Roschelle, J. (1990). Designing for conversations. In AAAI Symposium on Computer-Based Environments for Learning and Teaching, Stanford, CA, USA. April. American Association for Artificial Intelligence.

———

This paper examines the tole of computer-based representations in learning, distinguishing a dominant assumption in many analyses of representation use, epistemic fidelity, from an alternative, symbolic mediation. Epistemic fidelity refers to the quality of the denotational relationship between a computer display and the desired knowledge structures. A display with high epidemic fidelity is often thought to be a better learning tool because it should enable students to decode the denotational relationship and internalize the target knowledge. Symbolic mediation, in contrast, refers to the utility of a display as a resource for managing the uncertainty of meaning in conversations, particularly with respect to the construction of shared knowledge. When taken as a framing assumption for the design of learning technology, symbolic mediation can lead designers to construct external displays the will bridge the gap between commonsense and scientific interpretations of the world by providing an enriched physical situation to act in a talk about. I develop the contrast between these perspectives using the concrete example of a computer simulation of velocity and acceleration called the “Envisioning Machine”. I argue for a greather emphasis on designing mediational tools, and discuss guidelines for designing for conversations including minimalism, persistence, selective redundancy, direct manipulation for communication, and activity fidelity.