| Project | Articulate Software for Science Education |
| Contact | Kenneth D. Forbus |
| forbus@northwestern.edu | |
| URL | http://www.qrg.northwestern.edu/ |
| Project description | We are developing several kinds of articulate software, by which we mean software that has an understanding of the domain that it is about and can explain that understanding, for education at different levels.We have two efforts that involve using design tasks to teach fundamentals, using what we call articulate virtual laboratories. Our laboratory for engineering thermodynamics, CyclePad, is used at a variety of universities and schools, and we get a couple of downloads a day from new potential users, on the average. Our laboratory for teaching feedback is being prototyped with high-school students right now, and we hope to have activities ready for use in the Chicago Public School system next year, as part of the NSF Center here. These efforts are supported by NSF.We have also developed a new kind of simulation technology, self-explanatory simulators, that can be used to create new kinds of hypermedia and simulation environments. We are currently exploring how they might be used in middle-school science activities as well as training simulators.Details on these projects, and prototypes, can be found at http://www.qrg.ils.nwu.edu. |
| Theoretical background | Articulate software requires a combination of AI techniques, including constraint reasoning, truth-maintenance, and several others. The most important component is qualitative reasoning, which provides new conceptual and software tools for formalizing the modeling process itself. The educational software described above are made possible my group's work on qualitative physics, spanning well over a decade. In articulate virtual labs, the consequences of a student's assumptions, including their modeling assumptions, are made available for their inspection via a dynamic hypertext. This method of visualizing the results of their modeling decisions can be very effective.In self-explanatory simulators, the input to the simulation "compiler" is a high-level physical description. The compiler does the physical modeling, selects appropriate mathematical models, and writes the code. This analysis by the compiler is then boiled down into a compact explanation system that is part of the simulator. So while it takes some subtle reasoning to create a simulator (as might be expected), the simulators themselves can give good explanations without extensive runtime resources. For example, we have run them on HP200 palmtops, and as Java applets.The reason why qualitative physics will prove to be so useful in educational software, we believe, is that people appear to do substantial portions of their thinking about the physical world in these terms. For expressing theories involving continuous properties to students who do not yet know algebra, for instance, qualitative mathematics is the only propspect. |
| Challenges | We are addressing several issues right now:1. What sorts of design activities and curricula maximize the learning from articulate virtual laboratories? With CyclePad we are, in a sense, doing a guerilla operation: CyclePad is starting to have a calculator-like effect on thermodynamics instruction, where the traditional homeworks become "too easy". We are working with instructors at several institutions, including the US Naval Academy and others, to help develop new design-oriented activities that still fit within an engineering curriculum. With the Feedback AVL, since feedback is not currently taught in most high schools, one of the challenges we are thinking about is integrating it with other material being taught, so that the deep connections and explanations that feedback provides can help integrate different aspects of what students are learning.2. With self-explanatory simulators, there are three challenges that are absorbing us right now. First, we are trying to hammer out the technology in a way that makes them a transparent, easy-to-integrate medium within the web. [Anyone who is using Java heavily knows what a pain this is, given the current state of the technology.] Second, we are creating an infrastructure to make it easy for curriculum developers to use our compiler to generate self-explanatory simulators for themselves, including a variety of runtime shells they can customize, off-the-shelf domain theories, and an authoring environment. Third, we are developing sample curricula and activities so we can help instructors and curriculum/activity designers integrate the technology into what they are doing. |
| Partnership | We are looking for stronger ties to classrooms and curriculum designers. |