CAREER: Learning about Learning

拟议的项目将研究建立基于网络的智能辅导系统的方法，该系统可以在评估时进行教学。该项目的重点是创建一个系统，帮助八年级学生学习马萨诸塞州综合评估调查（MCAS）的数学。该系统提供了PI和他在CMU的同事们所说的辅助：计算机生成的个别辅导，将学习评估与对学生的辅助结合起来，这样当他们接受个性化的计算机帮助时，教师可以实时准确地监控学生遇到的困难种类和他们取得的进步。该项目将有5个主要的研究重点：在第一个研究重点中，设计认知模型，PI将使用困难因素分析和学习因素分析等方法，以创建可以预测学生学习和迁移的良好拟合模型。在第二个研究重点中，推断学生知道什么和正在学习什么，PI将研究新的模型学习和推理机制，以便尝试预测学习和迁移何时发生。这一努力将需要心理测量方法与智能辅导系统相结合。在第三个推动力——优化学习方面，PI将侧重于发现教师使用的教学策略能让学生更好地学习。通过对他的模型的迭代改进，他将设计出选择问题的机制，使预期的考试成绩获得与完成所需的预期时间的比例最大化。第四个重点是告知教育工作者，PI将研究如何最好地将细粒度模型中的信息呈现给教师。在最后的推力中，允许用户自适应，PI将尝试提供一种方法，允许K-12教师设计智能辅导系统中使用的教学法。在他的网站上，他将提供研究团队目前关于理想的八年级智能辅导系统的“最佳假设”。教师将能够根据自己的需要调整系统，改变现有的辅助或为学生创造新的问题。希望这样的系统最终能被大量的教师使用。

The proposed project will investigate methods of building web-based intelligent tutoring systems that teach as they assess. The project focuses upon the creation of a system to help 8th grade students learn math for the Massachusetts Comprehensive Assessment Survey (MCAS). The system provides what the PI and his colleagues at CMU call assistments: individual pieces of computer-generated tutoring that combine assessment of learning with scaffolded assistance to the students such that as they receive individualized computer help, teachers can monitor, in real time, precisely the kinds of difficulty students are having and what progress they are making. The project will have 5 main research thrusts: In the first research thrust, designing cognitive models, the PI will use methodologies such as difficulty factors analysis and learning factors analysis in order to create good-fitting models that can predict student learning and transfer. In the second research thrust, inferring what students know and are learning, the PI will study novel model learning and inference mechanisms in order to try to predict when learning and transfer occur. This effort will entail combining psychometric methods with intelligent tutoring systems. In the third thrust, optimizing learning, the PI will focus on discovering what pedagogical strategies used by teachers lead to better learning by students. Through the iterative refining of his models, he will devise mechanisms that select problems that maximize the ratio of expected test-score gain to the expected time needed for completion. In the fourth thrust, informing educators, the PI will study how best to present information from the finer-grained model to teachers. In the final thrust, allowing user adaptation, the PI will attempt to provide a means for allowing K-12 teachers to design the pedagogy used in an intelligent tutoring system. On his website, he will provide the research team's current "best hypothesis" of an ideal 8th grade intelligent tutoring system. Teachers will be able to adapt the system for their own uses, changing existing assistments or creating new problems for their students. The hope is that such systems could eventually be made available to large numbers of teachers.