International Journal of Artificial Intelligence in Education 13 (2003) 156–169IOS Press1560-4292/03/$8.00 © 2003 – IOS Press. All rights reservedAdaptive and Intelligent Web-based Educational SystemsPeter Brusilovsky, School of Information Sciences, University of Pittsburgh, 135 NorthBellefield Avenue, Pittsburgh, PA 15260, [email protected] Peylo, Software Logistik im Artland, Friedrichstr. 30, 49610 Quakenbrück,[email protected] and intelligent Web-based educational systems (AIWBES) provide an alternative tothe traditional “just-put-it-on-the-Web” approach in the development of Web-based educationalcourseware (Brusilovsky & Miller, 2001). AIWBES attempt to be more adaptive by building amodel of the goals, preferences and knowledge of each individual student and using this modelthroughout the interaction with the student in order to adapt to the needs of that student. Theyalso attempt to be more intelligent by incorporating and performing some activities traditionallyexecuted by a human teacher - such as coaching students or diagnosing their misconceptions. Thefirst pioneer intelligent and adaptive Web-based educational systems were developed in 1995-1996 (Brusilovsky, Schwarz, & Weber, 1996a; Brusilovsky, Schwarz, & Weber, 1996b; De Bra,1996; Nakabayashi, et al., 1995; Okazaki, Watanabe, & Kondo, 1996). Since then manyinteresting systems have been developed and reported. An interest to provide distance educationover the Web has been a strong driving force behind these research efforts. The researchcommunity was helped by the provision of a sequence of workshops that brought togetherresearchers working on AIWBES, let them learn from each other, and then advocate the ideas ofthis research direction via on-line workshop proceedings (Brusilovsky, Henze, & Millán, 2002;Brusilovsky, Nakabayashi, & Ritter, 1997; Peylo, 2000; Stern, Woolf, & Murray, 1998). Anumber of interesting AIWBES that were reported at early stages of their development duringthese workshops have since achieved the level of maturity. This double special issue capitalizeson the results of these workshops and assembles a collection of papers that represents the state ofthe art in the development of AIWBES.The goal of this introductory article is to provide a more systematic view to the variety ofmodern AIWBES and to discuss the role and the place of the AIWBES research stream in thefield of Artificial Intelligence in Education (AI-Ed). It provides a brief overview of knownAIWBES technologies classified by the field of their origin. It also attempts to distill the newdesign paradigm behind modern AIWBES and to compare this paradigm with a traditional designparadigm that has been dominating the field of AI-Ed for the last 15 years.ADAPTIVE AND INTELLIGENT TECHNOLOGIES FOR WEB-BASEDEDUCATIONAL SYSTEMSThe kind of advanced Web-based educational systems that this introduction attempts toreview are most often referred to as adaptive Web-based educational systems or intelligent Web-based educational systems. These terms are not really synonyms. Speaking about adaptiveP. Brusilovsky and C. Peylo/ Adaptive and Intelligent Web-based Educational Systems 157systems we stress that these systems attempt to be different for different students and groups ofstudents by taking into account information accumulated in the individual or group studentmodels. Speaking about intelligent systems we stress that these systems apply techniques fromthe field of Artificial Intelligence (AI) to provide broader and better support for the users ofWeb-based educational systems. While the majority of systems mentioned in this introductioncan be classified as both intelligent and adaptive, a solid number of systems fall in exactly one ofthese categories (Figure 1). For example, many intelligent diagnosis systems including GermanTutor (Heift, & Nicholson, 2001) and SQL-Tutor (Mitrovic, 2003) are non-adaptive, i.e., theywill provide the same diagnosis in response to the same solution to a problem regardless of thestudent’s past experience with the system. From another side, a number of adaptive hypermediaand adaptive information filtering systems such as AHA (De Bra, & Calvi, 1998) orWebCOBALT (Mitsuhara, Ochi, Kanenishi, & Yano, 2002) use efficient, but very simpletechniques that can hardly be considered as “intelligent”. The reason to focus on both intelligentand adaptive systems in this issue is that that the intersection is still large, the borders between“intelligent” and “non-intelligent” are not clear-cut, and both groups are certainly of interest forAI in Education (AI-Ed) community.Intelligent ESAdaptive ESFig. 1. Relationship between adaptive and intelligent educational systemsExisting AIWBES are very diverse. They offer various kinds of support for both students andteachers involved in the process of Web-enhanced education. To help in understanding thisvariety of systems and ideas, the author’s earlier review of adaptive hypermedia (Brusilovsky,1996) suggested focusing on adaptive and intelligent technologies. By adaptive and intelligenttechnologies we mean essentially different ways to add adaptive or intelligent functionality to aneducational system. A technology usually can be further dissected into finer-grain techniques andmethods, which correspond to different variations of this functionality and different ways of itsimplementation (Brusilovsky, 1996).An earlier review (Brusilovsky, 1999) identified five major technologies used in AIWBES(Figure 2). These technologies have immediate roots in two research fields that were wellestablished before the Internet age – Adaptive Hypermedia and Intelligent Tutoring Systems(ITS). Since their application in the Web context was relatively straightforward, thesetechnologies were the first to appear in AIWBES and can be considered as “classic” AIWBEStechnologies. According to their origin, the review (Brusilovsky, 1999) grouped the five classictechnologies into Adaptive Hypermedia technologies and Intelligent Tutoring technologies(Figure 2). The review also identified and grouped into “Web-inspired” AIWBES technologiessome new technologies that appeared on the Web more recently and had almost no direct roots inpre-internet educational systems. The lack of examples of