Before the computer age, the prevailing instructional approach was sufficient (e.g., one teacher transmitting information to about 30 students), but we now reside in a computerized world. Initial implementations of CAI mirrored this pedantic approach, and to some extent, so does the currently popular model-tracing approach in sophisticated ITS. Do we need to change our educational philosophies or systems?

We have most of the components necessary to advance educational reform. Not only is there great need for change, but also there are powerful, affordable technologies available to support it. Missing are definitive answers to the psychological controversies cited earlier. Basic research is actively being pursued to resolve these issues. For example, studies are beginning to consistently find that higher-order thinking skills are not acquired through didactic approaches (i.e., straight conveyance of facts), but rather, through learners' active involvement with the subject matter. This "constructivist" view of learning allows students to achieve intellectual accomplishments not possible under more traditional pedagogical approaches (Collins, Brown & Newman, 1989; Resnick, 1987).

The table 19-2 contrasts old versus new approaches to instruction (from Means, Blando, Olson, Middleton, Morocco, Remz & Zorfass, 1993).The table provides a clear direction for ITS research and implementation. That is, to get from "old" to "new," we need to open up learning environments that promote increased learner initiative and between-learner collaboration. We should assess learning as it transfers to authentic tasks, not standardized tests, and attempt to establish connections across various fields so topics are not learned in isolation of one another. As technologies emerge and advance, we can fit them into this framework. Furthermore, additional research is needed to validate the goodness of the new over the old approach to teaching-learning.

Look around you. Computer technologies have dramatically transformed the workplace, communications and commercial activities, as well as the entire business community. But education remains status quo. We need to harness the computer's potential and find ways to employ it in promoting educational change. Are current and prevalent ITS adequate for our purposes--now and in the 21st century (just right around the corner)? We believe that, as currently implemented, these systems may have asymptoted in utility. A philosophical shift has been suggested in this chapter, away from stand-alone instructional devices and toward using tools to aid in the more collaborative learning process. There are actually very few ITS in place in schools, yet they exist in abundance in research laboratories. We need to move on.

As we've discussed in this paper, reform can proceed along a number of pathways (perhaps in parallel). For instance, computer graphics are getting better every day; we can now develop 3-dimensional virtual environments where individuals can interact with any artificial world we choose to program (or purchase). Satellite transmissions can relay data to very distant locations; learners from different parts of the globe can access distal data, or even get together and jointly experience and solve various problems. Cognitive tools abound (e.g., simulators, hypertext/hypermedia formats, etc.), and we seem to be ready to recast our convictions about ITS. Rather than trying to create all-knowing, all-purpose teaching machines, a more fruitful approach may be to develop specific computerized tools. These tools can be specific for a given domain, or general purpose, applicable across domains. To paraphrase a well-known quotation: A person who is given a fish will eat for a day, but a person who learns how to fish will eat for a lifetime.

We can see the seeds of discontent growing. Go to any ITS-related conference and notice how researchers in the field have begun to discontinue using the term "ITS." Instead, in a show of semantic squirming, they refer to advanced automated instructional systems (formerly, ITS) as: Interactive Learning Environments, Cognitive Tutors, Individualized Teaching Systems, Computer-Assisted Learning, Automated Instructional Support Systems, Computer-Based Learning Environments, Immersive Tutoring Systems, Knowledge Communications Systems, Computer Tools, and so on.

Not only is the ITS construct too ambitious, but there is no universally-accepted definition of what comprises computer intelligence. While our working definition of intelligence is fairly specific, there exists a wide range of criteria in the literature related to computer-tutor intelligence. For instance, some say that for an automated instructional system to earn the label "intelligent," it must demonstrate the ability to learn by showing an evolving knowledge base. Yazdani and Lawler (1986) asserted, "No system which is too rigid to learn should be called intelligent." (p. 201). Others have argued that intelligent systems must provide for learner control during the learning process (Papert, 1980; Scardamalia, et al., 1989). Still others (e.g., MacKenzie, 1990) suggest that we reserve the word "intelligent" to describe only those systems showing truly impressive advances (e.g., intuition, empathy). Are these even realistic goals?

The fields of AI, psychology, and education have all greatly benefited from ITS research. But to continue (see Future 5), much more systematic research is needed to achieve some of the great potential offered by these systems. One suggestion is to begin a coordinated stream of systematic ITS research and development, altering specific features of existing systems and evaluating the results of those changes in accordance with a principled approach. According to Self (1989), "Once a sounder foundation for ITS has been specified, it becomes possible to identify the elements of a theory of ITS. These elements lie within (formal) AI, in areas such as belief logics, reason maintenance, meta-level architectures, and discourse models--areas from which ITS research has been divorced" (p. 244). Intelligent tutoring systems, as we now know them, may not exist 20 years from now, but we're on the right path, the motives are commendable, and the learner will ultimately profit.

As we began this review of ITS with the evolution of computer technology, so do we end it. ITS and related, developing technologies for education and training are constrained by two important factors: (a) the cost and power of computers, and (b) the pragmatic and theoretical knowledge of how best to employ them. Every month, computers are dramatically decreasing in cost and increasing in power; these changes bearing directly on consumer knowledge and application of the technology. While discussion of the interaction between these two factors goes beyond the scope of this chapter, we can make straightforward predictions about upcoming hardware and software developments. The MIPS (millions of instructions per second) curve is already converging on a BIPS (billions of instructions per second) curve in an exponential explosion that knows no limits. Desktop computers with 100 MIPS are currently available, and this raw horsepower makes a qualitative difference in computing possibilities. Soon, powerful systems will be available in notebook- and calculator-sized formats that fit into our hands, shirt pockets, and purses. Further, software tools enable us to learn from, and perform within, all major domains, such as algebra, biology, physics, art history, computer science, home economics, psychology, botany, calculus, accounting, and even manufacturing, medicine, and engineering. At our finger tips, we will be able to retrieve information, translate foreign languages, complete our tax returns, work out investment portfolios, analyze sales trends, and so forth. Software will be everywhere with embedded "assistants" to explain, critique, provide on-line support and coaching, and perform all of the ITS activities outlined in this chapter. Society stands at the edge of all this. Although the timeline for these exciting developments is uncertain, we do know that the research conducted so far is just a drizzle in comparison with the deluge to come.