Building A Teacher-AI Collaborative System for Personalized Instruction and Assessment of Comprehension Skills

学生需要基本的读写能力，尤其是阅读理解能力，才能成功地从事STEM学习和职业。许多研究探索了使用人工智能技术，如聊天机器人，通过让学生在阅读过程中参与互动对话来提高学生的阅读理解能力。这种方法尤其适用于处于发展阅读技能关键时期的年轻学生。然而，扩大这些人工智能资源，使其能够为不同的学习者和教师所用，仍然是一项挑战。该项目旨在利用人工智能的最新进展，特别是大型语言模型，使教师能够与人工智能合作创建交互式阅读资源，向学生提问，倾听和解释学生的反应，并在阅读期间为学生提供量身定制的反馈，重点关注幼儿园到二年级的学生。这将使教师能够贡献他们的专业知识，开发适合学生需求的人工智能资源。该项目将阐明，通过人工智能-教师协作，是否可以显著减少聊天机器人开发中通常由内容创作者、设计师和工程师执行的大量体力劳动，以及由此产生的聊天机器人是否有效地支持教师的教学并促进学生的阅读理解。该项目将分四个阶段进行。第一阶段涉及开发创新的人工智能模型，根据教师选择的阅读材料自动生成问答对。这些模型将根据教育环境的独特要求进行调整。在第二阶段，将通过上下文调查和参与式设计过程开发一个用户友好的教师-人工智能协作系统。该系统将使教师能够验证和修改人工智能生成的问答对，并随后将其整合到聊天机器人中，与学生进行对话。教师对问答对的修改会反馈给系统，这样人工智能模型就可以逐渐学习和适应每个教师的偏好。在第三阶段，研究团队将开发聊天机器人的自适应交互能力，使其能够进行对话，并根据学生回答的准确性和情感提供脚手架。第四阶段将包括检查教师-人工智能协作系统的可用性和有效性，以及由此产生的聊天机器人在支持个性化教学和评估方面的作用。为此，研究组将进行低功率随机对照试验的现场试验。五名教师和他们的大约150名学生将被招募参加，每个班级中有一半的学生被随机分配使用教师生成的聊天机器人阅读互动文本，而另一半则阅读没有聊天机器人的原始文本。对教师和学生的观察和访谈将揭示教师与人工智能共同创建的交互式阅读材料的可用性。学生的阅读后理解能力将被评估，为该系统的教育影响提供证据。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Students need fundamental literacy skills, especially reading comprehension, to successfully engage in STEM learning and careers. A number of studies have explored the use of AI technologies, such as chatbots, to improve students' reading comprehension by engaging students in interactive dialogue during reading. This approach is particularly promising for younger students who are in a critical period for developing reading skills. However, scaling up these AI resources to make them accessible and relevant to diverse learners and instructors remains a challenge. This project aims to harness recent advancements in AI, particularly large language models, to enable teachers to collaborate with AI in creating interactive reading resources that ask students questions, listen to and interpret student responses, and provide tailored feedback to students during reading, focusing on students from kindergarten to second grade. This will allow teachers to contribute their expertise to develop AI resources tailored to their students' needs. The project will shed light on whether the extensive manual labor involved in chatbot development, typically performed by content creators, designers, and engineers, can be significantly reduced through AI-teacher collaboration, and whether the resulting chatbots effectively support teachers' instruction and promote students’ reading comprehension. This project will be carried out in four stages. This first stage involves the development of innovative AI models to automatically generate question-answer pairs based on reading materials teachers select. The models will be tailored to meet the unique requirements of educational contexts. In the second stage, a user-friendly teacher-AI collaborative system will be developed through a contextual inquiry and participatory design process. This system will enable teachers to verify and modify the question-answer pairs generated by AI and subsequently incorporate them into a chatbot that engages students in dialogue. Teachers' modifications to the question-answer pairs will feed back to the system so that the AI models can gradually learn and adapt to each individual teacher's preferences. In the third stage, the research team will develop the chatbot's capability for adaptive interaction so that it can carry out dialogue and provide scaffolding based on both the accuracy and sentiment of students' responses. The fourth stage will involve an examination of the usability and effectiveness of the teacher-AI collaborative system and resulting chatbot in supporting personalized instruction and assessment. To this end, the research team will carry out a field test involving an under-power randomized controlled trial. Five teachers and their approximately 150 students will be recruited to participate, with half of the students in each class randomly assigned to read interactive texts with a chatbot generated by their teacher while the other half reads the original text without the chatbot. Observations and interviews with teachers and students will shed light on the usability of the teacher-AI co-created interactive reading materials. Students' post-reading comprehension will be assessed to provide evidence on the system's educational impact.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.