Visual Representations for Learning Modeling
Sponsor: Research on Learning and Education Program, National Science Foundation
Principal Investigator: Kenneth D. Forbus
Co-Principal Investigator: Bruce Sherin
- Karen Carney (Graduate Student, Education)
- Danielle Fisher (Undergraduate Student)
- Robert Harris (Programmer) [Alum]
- Thomas R. Hinrichs (Research Professor)
- Marisa Ketzler (Undergraduate)
- Leo C. Ureel II (Graduate Student / Software Engineer, Computer Science)
Project Summary: The goal of this project is to create a visual representation system with computer support that helps students learn how to articulate and reason with models of complex phenomena and systems. Learning how to create, test, and revise models is a central skill in scientific reasoning. There is ample experience suggesting graphical representations can provide a natural expression and communication medium for modeling ideas. Three families of graphical representations have been used in educational modeling environments (concept maps, dynamic system notations, and argumentation environments), each focused on different aspects of modeling, but none alone are sufficient for capturing the range of activities and knowledge involved in modeling. Moreover, none of them address three key issues in modeling:
- The importance of broadly applicable principles and processes
- Understanding when a model is relevant
- Qualitative understanding of behavior.
This project is using ideas from qualitative modeling to create a graphical notation, with computer support, that enables middle-school students to learn both particular areas of science and the process of modeling itself. The prototype modeling system we have created is called VModel. Students use VModel to create, analyze, modify, discuss, and disseminate their models. Ultimately, we think software like will be for modeling what word processors are to writing essays and spreadsheets are to mathematical analyses: A tool for creating, revising, and disseminating student ideas, creating an artifact that captures their evolving understanding and that can serve as a focus for discussion.
VModel uses a “construction kit” metaphor, enabling students to express their knowledge in an ever-expanding collection of principles and processes, thus emphasizing the underlying unity and systematic structure of scientific knowledge. VModel provides multiple levels of computer support, ranging from simple drawing and publishing tools to a coach that uses qualitative simulation to help students see whether or not their model supports their prediction. As many urban schools have outdated or minimal computer equipment, we have taken care to keep VModel lightweight, so that it does not require the latest computer hardware to run. VModel has been used in a number of experiments in Chicago Public School classrooms, with encouraging results. Some evidence that suggests VModel is functioning as intended include:
- Students can build qualitative models of interesting phenomena.
- Building models promotes transfer, as measured by re-use of prior models.
- Students use terms from our graphical language for physical phenomena even when
they are not using the software.
- Teachers report shifts in the kinds of student thinking with repeated use of the
software: More analysis of systems, and explaining behavior in terms of mechanisms.
This project interacts synergistically with the work we are doing in the NSF Center for Learning Technologies in Urban Schools (LeTUS). A key focus of the Center is creating and deploying an alternate middle-school science curriculum in the Chicago and Detroit public school systems that is inquiry-based and supports national standards, through the medium of work circles, collaborative arrangements involving researchers and teachers that support these efforts. We have operated two work circles to develop curricula that have been used to explore these ideas. One of these curricula, on solar houses, has been rolled out on a wider basis through Chicago schools via LeTUS. (This curriculum uses another of our technologies, self-explanatory simulators, to provide simulation laboratories that complement physical hands-on experiments.) The teachers who collaborated with us in these efforts have made a variety of valuable contributions to the project, and we would like to thank them here: Adam Dorr, Kiesha Korman, Beverly Miller, Carlos Rodrigues, Deborah Rogers-Green, Carol Scafide and Judy Whitcomb.
We are currently focusing our efforts in two directions. First, we are analyzing the data gathered in our classroom work and writing up appropriate publications. (For a selected list of publications click here.) Second, we are implementing more sophisticated coaching in VModel, off-loading more intense reasoning to an email-based server architecture to keep the classroom client software lightweight. (This RoboTA architecture was developed in our prior NSF-sponsored work on articulate virtual laboratories.) Using a combination of analogical and qualitative reasoning, the new software coaches will compare student models to normative models, point out discrepancies between the predictions of a student model and simulator behavior, and prepare assessment reports for teachers summarizing class progress and patterns of misconceptions. At the end of the project, all software and documentation will be made available as open source. We are already making the current version of VModel publicly available, so that interested teachers and researchers can use it.