域验证问题研究工作坊。
Gavin教授深入探讨了机器学习在教育中的应用，尤其是有关评估学生书面作业。他以案例将传统人工评估与机器学习自动化评估进行对比，表示需要把数据分为训练集和验证集，以此评估模型效能的重要性。同时他阐释了计算机科学与教育视角在验证方面的差异，指出前者注重性能指标，后者侧重模型实际效用，并拓展到机器学习在教育中面临的透明度、数据质量、见解解释等挑战。另外，他通过案例生动展示了数据稀疏性、高效评估需求及内部验证与外部有效性差异等问题，强调全面验证方法的重要性，呼吁应当提高数据质量、提供可操作信息、确保技术透明度和可用性。
工作坊结束后，Gavin教授与同学们展开了热烈讨论。以下是同学们思考后提出的问题。
李游源：“I see TEVV as a basic clue for machine learning or just a framework for machine learning. Can we use TEVV in the fundamental logic of science teacher training,not just the teaching students.Will there be any difference come into this issue？”
韩子彤：“As the professor mentioned before, due to practical factors, it is impossible for students to complete a questionnaire with 20 questions per day. They can only complete three questions, but their abilities cannot be fully measured. As an educational researcher and teacher, how can we solve this problem? Can we use generative AI to solve this problem？”
斯楚慧：“作为算法的初学者，很多人对机器学习和大语言模型的理解都是——经过大数据训练的算法模型尝试给出一个最具可能性的答案。但我们都不了解演算的过程和原理，也无法为演算的信效度评估负责。这就导致很多一线教师相比算法模型，更愿意相信自己在与学生的交谈、观察过程中产生的专业判断。那么，我们应该如何理解机器学习的原理？另外，美国的中小学教师通常如何使用AI工具？”
毛柯润：“The ML system in the lecture is constructed based on the skill network made in advance, so it should require teachers at least agree with the network for the system be helpful. Is that a thing in practice? ”
沈雨霏：“How would the focus of teachers' work shift if machine learning were introduced into formal classroom teaching?”
刘子涵：“What the difference between machine learning and AI learning ？”
蒋云杰：“When using machine learning to evaluate students written work.How to effectively solve the problem that the lack of transparency of the model makes it difficult for teachers to trust the evaluation results?”
冯靖然：“I still can't figure out where these large amounts of data come from when training this model?”
同学们踊跃提问，积极发言，Gavin教授对此一一做出了专业且细致的解答，充分回应了大家的困惑。此次工作坊帮助同学们循序渐进地提升自身能力，更好地适应人工智能时代的教育教学需求，让大家受益匪浅
编辑 | 李佳仪
审核 | 何 蕾 陈海垚
终审 | 黄 晓
