Learning Depends on Knowledge: Using Interaction Designs and Machine Learning to Contrast the Testing and Worked Example Effects

学习取决于知识：使用交互设计和机器学习来对比测试和工作示例的效果

• 批准号: 1824257
• 负责人: Ken Koedinger
• 金额: $ 71.24万
• 依托单位: Carnegie-Mellon University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2018
• 资助国家: 美国
• 起止时间: 2018-08-15 至 2022-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1824257&HistoricalAwards=false
• 关键词: Learning; Depends; Knowledge; Using; Interaction

A distinctive characteristic of human learning is its capability to flexibly acquire a wide range of rich and complex forms of knowledge (e.g., first and second languages) and acquiring new and accumulated knowledge (e.g., learning physics is easier after having learned algebra). An adequate explanation of human learning must address how existing knowledge changes the way we learn so that we achieve knowledge goals and in specific contexts. This project aims to discover and specify how human learning processes operate differently under different contexts, depending upon what content is being learned. This research contributes to improved educational practices by specifying how learning processes are influenced by knowledge acquisition in a systematic and replicable way. This research will enhance our understanding of successful learning and optimal performance.The project explores learning processes by contrasting learning from retrieval practice and learning from studying examples. The goal of this project is to resolve and clarify how these processes compete for cognitive resources, including attention and working memory, in ways that depend on the knowledge content to be learned. This research examines (1) the learning processes involved in learning from retrieval practice and from worked examples, (2) how these learning processes work differently when applied to different knowledge content, and (3) the computational mechanisms of learning that give rise to learning different content. The researchers use a combination of experiments in which the learning approach is varied along with the materials being studied and machine learning models. It will demonstrate knowledge-learning dependence by showing that one learning process (e.g., retrieval practice) produces better learning outcomes than another (e.g., example encoding) in some knowledge contexts but the reverse occurs in other knowledge contexts. By implementing the studies as part of a machine learning architecture, this research will provide computational evidence and theoretical insight into the hypothesized knowledge-learning dependence framework. The machine learning architecture developed may be used as an educational and research tool in learning sciences, and the project involves training new learning scientists.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.

人类学习的一个显著特征是它能够灵活地获得广泛的丰富而复杂的知识形式（例如，第一和第二语言）和获取新的和积累的知识（例如，学过代数后学物理更容易）。对人类学习的充分解释必须解决现有知识如何改变我们学习的方式，以便我们在特定的背景下实现知识目标。该项目旨在发现和说明人类学习过程如何在不同的环境下以不同的方式运作，这取决于学习的内容。这项研究有助于改善教育实践，具体说明如何学习过程中的影响，以系统和可复制的方式获得知识。本研究通过对比从检索实践中学习和从学习实例中学习来探讨学习过程。这个项目的目标是解决和澄清这些过程如何竞争认知资源，包括注意力和工作记忆，在依赖于知识内容的方式来学习。本研究探讨了（1）从检索实践和样例中学习所涉及的学习过程，（2）这些学习过程在应用于不同的知识内容时如何不同地工作，以及（3）引起学习不同内容的学习的计算机制。研究人员使用了一系列实验，其中学习方法随着所研究的材料和机器学习模型的变化而沿着变化。它将通过显示一个学习过程（例如，检索实践）产生比另一个更好的学习结果（例如，示例编码），但在其他知识上下文中发生相反的情况。通过将这些研究作为机器学习架构的一部分来实施，本研究将为假设的知识学习依赖框架提供计算证据和理论见解。开发的机器学习架构可用作学习科学的教育和研究工具，该项目涉及培训新的学习科学家。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

Investigating Differential Error Types Between Human and Simulated Learners

• DOI: 10.1007/978-3-030-52237-7_47
• 发表时间: 2020-06-09
• 期刊: Artificial Intelligence in Education
• 作者: Weitekamp D;Ye Z;Rachatasumrit N;Harpstead E;Koedinger K
• 通讯作者: Koedinger K

Self-regulated spacing in a massive open online course is related to better learning

• DOI: 10.1038/s41539-020-0061-1
• 发表时间: 2020-01-01
• 期刊: NPJ SCIENCE OF LEARNING
• 影响因子: 4.2
• 作者: Carvalho, Paulo F.;Sana, Faria;Yan, Veronica X.
• 通讯作者: Yan, Veronica X.

Square it up!: How to model step duration when predicting student performance

平方起来！：在预测学生表现时如何对步骤持续时间进行建模

• DOI: 10.1145/3303772.3303827
• 发表时间: 2019
• 期刊: Proceedings of the 9th International Conference on Learning Analytics & Knowledge
• 作者: Chounta, Irene-Angelica;Carvalho, Paulo F.
• 通讯作者: Carvalho, Paulo F.

Drinking Our Own Champagne: Analyzing the Impact of Learning-by-doing Resources in an E-learning Course

喝我们自己的香槟：分析电子学习课程中边做边学资源的影响

• 发表时间: 2021
• 期刊: Companion Proceedings of the 11thInternational Conference on Learning Analytics & Knowledge LAK20
• 作者: Hou, Xinying;Carvalho, Paulo F.;Koedinger, Kenneth R.
• 通讯作者: Koedinger, Kenneth R.

Toward Near Zero-Parameter Prediction Using a Computational Model of Student Learning

使用学生学习计算模型进行近零参数预测

• 发表时间: 2019
• 期刊: Proceedings of The 12th International Conference on Educational Data Mining (EDM 2019
• 作者: Weitekamp, Daniel;Harpstead, Erik;Rachatasumrit, Napol;Maclellan, Christopher;Koedinger, Kenneth R.
• 通讯作者: Koedinger, Kenneth R.