Feedback for Metacognitive Support in Learning by Teaching Environments
Feedback for Metacognitive Support in Learning by Teaching Environments
复制标题
DOI：
--
复制
发表时间：
2006
影响因子：
1.8
通讯作者：
Gautam Biswas;Jason Tan;Daniel L. Schwartz
中科院分区：
文献类型：
--
作者：
Gautam Biswas;Jason Tan;Daniel L. Schwartz
教学环境对元认知支持的反馈Jason Tan和Gautam Biswas（jason.tan，gautam. vanderbilt.edu）范德比尔特大学纳什维尔EECS和ISIS系，美国田纳西州37235丹尼尔L。Schwartz（丹尼尔. stanford.edu）教育学院，斯坦福大学，斯坦福大学，CA 94305 USA近似分析以估计近似答案。远远不一致的错误应该触发调试活动来解决差异。有证据表明，帮助孩子学会监控别人解决问题的能力，反过来也能帮助他们监控自己解决问题的能力和学习能力。例如，Palinscar和Brown（1986）发现，对监控的高度重视提高了学生的阅读和学习能力。同样地，监视别人的工作应该促使孩子们运行他们自己的过程来生成解决方案，然后将其与其他人的解决方案进行比较。这意味着学生不需要同时做这两个过程。这减少了认知负荷，发展了比较解决方案的意识和能力，随着时间的推移，更容易将这种能力转向内部。我们提出的研究利用这一假设使用可教代理（TA），这是软件环境，学生教计算机代理使用结构良好的视觉表示。助教根据所学的事实和关系进行推理，以回答问题和解决问题。使用他们的代理的表现作为一种动力，学生的工作，以弥补代理的知识，并学习自己更好。我们的一个助教，被称为贝蒂的大脑，已经成功地用于教河流生态系统在5年级的科学教室（比斯瓦斯等，2005年a，2005年b）。贝蒂的大脑的一个重要特性是，学生可以监控贝蒂如何回答问题，并在她犯错时纠正她（和他们自己）。然而，五年级学生在教学实践中既是领域新手又是新手，他们往往不具备必要的监控技能，他们往往无法分析知识中的错误和遗漏。这促使我们在助教环境中开发元认知线索作为外显反馈机制，以帮助学生发展监控能力。本文讨论了这些反馈机制在帮助学生监控和学习任务中的有效性。在一个五年级的科学教室里进行的实验研究的结果从学生的即时学习能力和他们对未来学习的准备方面进行了讨论（Schwartz和Martin，2004）。摘要过去对基于计算机的学习环境中的反馈的研究表明，纠正性反馈有助于立即学习，而指导性和元认知反馈有助于获得深入理解和发展知识转移能力。在发现学习环境中，反馈变得很重要，在这种环境中，新手学生经常被与学习和组织新知识相关的认知负荷所压倒，同时监控他们自己的学习进度。我们专注于反馈机制在可教代理系统，以帮助提高学生的能力，以监测他们的代理的知识，并在这个过程中，他们自己的学习和理解。我们的研究表明，有效的指导元认知反馈，为学生未来的学习做好准备。介绍元认知已被确定为一个关键的过程，支持学生的学习和解决问题（布朗-福特，布朗和科金，2000年）。Brown（1987）描述了两个组成过程：（i）监控一个人的认知活动的能力，以及（ii）当发现问题时采取适当的调节步骤的能力。这些步骤可以包括内部调节（例如，当读取硬材料时减速）和外部作用（例如，学习资源）。这两种能力都随着成熟而增加（Flavell，1987），但适当的教育机会可以促进元认知的发展并改善随后的学习。我们专注于开发学习环境，提供元认知的支持，并检查元认知干预是否提高学生的后续学习能力。自我监控（cf.自我解释（Chi，et al.，1994））是一种关键的元认知策略，它支持理解学习，以及将所学知识应用于问题解决任务的能力。然而，自我监控本身就是一项复杂的认知任务。在问题解决的背景下，它需要两个同时协调的“过程”：一个是开发解决问题的步骤序列，第二个是评估问题解决过程的正确性和效率。分析差异并进行纠正会进一步增加自我监控任务的复杂性。例如，当解决数学问题时，理想情况下，运行一个计算精确答案的系统程序，并使用概念图表示来教授第二个过程，该过程快速且准确地执行贝蒂的大脑贝蒂，如图1所示。学生们教她关于实体的知识，比如鱼和藻类，以及它们之间的关系，（例如，鱼类消耗溶解氧，
Feedback for Metacognitive Support in Learning by Teaching Environments Jason Tan and Gautam Biswas (jason.tan, gautam.biswas@vanderbilt.edu) Department of EECS &ISIS, Vanderbilt University Nashville, TN 37235 USA Daniel L. Schwartz (daniel.schwartz@stanford.edu) School of Education, Stanford University Stanford, CA 94305 USA proximate analysis to estimate an approximate answer. An- swers that are far out of alignment should trigger debugging activities to resolve the discrepancy. There is evidence that helping children learn to monitor others problem solving can, in turn, help them monitor their own problem solving and learning. For example, Palinscar and Brown (1986) found that a strong emphasis on monitor- ing improved students’ reading and learning abilities. Ide- ally, monitoring someone else’s work should prompt the children to run their own process to generate a solution, and then compare it against the other person’s solution. This means that students need not do both processes simultane- ously. This reduces cognitive load, develops the awareness and capacity to compare solutions, and with time makes it easier to turn this capacity inward. Our proposed research leverages this hypothesis using Teachable Agents (TAs), which are software environments where students teach a computer agent using well-structured visual representations. The TA reasons with the facts and re- lations it has been taught to answer questions and solve prob- lems. Using their agent’s performance as a motivation, stu- dents work to remediate the agent’s knowledge, and learn better on their own. One of our TAs, called Betty’s Brain, has been successfully used to teach river ecosystems in 5 th grade science classrooms (Biswas, et al. 2005a, 2005b). An important property of Betty’s Brain is that students monitor how Betty answers questions and can correct her (and themselves) when she makes mistakes. However, 5 th grade students, who are both domain novices and novices in teaching practices, often do not possess the necessary moni- toring skills, and they often fail to analyze relevant pointers to errors and omissions in their knowledge. This has led us to develop metacognitive cues as explicit feedback mecha- nisms within the TA environment to help students develop the monitoring abilities. This paper discusses the effective- ness of these feedback mechanisms in aiding the students monitoring and learning tasks. The results of an experimen- tal study in a 5 th grade science classroom are discussed in terms of the students’ immediate learning abilities and their preparation for future learning (Schwartz and Martin, 2004). Abstract Past research on feedback in computer-based learning envi- ronments has shown that corrective feedback helps immediate learning, whereas guided and metacognitive feedback help in gaining deep understanding and developing the ability to transfer knowledge. Feedback becomes important in discov- ery learning environments, where novice students are often overwhelmed by the cognitive load associated with learning and organizing new knowledge while monitoring their own learning progress. We focus on feedback mechanisms in teachable agent systems to help improve students’ abilities to monitor their agent’s knowledge, and, in the process their own learning and understanding. Our studies demonstrate the effectiveness of guided metacognitive feedback in preparing students for future learning. Introduction Metacognition has been identified as a critical process that supports student learning and problem solving (Brans- ford, Brown & Cocking, 2000). Brown (1987) describes two component processes: (i) the ability to monitor one’s cogni- tive activities, and (ii) the ability to take appropriate regula- tory steps when a problem has been detected. These steps can include internal regulation (e.g., slow down when read- ing hard material) and external action (e.g., consult learning resources). Both abilities increase with maturation (Flavell, 1987), but appropriate educational opportunities can propel metacognitive development and improve subsequent learn- ing. We focus on developing learning environments that provide metacognitive support and examining whether metacognitive interventions improve students’ subsequent abilities to learn. Self-monitoring (cf. to self-explanation (Chi, et al., 1994)) is a key metacognitive strategy that supports learning with understanding, and the ability to apply the learnt knowledge to problem solving tasks. However, self-monitoring is itself a complex cognitive task. In the context of problem solv- ing, it requires two simultaneous coordinated “processes”: one that develops a sequence of steps to solve the problem, and a second that evaluates the correctness and efficiency of the problem solving process. Analyzing discrepancies and making corrections adds further complexity to the self- monitoring task. For example, when solving math problems, one ideally runs a systematic procedure that computes a pre- cise answer, and a second process that does a quick and ap- Betty’s Brain Betty, shown in Fig. 1, is taught using a concept map repre- sentation. Students teach her about entities, such as fish and algae, and their relations, (e.g., fish consume dissolved oxy-