SBIR Phase I: Empowering Educators with AI During Distance Learning (COVID-19)

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，它将提供新的工具来评估和加强高阶批判性思维的教学。 该项目将利用先进的机器学习模型为学生阅读的任何文本生成批判性思维问题，分析他们的书面回答，为他们提供关于如何完善思维的即时反馈，并最终为他们的老师提供数据驱动的见解。使用人工智能（AI）自动解释开放式响应的能力是教育界的一个丰富领域。 该项目将加深教育界在这一领域的知识，并将研究结果应用于K-12教育。特别是，研究结果可能会导致一个新的评价架构，减少对多项选择问题和相关技术的依赖，通过一种有效评价和提供反馈的方法来加强教育。这个项目将解决使用多项选择评估来评估“学习”的问题，并转向更准确的方法来确定学习的细微差别。这个小企业创新研究（SBIR）第一阶段项目的重点是自动简答题评分，以解决一个紧迫的和基本上未解决的问题。自动评分的大多数工作都集中在较长的论文评分上，这与高等教育更相关。目前，还没有适用于短响应的通用算法。增加技术复杂性的是，该项目的机器学习方法需要适用于任何书籍或文本，并包括任何用户编写的问题。该项目将提高评估的准确性和细节，特别是对读者新文本的书面答复。该项目将需要深度学习自然语言处理（NLP）技术来完全建模语言表示。所提出的系统将在评估书面答复之前摄取主题文本和相关问题，而无需将先前提交的内容用作训练数据。该项目的原型将开发英语语言艺术教学之前，更广泛地部署在其他学科。这个奖项反映了NSF的法定使命，并已被认为是值得通过评估使用基金会的智力价值和更广泛的影响审查标准的支持。

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is that it will provide new tools to evaluate and enhance teaching of higher-order critical thinking. This project will leverage advanced machine learning models to generate critical-thinking questions for any text a student reads, analyze their written response, give them immediate feedback on how to refine their thinking, and ultimately provide data-driven insights to their teachers. The capability to automatically interpret open-ended responses using artificial intelligence (AI) is a rich area for the educational community. This project will deepen the education community's knowledge in this area and apply the findings to K-12 education. In particular, the results may lead to a new evaluation architecture relying less on multiple-choice questions and related techniques, enhancing education with a method to efficiently evaluate and provide feedback with open-ended and short-response questions. This project will address the problem of using multiple-choice assessments to assess ‘learning’ and move to more accurate ways to ascertain the nuances of learning.This Small Business Innovation Research (SBIR) Phase I project focuses on automated short-answer scoring to solve a pressing and largely unsolved problem. Most work in automated scoring focuses on longer essay grading, which is more relevant for higher education. At this time, there are no general-purpose algorithms available for short responses. Adding to the technical complexity is the fact that this project’s machine learning approach needs to work for any book or text, and include questions written by any user. This project will improve the accuracy and detail of assessments, particularly with written responses regarding texts new to the reader. This project will require deep-learning natural language processing (NLP) technology to fully model language representation. The proposed system will ingest the subject text and the associated question before evaluating the written response without prior submissions used as training data. The project prototype will be developed for English Language Arts instruction prior to broader deployment across other disciplines.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.