Research IssuesWe focus on the basic research activities critical to expanding our systems from research vehicles to powerful, integrated, widely disseminate instructional systems. Some of the critical issues addressed are:
1) Breaking Up Today's Monolithic Tutor ArchitectureIt is important to formalize tutor architectures so that the components can be separated, thus allowing economies gained by component reuse. Why should every tutor developer build a component, for example, a student model, from scratch when many serve the same purpose within tutors?
2) Tutor InteroperabilitySimilarly, we need to develop ways for one tutor to talk to another. Currently, remediation activities are built into each tutor on as as-needed basis. For modularity and ease of development, we would like to be able to make a call to another tutor to handle many kinds of remediation. For example, if a student is working with our Injection Molding Tutor and reaches a point where the tutor suggests revisiting some of the basics of thermodynamics, we would want the tutor to call a separate thermodynamics tutor, passing some state information stored in the student model that guides the thermodynamics session (e.g. a reasonable starting point given the student's misconception about thermodynamics). Once the misconception has been cleared up, the student is returned to the Injection Molding Tutor and continues. Such interoperability requires that some interesting issues in Tutor Knowledge Representation and Distributed Student Modeling (among others) be addressed.
3) Network Delivery of TutorsDelivering tutors over networks (e.g. WWW) allows widespread distribution in such a way that updates can be made quickly and multimedia materials can be efficiently stored and maintained. Furthermore, students are freed from having to invoke a tutor on a specific computer that stores his or her student model, and logs into a server instead which runs the tutor in a client/server mode; the student model is maintained on the server side. Before this is possible, issues involved in maintaining a distributed student model must be resolved (some of which overlap with those in 1).
4) Building ToolsTutors are difficult and time-consuming to build. Tools are needed to decrease development time and complexity. Many of the tools that exist can be improved (as we are now doing), and addressing the issues above will introduce a new set of tool requirements that we will explore.
5) Making Tutors SmarterMuch research needs to be done to expand the roles of simulation, discovery learning, 3-D visualization and multimedia in intelligent tutors. Additionally, there is still much basic research needed in AI to make tutors more intelligent. For example, our Protein Synthesis Tutor raises a number of Knowledge Representation issues about recognition of student intentions and understanding as they manipulate graphical objects (e.g. molecules) that have complicated spatial and temporal relationships.
6) Better Evaluation Methods
While some evaluation of these systems has been done by us and by others, no systematic attempt has been made to identify the gains in student performance that can be attributed to these systems. Part of our current research plan focuses on a systematic evaluation program.
To learn more about some of the research projects in which Beverly Park Woolf has been a major contributor, use the navigation menu to the right.