Artificially-Intelligent Tutoring: An Assessment for the Future
The human mind not only has the capability to acquire knowledge but also to comprehend and manipulate a level of understanding often implied only by the knowledge itself. Thus far, however, traditional work in the field of artificial intelligence (AI) aimed at modeling the mind through rule-based heuristics has not proven very successful. Nonetheless, if computers can somehow be endowed with human understanding, insight, and reason, the result might not be an accurate model of human understanding, but it would certainly relate to and interact with humans much more like a human teacher does. The field of artificial intelligence-based pedagogy is essentially gearing towards achieving this goal.
The traditional notion of an Intelligent Tutoring System (Intelligent CAI or ICAI) embodies a structured knowledge-base full of task-specific facts and relationships, plus the power to query that information flexibly. In addition, an ICAI program must foster an environment where students learn through guided exploration, or where students' vague conceptual understandings are transformed into meaningful learning experiences -- ones rooted in their own actions and hypotheses, yet skillfully guided by an intelligent coaching mechanism. Such a tutorial should be able to critique student efforts and avoid situations that could lead students in the wrong direction without substantially controlling their creativity. Most traditional ICAI programs rely heavily on "expert-system" technologies, consisting primarily of a body of commonly understood facts and heuristic strategies characterizing expert-level decision making and intuition. Therefore, the performance level of an expert system is primarily a function of the size and quality of the knowledge-base that it possesses.
While this approach represents an adequate model for dealing with "known" information, it does not really deal with those processes we use for acquiring information. Thus, a technology known as "Neural Networks" has been explored, utilizing patterning methods to draw conclusions from incomplete stimuli. In other words, once a concept is set in the mind of the network, new and unknown variations of that same information will be recognized as such by the system -- something currently impossible for an expert system. On the other hand, neural networks have no ability to explain or understand their actions. While this technology has thus far had little impact on the development of ICAI, it has made significant waves in the areas of perception and cognition. Ideally, since ICAI relies upon cognitive-based heuristics to effectively tutor and coach students, a fusion of ICAI and Neural Net processes through "Expert Nets" represents a potentially powerful model. Whatever the future holds for this technology, it will surely provide us a better understanding of human learning, and inevitably it will assist us in the development of more effective intelligent tutoring systems.
Despite the research that has been done in these areas, few music ICAI programs are currently available. These programs are computationally demanding, and only a few music scholars are involved in this research. For the most part, the amount of time required to develop the necessary skills has limited the field to computer scientists. Even for those with special skills, development time is staggering, often requiring many thousands of experienced work hours. We may begin to see more activity, now that some schools are stressing computer training in their advanced degree programs, but only if we are prepared to support this research. Breakthroughs in artificial intelligence continues to present us with many new insights; however, it is up to us to draw upon this knowledge to expand the quality and effectiveness of all instructional technologies.