Evidence-Based Adaptation of Learning: Supporting Informed Decision-Making

人们目前被信息淹没，需要支持才能在教室等复杂环境中理解这些信息。提供这种支持的一种方式是通过对人、他们的环境和他们的经历进行建模。这种建模使适应环境，用户体验或用户能力的元素可以得到改善。本提案的具体研究背景将集中在用户模型的教育应用上，这些用户模型将被称为学生或学习者模型。然而，由此产生的建模和可视化方法应该推广到其他应用领域，特别是当这些领域面临的问题没有正确或不正确的解决方案。相反，所提出的问题和所提供的解决办法是否适当是可以解释的。通过人们与教育技术的互动产生的大量数据可以被提取并用于推断有关学生，教师和学习环境的信息。然后，做出这些推断的算法的输出可以用于通过个性化来推动教育改进。不幸的是，由于两个原因，这一点没有得到有效的实现：*1）大多数为没有正确或错误答案的领域创建的学生模型都没有理论基础，这意味着无法对学习材料，学习过程或学习环境本身进行原则性调整，和 *2）现有的学生模型没有以支持利益相关者决策的方式进行沟通，因为它们没有传达基础学生模型的局限性，并且它们没有以利益相关者可以理解的方式传达模型信息。**为了缓解这些问题，该研究计划将认知科学的知识和方法与计算机科学的几个领域（人工智能，人机交互，用户建模和个性化以及教育数据挖掘）相结合，以开发人们的知识，信仰，技能和兴趣的理论基础计算模型。这些新的学生模型将以可理解的方式提供更细致入微的信息，以便人员和软件可以适当地调整和个性化为学生提供的支持和活动。这些技术的使用应该帮助人们学习更多，享受学习，并做出更好的决策，因为这些学生模型及其可视化将帮助人们了解他们的学习，以及教师的情况，他们的学生是如何学习的。这种使用证据来推动教育改进的方法将在真实的学习环境中进行评估。工具将被构建，以便它们与主要的教育技术兼容，一旦它们被证明是有效的，它们将被免费发布。

People are currently inundated with information and need support to make sense of this information in complex environments, such as classrooms. One way to provide this support is through the modelling of people, their environments, and their experiences. This modelling enables adaptation so elements of the environment, user experience, or user abilities can be improved. The specific context of study for this proposal will center on educational applications of user models, which will be called student or learner models. However, the resulting modelling and visualization approaches should generalize to other application domains, especially when the problems these fields face do not have correct or incorrect solutions. Rather the questions that are asked and the appropriateness of the solutions provided are open to interpretation.******The incredible amount of data produced through people's interactions with educational technologies can be extracted and used to infer information about students, teachers, and the learning environment. The output of the algorithms making these inferences can then be used to drive educational improvement through personalization. Unfortunately, this is not being done effectively for 2 reasons: ***1) most of the student models that have been created for areas where there is no such thing as a correct or incorrect answer are not grounded in theory, which means principled adaptations cannot be made to learning materials, learning processes, or the learning environment itself, and ***2) existing student models are not communicated in a way that supports stakeholder decision-making because they do not communicate the limitations of the underlying student models and they fail to communicate model information in a way that is understandable to stakeholders.******To mitigate these problems, this program of research synthesizes knowledge and methods from cognitive science with those from several areas of computer science (artificial intelligence, human-computer interaction, user modelling and personalization, and educational data mining) to develop theoretically-grounded computational models of people's knowledge, beliefs, skills, and interests. These new student models will provide more nuanced information in an understandable way so that people and software can appropriately adapt and personalize the support and activities that are provided to students. The use of these techniques should help people to learn more, enjoy learning, and make better decisions because these student models and their visualizations will help people to understand their learning, and in the case of teachers, how their students are learning.This use of evidence to drive educational improvement will be evaluated in real learning environments. Tools will be built so that they are compatible with major educational technologies and they will be freely released once they are shown to be effective.