SBIR Phase I: Teachpad - A Teach-to-Learn Paradigm for Difficult Math Concepts

这个小企业创新研究（SBIR）第一阶段项目的更广泛的影响/商业潜力是利用当前的学习科学研究来开发游戏，使中学数学具有吸引力和相关性。中学是学生教育历程中的关键时期，是学生能力自我概念趋于稳定的时期。研究表明，这也是学生开始对数学失去兴趣的时期，从六年级到八年级，对数学的负面情绪明显增加。根据国家报告卡，在美国公立学校的360多万名学生中，每年只有大约三分之一的八年级学生数学达到熟练或以上水平。该提案的目标是实施和测试一个名为“Teachpad”的脚手架式学习和教学环境的有效性，该环境可以帮助中学生学习困难的数学概念。多项研究表明，当学生不得不教别人时，他们会更努力地学习。该项目将扩展这些研究，并评估教学手册是否能提高学生的参与度，改善学习和长期记忆困难的数学概念。该计划采用“教与学”的框架，学生教电脑游戏中的角色如何解决特定问题，并利用他们从角色那里得到的反馈来评估他们自己对概念的理解程度。教学板使用自然语言输入，这意味着学生可以用简单的英语输入他们的指导，并且可以像对朋友或兄弟姐妹一样解释问题的解决方案。语义解析器从这些输入中提取有用的信息，并将其提供给机器学习算法，以教游戏角色如何解决问题。当角色试图解决一个测试问题时，学生也可以看到他们的指导是如何起作用的。该项目将实施操作顺序的教学手册，并将确定其与传统课堂实践的比较。数据将在真实的教育环境中收集，通过观察和访谈来研究学生的参与度和可用性，以及前测和后测（一次在治疗后，一次在学生从暑假回来时）来评估学习成果和长期保留。

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is to use current research in the learning sciences to develop games that make middle school math engaging and relevant. Middle school is a crucial time in a student's educational journey as it is the period at which their self-concept of ability becomes stable. Studies show that it is also the period at which students start getting disengaged with math and there is a marked increase in negativity towards math from sixth to eighth grade. According to the National Report Card, only about a third of 8th graders in US public schools, of over 3.6 million students, score at the proficient or above level in Math every year. The goal of this proposal is to implement and test the effectiveness of a scaffolded learning and teaching environment, called the 'Teachpad', that helps middle school students learn difficult math concepts. Multiple studies have indicated that students put more effort into learning when they have to teach someone else. This project will extend these studies and evaluate whether the teachpad increases student engagement and improves learning and long term retention of difficult math concepts.The proposed project uses a "teach to learn" framework in which students teach characters in a computer game how to solve specific problems and use the feedback that they receive from the characters to evaluate how well they have understood the concepts themselves. The teachpad uses natural language input, meaning that the students can enter their instruction in plain English and can explain the solution to the problem as they would to a friend or sibling. A semantic parser extracts the useful information from this input and feeds it to a machine learning algorithm to teach a game character how to solve the problem. A model of what the character has learned is also visible to the student who can see how their instruction is working when the character is trying to solve a test problem. This project will implement the teachpad for the order of operations, and will determine how it compares to traditional classroom practices. Data will be collected in authentic educational environments, through observations and interviews to study student engagement and usability, and pretests and posttests (once after the treatment and once when students return from summer break) to evaluate learning outcomes and long term retention.