创新背景

职前教师在学习培训期间总是缺乏真正可应对实际情况的实践机会，无法对自己的经验做出正确适当的判断和推理。

创新过程

德国慕尼黑大学教育和教育心理学教授Frank Fischer和Michael Sailer博士与达姆施塔特技术大学计算机科学教授Iryna Gurevych合作，进行了一项将人工智能与职前教师培训相结合的研究，开发了一种教师培训学习的新方法，研究成果《来自人工神经网络的自适应反馈有助于职前教师在基于模拟的学习中进行诊断推理》于2022年4月发表在《学习和指导》上。

研究开发了一种可根据职前教师的个人表现提供自适应反馈的人工神经网络，将人工智能于教师培训相结合。经过训练后，人工智能可以在线模拟培训过程中的经验，为之前教师所学的理论和技能提供更多的实践机会。
在模拟中，职前教师需要复杂的反馈来发展复杂的技能，如诊断推理。在线模拟可以训练职前教师的诊断推理，让受训者认识到学习过程中的某些困难，并提供书面理由，说明他们为什么怀疑某一特定困难。在由职前教师撰写的这些书面理由中，人工智能会辨别确定学习者做对和做错的事，给出相应的反馈。

研究进行了一项以178名职前教师为学习困难模拟学生的实验，将自动自适应反馈与专家提供的解决方案（静态反馈）进行比较，探究了基于人工神经网络的自动自适应反馈对职前教师诊断推理的影响以及诊断推理的准确性和理由是否合理充分。

并且，实验把控学习者是在基于计算机实验室的模拟中单独工作或以二元组的方式工作。实验结果表明，自适应反馈可以帮助提高职前教师在书面作业中的理由质量，但不能提高诊断推理的准确性，且静态反馈会对二元组学习的协作者产生不利影响。
模拟中的自动自适应反馈为高等教育培训中的大量学生提供可扩展的、精心设计的、以过程为导向的实时反馈。研究人员表示，使用人工智能和提供个性化反馈改善了实习教师的诊断推理。人工智能的部署增加了价值，特别是对于复杂技能的进步，有助于对复杂推理结果的学习。

创新关键点

结合人工智能和职前教师培训，帮助学习者发现自己的学习困难并增加实践机会。

Artificial intelligence combined with teacher training, adaptive feedback helps to identify learning difficulties

Frank Fischer and Dr Michael Sailer, professors of educational and educational psychology at the University of Munich in Germany, in collaboration with Iryna Gurevych, professor of computer science at the Technical University of Darmstadt, conducted a study that combined artificial intelligence with pre-vocational teacher training, developing a new method of teacher training learning, the result of which "Adaptive feedback from artificial neural networks helps pre-service teachers to make diagnostic reasoning in simulation-based learning" Published in Learning and Mentoring in April 2022.

Research has developed an artificial neural network that can provide adaptive feedback based on the personal performance of pre-service teachers, combining artificial intelligence with teacher training. After training, ai-tech can simulate the experience during the training process online, providing more practical opportunities for the theories and skills previously learned by teachers.

In simulations, pre-service teachers need complex feedback to develop complex skills such as diagnostic reasoning. Online simulations can train pre-service teachers in their diagnostic reasoning, make trainees aware of certain difficulties in the learning process, and provide written reasons why they suspect a particular difficulty. In these written reasons written by pre-service teachers, ai-tech will discern what determines learners are doing right and wrong, and give feedback accordingly.

In an experiment involving 178 pre-service teachers simulating students with learning difficulties, the study compared the automatic adaptive feedback with the solution provided by the experts (static feedback) to explore the impact of automatic adaptive feedback based on artificial neural networks on the diagnostic reasoning of pre-service teachers and whether the accuracy and rationale of diagnostic reasoning were reasonable and sufficient.

Moreover, the experimental control learner works alone or in a binary way in a computer-based lab-based simulation. Experimental results show that adaptive feedback can help improve the quality of reasons in pre-service teachers' written assignments, but it cannot improve the accuracy of diagnostic reasoning, and static feedback can adversely affect collaborators in binary learning.

Automated adaptive feedback in simulation provides scalable, well-designed, process-oriented, real-time feedback to a large number of students in higher education training. The researchers say that using artificial intelligence and providing personalized feedback improves the diagnostic reasoning of intern teachers. The deployment of AI adds value, especially for the advancement of complex skills, facilitating the learning of complex reasoning outcomes.