SBIR Phase II: Personalized Reading Instruction

这个SBIR第二阶段项目提出了发现数字方法来个性化阅读教学，使学生在阅读时理解更多，保留知识，并建立持久的技能。关于阅读的学术研究支持这样一种观点，即积极的阅读策略结合高质量的教学可以使学生受益。然而，教学通常不是个性化的，以满足特定学生的需求，即使当教育工作者与学生1：1工作，他们只能解释有限数量的信号，从学生，以帮助指导教学。该项目的目标是采取几个输入时，学生阅读数字和调查是否可以有效地创建和提供个性化的阅读指令，使学生得到额外的帮助时，他们的斗争和挑战时，他们可以自己成功。在美国，三分之二的学生阅读能力很差，他们在阅读时无法理解文章的大意。这些学生辍学的可能性是其他学生的四倍。批判性阅读的人在学校更成功，获得高质量的工作，并能够为扩大社会资源做出更多贡献。几十年来，研究人员和教育工作者一直试图解决“阅读鸿沟”，但直到现在才有技术使这成为可能。这个SBIR第二阶段项目提出使用独特的机器学习技术来个性化阅读教学。个性化教学的算法将确保额外的帮助或脚手架被分配给需要它的学生，当他们不再需要它或当它有可能成为拐杖时，它就会被移除。这种方法与其他机器学习算法不同，后者旨在最大限度地减少总体错误或最大限度地提高总体回报。然而，个性化阅读教学所需要的是不同的。算法必须学习帮助学生的程度，不是让他们在帮助下表现得更好，而是让他们在没有帮助的情况下表现得更好，因为目标是让学生成为更好的长期阅读者，而不是依赖脚手架阅读。该研究的目标是充分开发和商业化这个个性化的阅读系统，并将涉及数据科学，应用程序开发和内容创作。

This SBIR Phase II project proposes to discover digital methods to personalize reading instruction such that students understand more when they read, retain knowledge, and build lasting skills. The academic research on reading supports the claim that active reading strategies that incorporate quality instruction can benefit students. However, instruction is usually not personalized to meet the needs of specific students, and even when an educator works 1:1 with a student they can only interpret a limited number of signals from a student to help guide instruction. The objective of the project is to take in several inputs when students read digitally and investigate whether personalized reading instruction can effectively be created and delivered such that students get extra help when they struggle and are challenged when they can succeed on their own. Two-thirds of students in the U.S. are struggling readers; they cannot understand the main idea when they read. These students are four times more likely to drop out of school. People who read critically have more success in school, obtain high quality jobs, and are able to contribute more to expand social resources. Researchers and educators have been trying to solve the "reading gap" for decades, but only now does the technology exist to make this possible.This SBIR Phase II project proposes to use unique machine learning techniques to personalize reading instruction. The algorithms to personalize instruction will ensure that extra help, or scaffolding, is allocated to the students who need it, and removed when they no longer need it or when it threatens to become a crutch. This approach is different than other machine learning algorithms, which are built to minimize the overall error or maximize the overall reward. However, what is required for personalized reading instruction is different. The algorithm must learn how much to help a student not so they perform better with help, but so they perform better without it because the goal is for students to become better readers in the long term, not become reliant on scaffolding to read. The objective of the research is to fully develop and commercialize this personalized reading system and will involve data science, application development, and content authoring.