Articulate Software for Teaching Science and Engineering
Sponsor: DARPA Computer-Aided Education and Training
Initiative (CAETI)
Principal Investigator: Kenneth
D. Forbus
Project Summary: We believe that a synthesis of Artificial Intelligence techniques can enable the construction of new software architectures for educational software for science and engineering that dramatically improves student learning and instructor productivity. The pedagogical approach embodied in these architectures is learning by doing, specifically, learning science via virtual experiments and via modelling.
Learning Science via Virtual Experiments: The power of illustrative examples is well-known in education. Traditional media offer high authenticity but low interactivity. Textbook illustrations and posters can provide thought-provoking pictures, tables, charts, and other depictions of complex information. Movies and video can provide gripping dynamical displays. But none of these media provide interaction. A student intrigued by a picture of a steam engine in a textbook (or a movie of a steam engine) cannot vary the load or change the working fluid to see what will happen. They cannot ask for more details about explanations that they don't understand. They cannot satisfy their curiosity about how efficiency varies with operating temperaturesby testing the engine over range of values. Software that lets students experiment with simulations of physical systems that provides conceptual explanations as well as numerical results should enable students to learn more than from traditional simulation-based activities.
Learning Science via Modeling: Modeling, i.e., the application of principles to understanding phenomena, is at the heart of science. Students often bring misconceptions to the classroom that can interfere with learning. We believe that software that helps students articulate, test, and refine models can reduce such persistent misconceptions. Coaching students informing physically accurate mental models and in formulating mathematical models should help students to better understand both the phenomena being modelled and the process of science.
We are building two software architectures embodying these ideas, with prototypes and software tools that will enable curriculum developers and instructors to extend them:
- Active Illustrations help students learn
science by enabling them to experiment with simulated phenomena and systems,
receiving feedback in the form of conceptual explanations as well
as simulated behaviors and numerical values. The software will be
able to explain the consequences of student's experiments in
intuitive, qualitative terms, illustrating how the specific
responses are ultimately governed by general laws and principles.
Active Illustrations can be used as stand-alone learning tools,
as a medium to include in hypermedia systems, and as virtual
artifacts in shared virtual learning environments.
- Modeling Workbenches help students learn science by building, testing, and refining models. Given a phenomena or system, the software will help a student articulate (or build) a model. It will help test this model against the original example, and suggest new examples to challenge it. It will coach students in revising, extending, and refining their models.
Back to Structure-Mapping: A computational model of analogy and similarity page | Back to Self-Explanatory Simulators page | Back to Qualitative Process Theory page | Back to MAC/FAC: A model of similarity-based retrieval page | Back to Previous Projects page | Back to QRG Home Page