Customizing a Multimedia Tutor for Students Preparing to Take High Stakes Achievement Tests

Research suggests that multimedia and individualization of instruction are both effective ways to teach new material. This paper describes the integration of information about individual student cognitive characteristics into an intelligent tutor for high school geometry. The tutor presents a breadth of geometry problems with multimedia animations, embeds problems in narrative and fantasy and selects alternative teaching strategies based on a student’s cognitive abilities. We present an evaluation of the tutor that leads to conclusions of effectiveness and provide a guide for future adaptive versions of the tutor. Multimedia and Customization Both multimedia and customized teaching have been shown to be effective for learning. Multimedia has been shown to be effective in computer-based teaching and can range from simple graphics and animation to interactive lifelike pedagogical agents in rich, self-explaining 3D worlds (Lester et al., 1997). Advances in media technologies have created unprecedented opportunities for a new generation of knowledge-based learning environments. Some environments exploit artificial intelligence to support next-generation natural language dialogue. Animated pedagogical agents within learning environments can observe students' progress, provide visually contextualized problem-solving advice and play a powerful motivational role. Students also learn better when teaching material is customized to their learning needs. For example, research in human development points to the central role of one-on-one individualized instruction of the young learner by an older child, a parent, teacher, or other more experienced mentor (Greenfield & Lave, 1982; Lepper et al., 1993). Research suggests that novices construct deep knowledge about a domain through interaction with a more knowledgeable expert (Brown et al., 1994; Ericsson et al., 1993; Graesser et al., 1995). Multimedia techniques are common in commercial software (Beal et al., 2002) and have produced higher learning when present in educational software (Mayer, 2001). Yet, tutors that use intelligent agents to individualize teaching have not really taken advantage of the instructional possibilities of dynamic multimedia techniques. This paper describes the integration of both multimedia and customized responses based on knowledge about a student’s cognitive characteristics and gender. Although current intelligent tutors model a student's knowledge to provide effective help, only preliminary attempts have been built to incorporate knowledge of student group characteristics (e.g., profile of cognitive skills, gender) into tutors and to use this information to guide instruction (Shute, 1995; Arroyo et al., 2000). The benefits of customized tutoring are especially clear for those students with relatively poor skills and for under-represented students in some fields, e.g., minorities and women in science courses. Research has clearly shown that intelligent tutors support rapid learning and dramatically improve student performance. Tutors such as the Algebra Tutor (Koedinger et al., 1997) for algebra and AnimalWatch tutor (Arroyo et al., 2000; Beck et al., 2000) for arithmetic, indicate that students successfully master specific skills and that their attitudes towards math become more positive as a result of working with the software (Arroyo, 2003; Beal & Arroyo, 2002). Mathematics education is one discipline that has profited from the use of multimedia. Johari (2003) examined the effects of two inductive multimedia programs, one including use of a coordinate graph and its language, on university students' ability to conceptualize variables and create equations from word problems. The programs were designed to address the problem of syntactic and semantic translation misconceptions in mathematics problems. Results suggest that inductive multimedia facilitates the incorporation of many instructional strategies including inquiry learning from data, tutorials, schema, and core representational systems. The results are consistent with propositions recognizing the conceptual richness of visuals, specifically the coordinate graph, in mathematics education. Researchers using multimedia in mathematics education are cautioned to focus the multimedia on teaching inductive problem-solving strategies or scientific heuristics (for example, by working backwards, working inductively, or applying algebraic thinking to data) (Martinello & Cook, 2000) and to include the language of mathematics and math visuals (graphs) (Kaput, 1992). The SAT high stakes achievement tests have become increasingly important in the past years in the U.S.A. and a student's performance on such tests can have a significant impact on his or her access to future educational opportunities such as admission to universities and scholarships. New approaches are required to help all students perform to the best of their ability on these tests. Wayang Outpost, an intelligent tutor with real-time multimedia help reasons about a student’s cognitive characteristics while preparing them for the mathematics section of the a b