Deep cognitive diagnosis in intelligent tutoring systems in the framework of logic programming and meta-level reasoning

在领域专业知识、任务模型（解决问题的状态）和教学专业知识（预测学生错误）的背景下诊断学生的输入，对于在智能辅导系统中提供有效的脚手架和补救反馈至关重要。一个主要的见解是，学生的错误很少是随机的，而是由于正确执行错误的过程。因此，有效的教学依赖于深刻的认知分析来诊断并随后修复那些不正确的部分。本项目旨在发展逻辑规划和元级推理框架下的认知诊断方法。它将适应并进一步发展算法调试技术，以识别不正确（基于prolog）程序中的学生错误；它将有助于归纳逻辑编程的进步，从学生行为和背景知识的例子中合成Prolog程序，它将提出自动代码扰动来系统地引导错误以纠正程序代码。这些技术将用于构建软件，以帮助教师提高诊断技能，并帮助学生学习基本的算术和符号微分。基于逻辑的方法不仅可以补充现有的认知诊断方法；它还将揭示它们的性质，增加比较分析，从而有助于更系统地深入分析智能辅导系统中学生输入的方法。

The diagnosis of student input in the context of domain expertise, task model (problem solving state), and pedagogical expertise (anticipating student errors) is central for the provision of effective scaffolding and remedial feedback in intelligent tutoring systems. A main insight is that student errors are seldom random, but result from correctly executing an erroneous procedure. Effective teaching, thus, depends on deep cognitive analyses to diagnose and subsequently repair those incorrect parts. This project aims at developing methods for cognitive diagnosis in the framework of logic programming and meta-level reasoning. It will adapt and further develop the technique of algorithmic debugging to identify student error in incorrect (Prolog-based) procedures; it will contribute to advances in inductive logic programming to synthesise Prolog programs from example student behaviour and background knowledge, and it will propose automatic code perturbation to systematically inict bugs to correct program code. The techniques will be used to build software to support teachers to improve their diagnostic skills, and students to learn basic arithmetic and symbolic differentiation. The logic-based approach will not only complement existing methods for cognitive diagnosis; it will also shed new light on their nature, add to a comparative analysis, and thus contribute toward a more systematic approach for the deep analysis of student input in intelligent tutoring systems.