Learning Analytics for Adaptive Systems supporting Self-regulated Social Learning

今天，在互联网和社交媒体上以各种形式促进学习的机会，从小学到中学后、工作场所和终身学习，以及在不同层面上都得到了广泛承认。然而，互联网上提供的信息和社交媒体上的学习活动，即使是通过正式的学术课程，主要也不是基于来自学习科学的经验证的建议。通常，学习者只能“自己”找出最适合他们的学习策略。正如学习科学研究报告所述，这是一个严重的威胁，因为大多数学习者的自我调节学习技能较弱。这可能会对他们的学习成功产生严重影响，带来许多社会后果，如高辍学率和劳动力对职业过渡缺乏准备。 自适应学习系统被认为是一种很有前途的方法，通过基于观察到的学习进度和特征的实时适应来增强(自我调节)学习。在现代系统中，适应集中在一个特定的学生群体上(例如，那些处于危险中的学生)，通常是通过测量学习者的进步与学习者应该具有的“平均”特征之间的距离来实现的。为了充分发挥社会适应性学习系统的潜力，应开发利用学习系统记录的大量跟踪数据的深刻方法和技术，以便优化不同学习亚群的学习。因此，本研究的目标是：(I)通过挖掘关于行动中学习的定量(例如，页面访问次数)和定性(例如，非结构化文本)众源数据，提出在社会背景下分析自我调节学习的方法；(Ii)提出通过考虑学习者的个体差异(例如，动机和目标定向)来设计能够激发更深层次学习(例如，对话、同伴讨论或自我测试)的适应性学习系统的方法。这些方法将被用来开发新的和扩展现有的(开源和商业)自适应系统，以支持自我调节的社会学习；以及(Iii)在学习者的实证研究中验证所提出的方法的有效性。

Opportunities to facilitate learning, in all forms from elementary schools to post-secondary, workplace, and life-long learning and at different levels, on the Internet and with social media are widely recognized today. However, information offered on the Internet and learning activities in social media, even via formal academic courses, are mainly not based on empirically-proven recommendations from learning sciences. Often, learners are left "on their own" to figure out which study tactics best work for them. This is a serious threat given that most learners have weak skills for self-regulated learning, as reported in learning science research. This can have severe effects on their learning success with numerous societal consequences such as high dropout rates and workforce unprepared for career transitions. Adaptive learning systems are seen as a promising approach to enhancing (self-regulated) learning through real-time adaptations based on observed progress and traits of learners. In contemporary systems, adaptations are focused on one specific group of students to assist (e.g., those at risk) typically by measuring the distance between a learner's progress and the traits expected that learners should have "on average." To unlock the full potential of social adaptive learning systems, profound methods and techniques for harnessing the wealth of trace data logged by learning systems should be developed in order to optimize learning for diverse learner subpopulations. Therefore, the objectives of this research are to (i) propose methods for analysis of self-regulated learning in social context by mining quantitative (e.g., page visit counts) and qualitative (e.g., unstructured text) crowd-sourced data about learning in action; (ii) propose methods for designing adaptive learning systems that motivate deeper learning (e.g., dialogue, peer-discussion, or self-testing) by considering learners' individual differences (e.g., motivation and goal-orientation). These methods will be used to develop new and extend existing (open source and commercial) adaptive systems supporting self-regulated social learning; and (iii) validate the effectiveness of the proposed methods in empirical studies with learners.