2. Benefits for Learning and Instruction Delivery

Information technologies might improve learning and teaching in two distinct ways. Just as they have made workers more productive in business and commerce, they could reduce teaching costs or increase the speed with which learners acquire knowledge. Alternatively, such technologies might help improve the quality of learning, rather than simply making it faster and cheaper. In this section we first review evidence that information technologies can improve the efficiency of learning and teaching; then we consider how they might lead to better educational outcomes.

Reducing Costs and Increasing Productivity

Most discussions of computers in education look at how new technologies might improve "instruction delivery," a management theorist's way of saying how teachers teach and how learners learn. Therefore, as resources for higher education continue to dwindle, when most people think of information technologies, it will usually be in terms of how such technologies might reduce the number of teachers needed or cut the time (and money) it takes learners to acquire skills. A collection of cases in current practice suggests that some technologies do indeed help improve productivity in these ways. (See Figure 2.1.)

Current successes are subject to several qualifications. Some cost savings are modest, others come with hidden prices (for example, high development costs or lower graduation rates), and all depend on careful attention to implementation in the classroom. But, although the data are still unclear, the economic logic behind the view is relatively straightforward: Cut costs by doing more with less; that is, hold student outcomes roughly constant (or improving) while displacing labor (staff) with capital (here, information technology). With technology costs continually dropping while labor costs only rise, this trade-off appears a good one for increasing productivity.

CAI and CBT

Computer-aided instruction (CAI) and computer-based training (CBT) have shown some ability to do more with less, by providing learners drill-and-practice systems that replace teachers in routine coaching tasks. These systems are most frequently found in secondary schools but also play a role in remedial math, science, and language courses at the collegiate level, as well as in business and military training (Figure 2.1). However, as convenient as they might be for self-paced learning, CBT programs often do not yield dramatic cost savings. Development costs for multimedia courseware are usually very high, often offsetting savings in delivery, and the price of a stand-alone computer system for each student is sometimes so dear that labor costs must fall dramatically before any net savings appears. The latter situation may change in the future, however, as powerful machines drop roughly to the cost of a top-of-the-line TV.

Distance Learning

A few intelligent tutoring systems (ITS) also have reported cost-reductions in training. (See Figure 2.2.) But, as with CAI and CBT, development costs currently impose strong limits on their productivity gains. In fact, the main attraction of ITS is that they may improve the quality of learning rather than reduce costs, as we discuss later under "Increasing Quality and Productivity."

Today, distance learning--whereby students and teachers rely on electronic and hardcopy media, rather than on face-to-face contact, for at least some of their communication (correspondence courses by postal mail are perhaps the oldest form of distance learning)--provides probably the best examples of cost savings that are at least partly attributable to information technologies. In this country, the PLATO system,[1] among others, pioneered distance learning in the 1960s, connecting as many as one thousand student-terminals at a time to mainframes. Only relatively recently, however, have these trailblazing ideas translated into reduced education costs. Several studies (Figure 2.1) show that distance-learning courses in Asia cost from 45 to 90 percent of the cost of conventional college classes, with comparable student performance; costs can drop to as little as 8 percent of on-campus courses, although completion rates then also seem to dip. Using telecommunications to offer distance learning for over three decades, the Open University in Great Britain--one of the first very large-scale experiments in distance learning at the higher-education level--boasts average per-student costs of around one-half those of conventional campuses. (See Figure 2.2.) We briefly examine its success to show why, at least right now, distance learning can help reduce costs only under relatively narrow circumstances.

The Open University: An Example in Achieving Cost Reductions. First, the technology must allow many students to be taught at once. Class sizes at the Open University usually exceed 200. The larger the class size, the greater the student-teacher ratio, which lowers labor costs and spreads high development and initial technology costs, leading, in turn, to productivity improvements. Break-even points--class sizes for which the cost of distance education and traditional delivery are roughly equal--will vary depending on, among other things, the size of technology investment and the course being taught. But it is a safe bet, at least for the near term, that distance-learning classes of less than 20 will rarely lead to substantial per-student cost reductions.

Second, a substantial amount of labor must be displaced by technology. Typically, labor is over 95 percent the cost of the class; so, modest trades of faculty for machines will not substantially change overall prices. Consequently, distance learning works best in relatively "standard" courses--math and the sciences--for which knowledge is relatively routine, can be embedded in the technology, and can have much teaching delegated to it. Alternatively, to expand the range of courses that can be taught at a distance, the use of technology should be carefully designed so that expensive labor at least can be replaced by cheaper help. For example, the Open University makes liberal use of tutors (replacing many full professors), either in person or on the phone.

Third, technology should substitute for labor as simply as possible. As the example of the Open University suggests, most low-cost distance learning does not fundamentally change the learning experience for the student, or the teacher. A computer application that promised to play a substantial role in teaching but that also demanded broad changes in faculty skills and activities (several of which are discussed in Chapters Three and Four) could actually increase labor costs, even while guaranteeing a savings in teacher time.

The bottom line, then, is that today's distance-learning technology is largely limited to imparting relatively routine skills and knowledge, taught en masse, through relatively traditional methods. Most of the best examples actually come from military and corporate training, rather than from higher education. For example, large companies such as Ford now contract for distance instruction to help their employees learn how to use tools such as spreadsheets and word processors, as well as specialized equipment. And Hewlett-Packard estimates that it saves millions of dollars each year using distance learning to train employees more effectively and more efficiently than with conventional methods. These examples explain why well over one-half of distance learning now goes on not in academic settings but in businesses--a market that is expanding so rapidly that whole TV networks are now dedicated to corporate distance learning.

The Internet is increasingly being viewed as a medium that can bring distance learning and its cost benefits into the higher-education sector. In the next subsection, we look at the cost and other advantages the Internet in particular can provide as a distance-learning tool for higher-education institutions.

Internet Impact: Distance Learning, Cost Reduction, and Increased Access

At least partly to cut costs, almost all higher-education institutions in the United States are considering distance-learning programs, dozens already have substantial offerings in place, and many universities are now moving in this direction. In the future, we believe that the Internet will be the most cost-effective way to deliver a new generation of distance-learning courses. By winter 1996, the U.S. Distance Learning Association (http://www.usdla.org/home.html) listed many dozens of universities with homepages outlining their WWW and Internet-based courseware. At the same time, the World Lecture Hall, a Web site managed by the University of Texas, has become a "virtual repository," organizing links to hundreds of other sites that are making Web-based class materials available to anyone who is interested. (See Chapter Three for a more complete description of the World Lecture Hall.)

Because the WWW first came online only in 1991, all these courses are new, and there are no solid data proving the effectiveness of Internet-based distance education. But a chemistry curriculum drawn almost at random from the World Lecture Hall[2] illustrates some of the reasons why Web-based distance learning seems very promising, at least "on paper." (See Figure 2.3.) To begin with, emerging Internet courses can include a wealth of material about the class (lectures, exams, assignments) and references to relevant online literature (books, programs, tools such as the periodic table), all accessible at the click of a mouse on a hypertext link. Each of these sources can be constructed as multimedia documents, including pictures, digitized audio versions of lectures, presentation overheads--even simulations, which can be run to give a much more dynamic impression of processes and structures. Full high-bandwidth interactivity will be available soon, to support multi-person, multimedia dialogues in real-time. The new generation of Internet-based distance-learning courses, in short, can mix the graphics and video of instructional TV with a much more intimate style of interactive mentoring than teleconferencing ever could. Therefore, it should provide substantially richer learning experiences for students than the current generation does.

Web-based distance learning might enhance productivity in higher education in several ways. Most obviously, by providing a more powerful and flexible medium, the Web should enable developers to construct higher-quality courseware for a wider range of classes than ever before. (See Figure 2.4, top panel.) With richer courseware, classes for which computers once replaced only a modest amount of labor might shortly be taught almost exclusively through technology; and in courses in which information technology previously had no part, it may play at least a supporting role. All this means that Web-based learning tools, appropriately used, promise to replace more labor and, consequently, to lower faculty costs.

But less-visible features of Internet-based distance learning--costs and standards--may enhance productivity even more dramatically. It is reasonable to assume that if courseware built on the Web is richer than previous distance-learning products, it might also be more costly to deliver--reasonable, but probably wrong. Although few realize it, the Internet is not only a better medium for learning and interaction but a cheaper way to deliver instruction (Figure 2.4, bottom panel)--something providers of corporate training have already figured out. For example, corporate providers are quickly trading their old on-demand video systems for more-flexible and less-expensive high-capacity intranets, which use Internet-style communication on private networks to deliver full-motion video to hundreds of viewers at once, at the click of a mouse. (See Figure 2.5.)

Schools and homes can expect similar benefits, even if they cannot afford the high-capacity LANs of big business. High-bandwidth yet low-cost connections to the WWW will soon come in several forms. Many homes today can access the Internet cheaply through ISDN phone connections, providing about 5 to 10 times the effective bandwidth of ordinary phone lines. More interestingly, successful trials of cable modems in 1996 promise very high-speed Internet connections, at roughly the cost of standard cable service. (See Figure 2.6.) If this promise is realized, the much-heralded-but-never-accomplished "500 channels of cable TV" finally could turn into something really useful. And, because they require no new and costly infrastructure development, wireless connections, such as DirectPC (http://www.helius.com/), may be the most promising of all. In short, the technical problems facing the installation of inexpensive wide-bandwidth Internet connections are all but solved; the policy barriers, however, loom large, as we discuss in Chapter Six.

In addition to reducing educational costs, the shrinking price of delivering Internet-based distance learning should help address other challenges for higher education. For one thing, Internet-based distance learning could be conveniently accessible to many different kinds of learners. Part of this improved access simply means putting more people in touch with more learning resources. A new wave of cheap "Internet appliances" (dumb terminals dedicated to Web browsing) should bring Internet access within almost anyone's budget; anyone who can afford Nintendo should be able to afford the WWW.[3] And once on the Internet, all resources are equally inexpensive to browse, whether they happen to be on a homepage next door or thousands of miles away. Equally important for improved access is the ability of Internet distance learning to reach consumers at any place and any time. This flexibility will be especially important, for example, to workers who want to "retool" skills at night while working during the day.

Reflection: The Death of Distance As a Barrier to Learning

At the broadest level, an Internet style of learning or instruction delivery will, we believe, be so common that it makes little sense to view distance learning as a special form of education delivery. Since Internet and Web tools will continue to drop in cost and increase in functionality for the foreseeable future, technology should, over time, displace more and more faculty labor. Further, what labor cannot be displaced will at least not have to be nearby: The Internet can connect students with faculty and peers as easily as it connects learners with multimedia documents. Eventually, then, most learning will be distance-independent, and the special cases will be those that are geographically bound.

Perhaps more to the point, distance learning should be viewed as a common tool in higher education, not as a separate--and esoteric--option. And most courses should be designed to use distance learning in some capacity, and often in a variety of capacities. A wide range of experiments in distance learning are already under way, and they come in many forms, rather than all fitting one mold. (See Figure 2.7.) The degree to which direct interactions are replaced can vary significantly: Telecommunications are sometimes little more than an add-on to classroom lectures; conversely, taped lectures may be the centerpiece of instruction, with face-to-face tutoring serving as a supplement; in still other cases, all discussion happens at a distance.

There is no single, obvious way to order courses from light to heavy users of distance-learning technology. Some courses that are, for example, offered exclusively at a distance (no face-to-face interaction) still make less use of telecommunication connections than classes that also include some direct contact. We ordered the following list roughly from courses that provide narrow connectivity, overall, to those that provide rich interactions. The list in Figure 2.7 is only illustrative; many distance-learning courses already mix and match features and functionality in ways we do not include here.

Less obviously, examples also differ in the bandwidth of communication, as well as the degree of interactivity. Early distance-learning courses (through postal mail) trafficked only in text, and a single interchange required days, if not weeks--not at all interactive by today's standards. Videotaped lectures distributed from a centralized source offer much greater bandwidth but no more interactivity; on the other hand, e-mail discussions between teachers and students are highly interactive but narrow in bandwidth. Some of the newest Internet-based distance learning enjoys both broadband, multimedia connections and real-time interactions, often linking not only faculty with students, but students with one another and with shared online resources. While earlier distance-learning models delivered knowledge from a centralized teacher (or "server") to students through narrow, slow, and separate connections, newer models look much more like an electronic community, tying together students, faculty, and distributed tools in a rich, speedy web of digital contacts.

With so many variants, it is no surprise that what makes an application of information technology an instance of distance learning is becoming less and less clear. For example, courseware such as the chemistry curriculum in the World Lecture Hall might be used in a distance-learning class; but right now, it is simply online material for an on-campus course. However, merely by skipping lectures and labs and relying on the rich Web documents to compensate for reduced in-person contact, individual students may turn it into a virtual distance-learning course. Such individual experiments, which probably have gone on informally for decades, now are leading to large-scale studies that will try to gauge the importance of distance as a barrier to learning in technology-intensive courses. (See Figure 2.8.)

It is useful to view this increasing freedom from spatial location in terms of the progressive "unbundling" of teaching functions in higher education as educational technologies have changed across time. Centuries ago, when knowledge was conveyed exclusively through an oral tradition, location was everything. With the advent of written texts, students were freed from the need to attend lectures (logically, but not always legally); however, other important functions, including tutoring, counseling, and evaluation, remained essentially face-to-face activities.

Now, just as books (and cars) have weakened the links between distance and learning in the past, today new information technologies can almost break them. Since the Internet can connect students with teachers as easily as it connects students with course material, tutoring and counseling can be (and are) done through e-mail and real-time chat rooms. (See Figure 2.4.) Similarly, as reference materials increasingly become available in digital form, students will no longer need proximity to libraries. Evaluation and assessment also can be done at a distance. Written products, such as essays, already are created routinely with word processors; sending them across a network for grading can be easier than printing them and walking them into a professor's office. Multiple-choice as well as short-answer tests have been computerized for years. In a pinch, even timed, closed-book exams could be conducted using simple keystroke logging or more-sophisticated Internet video products such as CU-SeeMe (http://cu-seeme.cornell.edu/) for proctoring.

Again, the point is not that learning should be conducted at a distance, just that creative applications make distance less and less of a barrier to learning. The challenge will be to discover the most effective applications of distance-learning technology and to blend them together with traditional face-to-face transactions.

Increasing Quality and Productivity

While many applications of information technology in education try to reduce costs, others focus on the output side of the productivity equation: helping students learn more or better, or to learn new skills--applications rarely included in higher-education curricula. A collection of examples (Figure 2.9) illustrates several ways in which information-technology applications could lead to better learning and teaching outcomes.

CAI and CBT

As discussed, most successful computer-aided instruction and computer-based training applications boast of cost reductions (Figure 2.1), but a few also claim gains in the quality of student learning. Some of these applications rely on distinctly cutting-edge technology, making lavish use of multimedia, and engaging students in highly interactive discussions. However, other ideas are surprisingly simple. For example, The Teaching Company (TTC) thinks that the videotapes and audiotapes it offers can help college students improve grades in everything from history to psychology to physics. Of course, these tapes lack the interactivity of a good tutoring session, but TTC is gambling that the quality of its tapes will more than compensate: The lectures are delivered by elite, "superstar" faculty in each field, not by merely competent professors.

Some very positive reviews of these products (Figure 2.10) may be too optimistic. However, in the future, Web technology could clearly make educational packages featuring superstar instructors easier to develop and more broadly accessible, perhaps leading to significant improvements in teaching and learning productivity. Imagine, for example, a course built around master lectures, digitized and put online, and combined with complementary assignments, projects, tutorials, and group discussions--like those already found in World Lecture Hall courses. Taking this course through the Internet might not be quite the same as being there in person, which would require being a student at MIT, Stanford, or a similarly elite and pricey institution. Nonetheless, for less-fortunate students, such an online course could be better than an in-person course offered by a second-tier university. Here, as in other areas, information technology might lead to improvements in productivity by making truly great--not merely competent--practitioners and their performances available to almost everyone.[4]

Intelligent Tutoring Systems

While replicating outstanding teaching performances on video might lead to improvements in learning productivity, a much more ambitious approach is to try to capture, in software, the reasoning of teachers. Of course, this cannot be done just by pointing a camera at a smart talking head. Rather, the process of developing so-called intelligent tutoring systems (ITS) begins with an expert system--a collection of heuristic rules--that is capable of answering questions and solving problems in a given subject matter. For instance, the expert system inside a good calculus ITS would be able to solve arbitrary symbolic integration problems, and would do so in a way that approximates how an "ideal" human mathematician would reason. The expert system in an ITS is often supplemented with a student-modeling capability, which can examine a student's reasoning, find the exact step at which he or she went astray, try to diagnose the reasons for the error, and even suggest ways of overcoming the impasse.

The potential value of such highly intelligent systems is obvious. Indeed, for decades, the idea of supplying students with their own automated tutor, capable of tailoring learning experiences to students' needs at a fine level, has been the holy grail of teaching technology--with good reason. Many studies have shown that one-on-one tutoring is the best way to learn. And, as the examples in Figure 2.9 demonstrate, for a few academic subjects, ITS have nearly matched the capabilities of their human-tutor counterparts, helping to raise students' scores one letter grade or more.

However, while such systems show that ITS can raise the output side of the productivity equation in education, their successes to date are very limited. Because they try to embed detailed human-reasoning skills, both in subject-matter areas and in teaching expertise, they rely on advances in cognitive theory and software engineering. But these disciplines have yielded a detailed understanding of thinking in only a very few well-defined subjects; consequently, ITS currently are confined to relatively simple parts of algebra, geometry, computer programming, physics, and other sciences.

In higher education, then, ITS probably would be most appropriate in basic freshman and remedial classes. And, needless to say, even in these subjects, automated tutors come nowhere near the ability to mimic superstar teachers. On the contrary, ITS are frequently called "brain damaged" by most computer scientists, their developers included. Over time, however, ITS will get smarter and smarter. Buttressed by advances in cognitive science, they will capture an increasing share of human-teaching expertise, and they will extend to a wider range of subjects.

However, all evidence suggests progress here will be slow. While the speed of computer hardware continues roughly to double every two years, the intelligence of computer software, however you measure it, creeps ahead at a snail's pace.[5] But this does not mean that information technology is destined to play only a minor role in helping improve the quantity and quality of learning in higher education. It just means that some of the most obvious ways in which technology attempts to partially replace faculty as lecturers or tutors in a traditional classroom setting may not be the most promising ways, at least in the short run.

Interactive Learning Environments

A glance back at Figure 2.9 suggests that some of the most successful applications of information technology follow a very different principle: Instead of mimicking human tutors, they provide rich, simulated environments that enable students or trainees to practice skills intensively.[6] Flight simulators, the first such learning environments, are still perhaps the most compelling ones. For decades they have been singularly effective technologies for improving the quality of pilot training. In the protective environment of a virtual cockpit, pilots can perfect difficult maneuvers, trying them over and over without fear of the deadly effects of a real-world mistake.

SHERLOCK (Figure 2.9), a descendent in spirit of these early simulators, demonstrates similarly impressive outcomes: Students can learn electronics troubleshooting and fault-diagnosis skills in less than one-tenth the time they previously required. While SHERLOCK includes some important coaching supports, the basic reason that trainees acquire skills so rapidly is simply that the system can present students with simulated problem situations that arise only rarely in the real world--namely, troubleshooting the complex hardware that diagnoses faulty F-15 avionics.

Trainees do not learn these troubleshooting skills very quickly in the field, because problems that might foster new knowledge naturally occur, say, once every few weeks. Using SHERLOCK, such problems can be made to order every few minutes, if necessary. More generally, this and other virtual worlds rely on the fact that one of the most powerful ways to learn is by doing; but learning-by-doing works well only if the world cooperates by providing the learner with a large number of practice cases, and cases suitable to the learner's current level of expertise. In situations where the real world cannot be easily controlled or does not readily cooperate, computers can now provide rich simulated worlds, at an arbitrarily high degree of fidelity, that can be controlled with the click of a mouse.

Such environments seem promising tools for training, but can they work in higher education? While it is relatively easy to imagine how simulated worlds might be useful in teaching concrete subjects such as electronics or even physics, it is less obvious how simulations would help in abstract or symbolic disciplines such as history, philosophy, or the social sciences. However, some innovative systems are already pointing the way. For example, CoVis (Figure 2.9), Rebel! (Figure 2.11), and even strategy games such as SimCity, are built on top of complex computational models that capture key features of the real-world dynamic systems underlying policy decisionmaking, weather, and historical processes. True, these models do not simulate concrete physical systems; but, at an abstract level, the conceptual systems they represent also comprise dynamic cause-and-effect relationships among a collection of variables. What a new generation of simulation-based learning environments are doing, in essence, is combining computational models of such processes with highly graphical and easy-to-manipulate "visualization" front-ends. Collectively, these environments give the learner a virtual world in which many what-if experiments can be performed.

In theory, then, just as SHERLOCK provides trainees with a large number of rich avionics cases, these simulations might present learners with, say, a series of hypothetical history scenarios, some of which may have analogs in real historical events and some not. These technologies appear very interesting, but they are also very new. Whether they can lead to the dramatic gains in the quantity of learning we find with SHERLOCK remains to be seen.

Internet Impact: Turning Expensive Stand-Alone Systems into Cheaper Distance-Learning Technologies

Some of the applications that promise to help raise educational productivity by improving student learning, rather than by cutting costs, are already network-based systems. In CoVis, for example, students use the Internet to communicate with distant mentors, collaborate on projects with peers, gather data from remote sites, and talk with experts in atmospheric and environmental sciences. (See also Figure 2.12.) But many of the cutting-edge applications are stand-alone systems that run on only one or a few computer platforms. For instance, one might work on a Macintosh but not on a PC. Even worse, since many of these systems are still research prototypes, they may run on only relatively high-end machines (Sun Workstations or even supercomputers, for example). All this would seem to mean that some of the most interesting new information technologies for higher education will be, for the foreseeable future, either inaccessible to most students or prohibitively expensive for schools or students to acquire.

Not necessarily. One way around this problem, again, exploits the Internet and the WWW. It is already possible for stand-alone software systems to piggyback on the near-standard protocols for creating WWW documents. In doing so, these systems can be made available very widely and relatively cheaply.

As anyone who uses the WWW knows, Web browsers and servers, unlike most software, run on a wide range of machines. The considerable complexities of making key languages (such as HTML) and protocols (such as HTTP) de facto standards are hidden from the average user, who can happily create documents that will appear on any machine just about the same as they look on his or hers.[7] Technically, only WWW documents enjoy this relative machine independence. However, using scripting languages (e.g., CGI) that enable programs to be initiated on remote machines and new homepages to be created and loaded on the fly, it is possible to turn almost any piece of educational software into a distance-learning application.[8]

The cleverness needed to turn isolated software into sharable distance-learning systems on the WWW might be of technical interest to some readers. (Certainly the exercise has fascinated the authors.) But such technical subtleties miss the key point of the example: Already, the Internet and WWW permit high-powered applications that run only on high-powered machines to be accessed by students who might have available only low-end machines--and almost any kind of low-end machine, not just those that the original application developer envisioned.

The future looks even more promising, on several scores. First, as already noted, ultra-cheap Internet appliances should bring down access cost even further. Second, the protocols needed to make products widely sharable across the Internet, using almost any hardware platform, are quickly falling into place. Tricks now required to turn the WWW into an interactive medium will soon lose their appeal, primarily because new Web tools, and languages such as Java, for building "live computational objects" (it is misleading to call them "documents" any longer), will soon provide primitives that make this kind of interactivity easy to achieve rather than a challenging problem to solve.

All this may mean that higher education can have its cake and eat it too. Most of the current cutting-edge information-technology applications we reviewed promise to help students learn better or faster, or both. Improving educational productivity in this sense is clearly an important goal for higher education. But, until recently, this goal seemed at odds with others: reducing the cost of delivering education and increasing access to education by an increasingly diverse population of students. Without question, in the future, most applications of technology in education that offer exciting possibilities for transforming teaching and learning will continue to be developed in corporate and university research settings on high-end machines. And these will probably remain expensive, well beyond the means of almost all students and even most higher-education departments and computer labs. However, high-bandwidth networks, coupled with increasingly standardized software languages and communication protocols, and with ever-cheaper Internet terminals, should bring these cutting-edge applications within reach of learners and teachers who do not have access to the newest and fastest technologies.

This does not mean that poor students will enjoy all the same access to technology-based learning as wealthy ones, of course. There will always be computers that offer more functionality at a higher price. And even if high-quality educational materials could be accessed by anyone on any machine at any time, nothing right now prevents providers from metering this access and charging whatever the market will bear. (Nor, necessarily, should providers be prevented. These and related policy issues will be examined in Chapter Six.) The point, however, is that information technology by itself will no longer be a significant barrier to access; if anything, the WWW and Internet should be a force toward greater access to and equity in education.

Reflection: Many Interesting Options, Many Hard Choices

Certainly information technologies do more than provide better tools with which to deliver instruction and learning materials in higher education, but such tools still dominate the field of educational technology. We have discussed (or at least pointed to) a wide range of these applications, focusing especially on those that exploit the WWW, or could do so. Stepping back from the many specifics, several general points should be emphasized.

First, it should be obvious that information technologies offer many different ways to improve instruction delivery, not just one. Looking just at distance learning, there is a rich set of options to choose from. Selecting from the different choices demands attention to the (often implicit) educational goals the technologies address. Some might help higher-education institutions reduce costs; others look better-suited to increasing access to educational resources, or to increasing the quantity, quality, or speed of student learning. Overall, the proliferation of Internet and WWW resources should steadily bring down the cost of all educational software products and services. Yet, some will be more costly than others, and a few will still be very expensive for the foreseeable future.

In general, then, it is a mistake to assume that information technology provides one or a few ways to improve learning and teaching, and it is equally wrong to believe that technology applications will affect just a single educational goal: cost reduction.

A third assumption, probably more widely believed, is also wrong: that information technology is mainly a vehicle to enhance educational productivity. Look back for a moment at the last few information-technology applications described in Figure 2.9. While some, such as SHERLOCK, clearly intend to help speed up learning, or to improve learning of well-defined subject areas, it is much less obvious how others are trying to improve educational productivity in any simple sense. Instead, they are generally trying to transform the processes and products of learning. For example, they might exploit new visualization technologies to enable junior students to learn about complex systems that previously only graduate students found comprehensible; and they might foster deeper understanding through inquiry-based learning rather than traditional lectures, or through collaborative-learning methods that simply never existed prior to high-bandwidth networks.

In short, here information technology is a driver of educational reform, not of productivity enhancement unless productivity is redefined to include a wide range of qualitative changes, not just quantitative improvements.

That these applications transform processes and products of learning more than they enhance productivity has several implications for their adoption, implementation, and costs. Consider, from the point of view of the developer, how many different facets of an educational context might need to change for the potentially impressive benefits of such applications to be realized. (See Figure 2.12.) Clearly, the technology itself--the tools actually used by students in learning--is just the tip of the iceberg. Other possible changes range from the roles of teachers, to methods of evaluating learning, to deeper philosophical views of how learning happens and what constitutes knowledge.

All of this is roughly consistent with a simple rule we have distilled from our own research in developing and implementing educational technology: If a technology fits into an existing educational process and tries to enhance existing educational products, you spend 75 percent of your time developing the technology and 25 percent worrying about all other implementation issues; if a technology tries to transform educational processes and products, you spend 75 percent of your time attending to these nontechnical, often complex, and frequently invisible changes in institutional practice and structure.

If anything, our 75/25 rule understates some of the complexities and costs of integrating cutting-edge applications into the classroom. True, it does suggest that deep institutional changes might be necessary to make the best use of some information technologies. However, it also seems to imply that we can anticipate the best ways to use these tools in advance, and that the biggest challenge will be to change the educational system to reflect what society knows it needs to learn and teach.

Although this statement is probably far too optimistic, who would have imagined a decade ago that visualization technologies or simulated worlds would open new educational experiences to learners? By the same token, it is extremely difficult to predict with any confidence what new and interesting applications of information technology for learning will surface over the next decade, or to imagine the kinds of institutional changes that might be necessary to best exploit these new tools.

This is not to argue that all good applications of the Internet and WWW to deliver education are shrouded in mystery. Again, there will be many options to choose from. Some applications--perhaps simple Internet distance-learning tools and the next generation of ITS--may continue to fit well into existing higher-education structures and to reduce costs or increase teacher productivity, others may give rise to new structures and cause us to rethink our ideas of educational productivity, and a few may transform learning in completely unexpected ways. Shortly, we will see additional examples of unexpected applications of information technology. Chapter Six follows up a few policy implications of this uncertainty.

[2] This curriculum was found under the "Chemistry" entries on the World Lecture Hall homepage (http://www.utexas.edu/world/lecture/index.html) on February 22, 1988. The course itself was at Brown University (http://jcbmac.chem.brown.edu/baird/Chem22I/chem22i.html) on the same date.

[3] Although dropping technology costs will make the Web accessible to make many, it may be still beyond the means of many urban poor--a group who arguably might benefit most from access to low-cost educational services. As of 1993, only 7 percent of households in the bottom quartile in income owned a computer. Further, recurring Internet connection costs (still at least $20 per month in 1998) are substantial. See How Will Individual Learners Access Internet-based Education? below and, especially, Anderson, Bikson, et al. (1995) for further discussion of these access issues.

[4] See, for example, Rosen (1981) for an economic analysis of how information technology makes great performances widely available, increases the wealth of the elite (at the expense of the merely gifted), and therefore shrinks the number of people who can make a living providing information services to a small contingent. Rosen claims that this phenomenon applies not only to entertainers and sports figures, but (to a lesser extent) to lawyers and business executives, whose range of influence and availability is expanded by modern telecommunications.

[5] Well-publicized successes of "man over machine," such as Deep Blue's recent win over world chess champion Gary Kasparov, are false counter-examples. Deep Blue's performance relied almost exclusively on fast computer hardware. The program uses brute-force searches and "understands" virtually nothing about the way humans play the game. It is just this kind of understanding that automated tutors must possess if they are to help humans learn.

[6] Interactive and simulation-based learning environments such as those we discuss here are certainly not the only alternatives to ITS and systems that try to be "smart" or human-like. A few, listed in Figure 2.9, take different approaches; for example, some offer new "power tools" for collaborative learning.

[7] The universality of Web standards such as HTML and HTTP is more fragile than it might appear. Even the simple current versions of these protocols do not run on all machines. Future versions threaten to be even less general. For one thing, different groups are extending the basic standards in many incompatible directions; for another, the pressure to develop highly interactive Web objects quickly is speeding up the pace of innovation, which works against the slow processes of achieving a broad-consensus standard. That said, the current HTML and HTTP standards are far closer to a universal standard for complex computational objects than anything we have seen before.

[8] See Figure 2.11 for an example of how we turned Rebel! into a Web-based history simulation that could be run on one (high-end) machine but displayed on a variety of other (low-end) ones.