Self-explanatory simulators: making computers partners in the modeling process

Digital computers have dramatically aided the application of mathematical models to problems of all kinds. For the most part computers have served as calculating engines, data retrieval and sorting devices, and lately as a powerful medium for visualization. Rarely have computers been used to assist in the more conceptual stages of investigations, for example in formulating the models to be analyzed, or in helping to interpret their results. Qualitative physics provides exactly the kinds of representations and reasoning techniques that can allow computers to shoulder more of the burden of creating and interpreting simulation models. This paper outlines the idea of self-explanatory simulations, an integration of qualitative and quantitative techniques, which provides one example of how qualitative physics can do this. A self-explanatory simulator combines the precision and efficiency of numerical simulation with the flexibility and interpretive power of qualitative representations. This synthesis makes them easier to use, and makes their results easier to understand. Moreover, such simulators can be written automatically from a physical description of the situation to be simulated, by using the qualitative analysis of the situation to guide the application of the appropriate quantitative and numerical models. Consequently, computers can become more like partners in the modeling process. This paper describes the basic ideas of self explanatory simulators, the results we have achieved so far, and our plans to push this technology towards applications.