This project focused on an intelligent tool that adapts to students' individual approaches to collaboration across different types of digital platforms in ways that encouraged middle school mathematics learning. While others have worked towards supporting collaboration between students in a single digital context, the project is unique in how we create collaborative support across different platforms. It focuses specifically on help-giving as an essential collaborative skill that is broadly applicable. 

The project produced a middle school mathematics curriculum that included collaboration opportunities face-to-face and across three digital platforms. The curriculum was based on Modeling (TM) pedagogy, where students model real-world phenomena in small groups and then come together as a class to reflect. Three digital learning environments supported this curriculum. First, Modelbook was custom-built, an interactive digital textbook that facilitates students in posting pictures of their own work and commenting on their classmates' work. Second, students posted questions and answers to the Khan Academy (www.khanacademy.org) Q&A forum as their knowledge evolved. Third, students explained math problems to a teachable agent in natural language, and the agent responded with questions and explanations. The curriculum and learning environments have been used as part of in-person, fully remote, and hybrid classroom instruction.

One objective of this project was to use design methods to understand how students give help to each other across platforms and how best they can be supported. This project included three years of co-design, which consisted of 4-6 sessions with each year's cohort of approximately 15 students. The co-design helped build understanding of how relationships motivate help-giving, barriers to giving help, and what supports might be positively received (e.g., badges). The project also involved an initial design-based research study involving deployments of the system in three different units in a single class over the course of the school year. This study showed the importance of supporting multiple platforms for giving help, as some students were more comfotable giving help on one platform compared to another. Interviews revealed four motivational factors that influence students in giving help differently across different contexts: math self-concept, help-giving self-concept, familiarity with ones' peers, and conscientiousness. Based on the data collected, a logistic model was developed predicting whether students would give constructive help at a given opportunity based on platform features (e.g., synchronous vs asynchronous interaction) and students' motivation.

This project also included the development of a suite of tools that enabled adaptive support for help-giving. Noticing that surveys prior to the interaction did not always reflect students' motivation, an Interactive Persona Tool was designed to elicit factors related to students' motivation to collaborate using a narrative-based approach. Students could modify their personalized personas prior to engaging in a collaborative activity. Other adaptive supports consisted of sentence starters, chosen based on student motivation, that prompt students to participate constructively, and badges that reward positive help-giving behaviors, which were detected across the platforms using a combination of an off-the shelf classifier and custom keyword-based algorithm.

Two classroom studies were ran after our adaptive support was developed, a design study to examine student use of the evolving tools, and a final study where a control condition where the platforms were not used was compared to two versions of the digital tools (one where motivation was assessed based on the personas and one where motivation was assessed based on the surveys). While there were no learning differences between the three conditions, there was some evidence that the tool design contributed positively to students' experience during the intervention conditions. For example, motivation collected using the Interactive Persona Tool was more predictive of student participation than motivation collected using surveys. Sentence starter and badge use was predictive of constructive participation, controlling for pretest, although only during use of Modelbook. Future work in this area would further tune the sentence starter and badge support to individual and contextual factors.

Intellectual merits of this work consist of learning sciences contributions in terms of a deeper understanding of what motivates students to give help and how adaptive support might influence help-giving behaviors. There are computer science contributions with respect to the computational model of help-giving and design of the adaptive supports. There are methodological contributions with respect to the Persona-based motivation tool and through exploring how co-design can meaningfully engage students in research methods.

Broader impacts of the work include practical insights about how collaboration can be supported in classrooms for individual student needs and how digital tools can be employed in different types of learning ecosystems: in-person, remote, and hybrid. Over 150 students interacted with our learning environment or participated in its design over the life cycle of the project. Our curriculum is freely available for use and linked to via the American Modeling Teacher's Association, and we have designed and conducted professional learning workshops with approximately 20 teachers to support its use.

Last Modified: 05/19/2023
Modified by: Erin Walker