Lessons Learned from Scaling Up a Web-Based Intelligent Tutoring System

Client-based intelligent tutoring systems present challenges for content distribution, software updates, and research activity. With server-based intelligent tutoring systems, it is possible to easily distribute new and updated content, deploy new features and bug fixes, and allow researchers to more easily perform randomized, controlled studies with minimal client-side changes. Building a scalable system architecture that provides reliable service to students, teachers, and researchers is a challenge for server-based intelligent tutors. Our research team has built Assistment, a Web-based tutor used by hundreds of students every day in the Worcester and Pittsburgh areas. Scaling up a server-based intelligent tutoring system requires a particular focus on speed and reliability from the software and system developers. This paper discusses the evolution of our architecture and how it has reduced the cost of authoring ITS and improved the performance and reliability of the system.

[1]  Kenneth R. Koedinger,et al.  An architecture for plug-in tutor agents , 1996 .

[2]  Peter Brusilovsky,et al.  Adaptive and Intelligent Web-based Educational Systems , 2003, Int. J. Artif. Intell. Educ..

[3]  Daniel D. Suthers,et al.  Representational support for collaborative inquiry , 1999, Proceedings of the 32nd Annual Hawaii International Conference on Systems Sciences. 1999. HICSS-32. Abstracts and CD-ROM of Full Papers.

[4]  Vladimir Zadorozhny,et al.  A User Modeling Server for Contemporary Adaptive Hypermedia: An Evaluation of the Push Approach to Evidence Propagation , 2007, User Modeling.

[5]  Jim Kurose,et al.  Intelligence on the Web , 1997 .

[6]  B. Bacic.  Constructing intelligent tutoring systems: design guidelines , 2002, ITI 2002. Proceedings of the 24th International Conference on Information Technology Interfaces (IEEE Cat. No.02EX534).