Self- and Co-Regulating Anthropomorphic Pedagogical Agents

想象一下，一个软件代理可以根据你自己的习惯和优势，以及成千上万其他成功学生的学习分析模式，学习如何为你制定个性化的教育和培训策略。一个值得信赖的朋友、伙伴和盟友，可以帮助你确定目标，并根据你自己的学习风格，与你自己最深刻的价值观、个人、社会和其他标准保持一致，规划出实现这些目标的最可能的途径。这项研究的目的是建立和测试一套这样的智能代理，称为拟人化代理，模仿人类的特征，并帮助学习者完成他们的监管任务。此外，这些代理人自我调节自己的拟人化特征，如主动性和主题能力，以激发高度的信任，共享观点和实体-从而增加学习者将他们视为教育过程中称职的合作伙伴的意愿。研究表明，混合主动性的激励策略可以用来支撑学生的情感、认知和能力体验，从而提高监管的成功率。这是通过规避非生产性精神状态和/或维持更有利于执行监管任务的状态来实现的。其他调解行为包括协助拟订目标和查明长处和短处，挖掘数据以评价目标进展情况，以及选择最佳战略和适应计划，这些战略和计划以最可能激发所需的改变和调整的首选格式和方式提供。目前的监管理论并没有模拟拟人化代理人的影响。这项研究的结果将提供新的信息，填补现有知识体系中的这一空白。阻止人们享受监管带来的好处的真正原因包括低工作记忆容量、压力或监管资源枯竭、执行功能障碍（如多动症）、糟糕的时间管理技能、个性和气质，这些都可以通过拟人软件代理有效地解决。

Imagine a software agent that learns how to develop personalized education and training strategies for you based on your own habits and strengths, and based on the learning analytics patterns of thousands of other successful students that have come before. A trusted friend, partner, and ally that can help identify goals, and map out the most likely path to achieve those goals, based on your own learning styles, consistent with your own deepest values, personal, social, and other standards. This research aims to build and test a suite of such intelligent agents, called anthropomorphic agents, that mimic human-like traits and will assist learners in their regulatory tasks. Further these agents self-regulate their own anthropomorphic traits, such as initiative and subject matter competency, in order to inspire a high level of trust, shared perspective, and entitativity - thus increasing the learners' willingness to view them as competent partners in the educational process.Research shows that motivational strategies through mixed-initiatives can be used to scaffold students'emotional, cognitive, and competence experiences so as to increase the success of regulation. This is accomplished by circumventing unproductive mental states and/or maintaining states that are more conducive for carrying out regulation tasks. Other mediative acts include assistance with goal formulation and identifying strengths and weaknesses, data-mining to evaluate goal-progress as well as selection of optimal strategies and plans for adaptation proffered in a preferred format and manner most likely to inspire the required changes and adjustments. Current theories of regulation do not model the impact of anthropomorphic agents. The results of this research will provide new information that fills this gap in the existing body of knowledge. There are real reasons that prevent people from enjoying the benefits of regulation that include low working memory capacity, stress or depletion of regulatory resources, disorders of executive function (e.g., ADHD), poor time management skills, personality and temperament, which can be effectively addressed with anthropomorphic software agents.