Expanding the Scale and Scope of Information Technology Research
The twenty-first century begins with unprecedented opportunities to use information technology (IT) to meet a growing number of societal needs. The payoff from investments in IT research initiatives launched a half a century ago have become manifest: IT has moved from the laboratory to the office, store, and home and has been incorporated into personal belongings of all types. It is also transforming countless aspects of business, government, and social interaction from education to health care to commerce, and as the potential of IT grows, so, too, do users' expectations for it. More people want IT to do more things, more easily, with more trustworthiness, and with more reach into society. The IT research community has many ideas about the ways in which computer science and engineering can be developed and applied to meet those expectations and new ones that will emerge as further progress is made. Paradoxically, however, it looks as if it will be more difficult to make practical progress. The potential of IT and the challenges that must be overcome to realize that potential are rising in tandem, boosting the level of required investments. Everyone can see the growth of markets for IT goods and services, but the need for more research is less obvious. Nevertheless, the range of emergent issues for IT research to address is increasing, calling for an expanded research agenda.
Many reports call for more funding for IT research. Such pleas come in the face of the misperception that companies that sell large amounts of IT products and services and that generate large revenues or market valuations are capable of the types of research needed. Many reports also
point out that academic research, funded largely by the federal government, is essential for expanding the knowledge base.1 Building on that work, this report also advocates a large research effort that fosters advances in a number of important, specific areas of IT. At the same time, it goes beyond previous appeals by arguing that the metamorphosis of IT from distinct devices (both hardware and software) into complex, large-scale societal infrastructure calls for a shift in the emphases of IT research, which in turn requires different approaches to the organization and conduct of IT research. Traditional mechanisms for funding and conducting IT research are not necessarily attuned to today's research challenges, so new mechanisms must be developed. Attention to the ways in which research is supported and conducted is more important than ever, given the allure held by industry over the past few years for the talent that historically filled the ranks of graduate students and faculty. Today's reality is that research competes with other activities, and research programs must emphasize compelling and important problems that will attract and retain talented individuals.
This chapter summarizes the study committee's primary recommendations for ensuring the continued well-being of the nation's research base in IT. The premise, based on the arguments put forth in the preceding chapters, is that IT research can no longer focus almost exclusively on the IT components that have been the hallmark of past initiatives but must expand to include commensurate research efforts on the problems of large-scale systems and the social applications that they support. The chapter recommends the continuation of ongoing and substantial funding and support for traditional areas of IT research, but it also calls for new research that emphasizes large-scale systems and social applications. It recommends both meaningful efforts to promote something that has proven difficult to achieve —interdisciplinary research—and diversification in the modes of supporting and conducting such research. If successful, these research efforts could ultimately improve education in large-scale systems and social applications, helping to create a workforce better able to research, develop, and use IT systems.
Implicit in the recommendations is a recognition that the IT research community is stepping into uncharted territory in which many new research challenges are arising and even familiar problems can assume new forms. The recommendations therefore seek to foster experimentation with a variety of approaches to bringing together diverse communities with different sets of expertise and different perspectives on the issues of IT research—characteristics that have contributed to the nation's history of success in IT (CSTB, 1999a). They lay out the desired characteristics of research mechanisms but recognize that multiple approaches may need
to be tried, evaluated, and fitted to the particular characteristics of the institutions involved.
The recommendations are organized according to the group of stakeholders, decision makers, and policy makers to which they are directed: those in government, universities, and industry. As the discussion notes, however, these groups will have to interact to ensure that the set of research programs put in place will meet society's growing needs for IT and assure that this and future generations can safely depend on IT systems.
RECOMMENDATIONS FOR GOVERNMENT
The federal government has an important role to play in helping to expand the IT research agenda. Government agencies have long been an important source of funding for the IT research community as well as for the other research communities that may need to become more closely integrated into IT activities. Although federal expenditures for IT research are dwarfed by those of industry, they constitute the overwhelming majority of research funds provided to universities and thus are critical in supporting long-term fundamental research that can benefit a wide range of companies, both established companies and start-ups. Moreover, because federal funding is so pervasive in academia, it can be used as a lever to help direct academic research toward needed areas. In doing so, it can lay the groundwork for more subtle transformations of research and development (R&D) in industry as well. The committee makes four recommendations to help guide this process.
Recommendation 1. The federal government should continue to boost funding levels for fundamental information technology research, commensurate with the growing range of research challenges.
The first step in expanding the scope of IT research is to ensure the availability of sufficient funding to address the growing range of research problems that must be tackled. The scope of inquiry for IT-related research is clearly growing (as is the need for more creativity and flexibility in how funds are allocated and spent). The unprecedented growth in the complexity, size, and social engagement of the IT systems now being deployed calls for fundamentally new concepts, abstractions, and methodologies to master and harness IT for the good of the society. Attaining those concepts, abstractions, and methodologies is the goal of fundamental research. Hence, increased funding is needed not only to support continued advances in the capabilities of IT components (e.g., increased processing power, storage capacities, and communications bandwidth)
but also to develop solutions to the problems posed by large-scale systems and social applications of IT. The committee views components, systems, and applications as three equally important areas for research, noting that they play roles in IT analogous to the roles of biomedical research, physiology, and medicine in the health sciences. Each area informs work in the others, and fundamental scientific understanding is needed in all three areas to ensure a properly functioning system.
As noted above, the government is the primary vehicle for support of fundamental IT research. Although industry funds a considerable amount of research (some of which is fundamental research) in its own laboratories and in universities, intense competitive pressures and the need to generate positive returns for investors force companies to direct more of their R&D funding to projects with more certain results and more obvious applicability to market needs. The potential social return on investments in research is enormous, but these investments will not be made without the government's lead.
It is not feasible to specify a precise dollar amount by which IT research funding should increase, but the increases recommended by the President's Information Technology Advisory Committee (PITAC) and requested by the Clinton Administration for fiscal years 2000 and 2001 are representative of the magnitude of the annual increases needed for some time to come (PITAC, 1999). Government program managers report that they receive far more high-quality research proposals than they can fund (a situation that is common in other fields, too). How the money is spent is at least as important as the amount. Researchers in the field observed, in testimony to the committee and in other contexts, that the allocation of federal funding shows less vision and more emphasis on process than it did in the middle of the twentieth century. The historic comparison is important, because there is a correlation between the approach to funding management and the yield. Today's circumstances demand a more visionary, less process-bound approach, as will be discussed below, because the needs of large-scale systems and social applications can be met only with innovative, revolutionary work. The Defense Advanced Research Projects Agency (DARPA), in particular, has a history of supporting revolutionary work, and this orientation should be reinforced and encouraged.
Recommendation 2. The National Science Foundation and the Defense Advanced Research Projects Agency should establish significant programs of fundamental research in large-scale information technology systems.
Large-scale IT systems pose difficult technical (and nontechnical) problems that are manifested in a variety of ways: delays in designing
and deploying new IT systems, failures of operational systems, inability to add new functionality to existing systems, and unexpected behavior. The direct cost of these failures is high.2 By some estimates, between 70 and 80 percent of all major system development efforts are never finished, seriously overrun their cost and development time objectives, or fail to provide the desired functionality. Hundreds of millions—even billions—of dollars are spent on such failed efforts. The indirect costs of fragile systems are even greater because of the potential for widespread damage from failures of critical IT infrastructures used for controlling the electric power grid, communications systems, or financial transactions.
Problems with large-scale systems are not new, but the steady push to build and use such systems means there is an imperative to address those problems now. System failures are not only the stuff of daily news, they are also the stuff of congressional hearings and inquiries by regulatory agencies. Improved techniques for designing and implementing large-scale systems will require fundamental research to build a stronger scientific base for understanding such systems. A stepped-up research program should include not only case research that examines particular systems also methodology research that seeks common architectures, techniques, and tools that can influence a wide range of large-scale systems. Two elements are key to success: the enabling resources (i.e., funding and the talent it nurtures, infrastructure, access to appropriate artifacts) and the cultivation of a motivated community of researchers.
Although all federal agencies face problems with large-scale systems, DARPA and the National Science Foundation (NSF) are the best positioned to lead research efforts in this area because it needs fundamental science and engineering and because they have the longest history of managing related research and the closest contact with the appropriate research communities. Both DARPA and NSF have a number of programs in place that address elements of the large-scale systems problem, but their programs have not been linked to form the larger thrust that would give this problem area the high profile it deserves. A stepped-up program would help create the critical mass of researchers needed to address large-scale systems issues and help form a research community around this set of problems. Strong federal leadership of the sort that DARPA and—to a lesser, but growing, extent—NSF have demonstrated in the past could bring about a more comprehensive approach.
The development of the fundamental Internet technology provides a model: effective program management led a dispersed group of researchers to work on separate projects toward a shared goal. A dynamic mix of people and institutions participated in different ways and at different times according to their interests and capabilities and the needs of the project or program. Both DARPA and the NSF should experiment with
program management to promote such leadership, creative research, and community efforts to tackle today's large-scale systems challenges —which are, of course, similar to some of the early Internet challenges. These agencies will need to ensure the quality of the research while allowing researchers sufficient freedom to pursue fundamental, visionary work.
The research conducted under a broad program on large-scale systems should pursue a diversity of problems and approaches. As described in greater detail in Chapter 3, it should do the following:
Support both theoretical and experimental work;
Provide small, medium-size, and large awards to support individual investigator research, small teams of researchers, and larger collaborative efforts;
Pursue a range of approaches to large-scale systems problems, such as improved software design methodologies, system architecture, reusable code, and biological and economic models;
Attempt to address the full scope of large-scale systems issues, including scalability, heterogeneity, trustworthiness, flexibility, and predictability;3 and
Provide academic researchers some form of access to large-scale systems.
Access to large-scale systems could be provided in any of the several forms: researchers could team with organizations that deploy or operate such infrastructures; they could create separate experimental testbeds that would allow them to develop, demonstrate, and test new techniques without worrying about interfering with operational systems; or existing systems (such as portions of the Internet) could be better instrumented to provide researchers with the kinds of data they need for analysis. Some combination of these approaches will undoubtedly be needed.
Given the broadening circle of agencies interested in and involved with IT research, DARPA and NSF should attempt to involve in this research program other federal agencies that operate large-scale IT systems and that would benefit from advances in their design. In other words, participants should include not only other agencies with a history of funding IT research (such as the Department of Energy, the National Aeronautics and Space Administration, and the National Institutes of Health) but also agencies that have not traditionally funded IT research but have large budgets for developing and procuring IT systems (such as the Internal Revenue Service, the Federal Aviation Administration, and the Social Security Administration). Doing so may help to ensure that researchers have access to research facilities that would allow them to
better understand the problems faced in large-scale system design, development, and operation. Agencies such as the Deaprtment of Health and Human Services and the National Institute of Standards and Technology might be able to expand their support for fundamental research and system problems, too.
One approach for engaging a diverse set of agencies in an IT research effort would be to build a vigorous program within NSF to understand the Internet and make it more robust. This effort could build on ongoing activities, such as the Next Generation Internet program, which involves several federal agencies, including DARPA, NSF, NASA, DOE, and NIH (via the National Library of Medicine).4 The idea would be to use the Internet as the experimental testbed and build an institutional structure that facilitates research on it. In addition to generating useful technical results, this Internet-as-testbed approach could evaluate whether the approaches the committee suggests are effective in practice.
Recommendation 3. Federal agencies should increase support for interdisciplinary work on social applications of information tech nology that draws on the expertise of researchers from IT and other disciplines and includes end users of IT systems.
People use IT. That has long been the case. What is different now and for the future is that, as more people use IT, and in more ways, they are less likely to be expert users or interested in how IT works (as opposed to how well it works); furthermore, they want to make more and better use of IT in ways that affect their lives more intimately and directly than the early systems did in scientific and back-office business applications. These are issues with which the traditional IT research community has little experience. Successful work on the social applications of IT will require new computer science and engineering as well as research that is coupled more extensively and effectively to other perspectives—perspectives from other intellectual disciplines and from the people who use the end results, that is, the goods, services, and systems that are deployed. For some time now, the Computer Science and Telecommunications Board (CSTB), echoed by PITAC and the Clinton Administration, has been calling for research on the societal impacts of IT; the present committee is emphasizing a complementary, technical sort of research that will result in IT with fewer adverse and more positive impacts.
The ongoing Information Technology for the Twenty-First Century (IT 2) initiative, in particular the social, economic, and workforce component of the IT Research (ITR) program at the NSF, provides a basis for addressing both the impacts of the systems and new approaches to them; the committee, however, envisions a more substantial effort that interacts
with the rest of the IT research effort. The bases for such work are already evident within programs supported by DARPA, NSF, and other agencies. The NSF's Computing and Social Sciences program and Digital Government program are important steps in the right direction, and they could serve as models or launching points for more expansive efforts if they prove successful in the long term, but at present, they are limited in size and scope. Broader programmatic support is needed, along with the attendant nurturing of a research community. The major expansion that is warranted should have the following characteristics, which are discussed below:
Research support provided through a variety of mechanisms;
Explicit participation of end users and systems integrators in the research process;
Participation of federal agencies that are major users of IT systems, not just the traditional funders of IT research;
Access to operational systems and support for testbeds;
Management and oversight provided by traditional funders of IT research; and
Pursuit of fundamental knowledge of the interaction between IT and the context in which it is deployed.
Research support mechanisms should range from small, single-investigator grants to medium-size collaborations among researchers from different disciplines, to activities characteristic of research centers that enable large numbers of researchers to interact for extended periods of time and across multiple projects. Significant progress will not be made on social applications through single-investigator research alone; efforts will be needed across the full range of program scales and scopes. The NSF has recognized this need in its solicitation for the ITR program, which calls for grants of many sizes, including for the establishment of centers to focus on social, economic, and workforce issues. The Computing and Information Science and Engineering (CISE) directorate has limited experience in managing such efforts, but it can build on NSF's experience with science and technology centers, engineering technology centers, and supercomputing centers. It will need to ensure that centers maintain a strong focus on research and produce high-quality results through periodic evaluations.
Another way to organize larger research programs would be to cluster activities around a project rather than around a center per se, as DARPA and, to a lesser extent, NSF have done in the past. Program managers could pick specific problems to work on (e.g., a design project, such as air traffic control, control of the electric grid, or payment for health care
services), establish research goals to be pursued, facilitate the relationship between the research community and the system to be studied, and issue a call for proposals to participate—much as has been done with the Digital Libraries Initiative.5 The life of such projects, unlike that of the center, would not be fixed at 5 or 10 years but would be linked to the natural cycle of study implied by the testbed artifact. Such a project would be more dynamic than a center, allowing a changing mix of researchers and institutions to participate in different phases of the project as their interests and the needs of the project suggest as opposed to binding them together for a set period of time. Program managers at DARPA and NSF would exercise leadership by establishing a common direction for all the researchers, even if they were not all working under the same contract. A process of this kind would draw on a set of programmatic goals to motivate research rather than emphasize a particular form of organizing researchers. It would more closely resemble the organization of early research on the Internet rather than the organization of NSF's Science and Technology Center program.
Research programs motivated by social applications must engage end users and system integrators in order to better understand the problems that people and organizations are confronting with IT systems and the range of potential solutions and to gain the inspiration for research. It may be unrealistic to expect end users and systems integrators to actually conduct research, at least initially, but they should at least participate on advisory boards or otherwise contribute actively to the research process. Over time, they should play a more active role in funding research and perhaps even in conducting it. The differences between them and the research community in culture and approach to the use of time and resources may make it difficult to engage them, but once the benefits are demonstrated, the committee expects that end users and system integrators will become more supportive of such work.
The participation of federal agencies that are major users of IT systems is also important. Government systems are, by definition, large-scale artifacts of public interest; they are, by observation, artifacts that present problems and opportunities that challenge the state of the art.6 The difficulties experienced in getting these systems right show the limitations of current technology and of the skill base in industry. Government agencies would save money and improve their productivity and service quality if there were a better understanding of ways to reliably and efficiently design, operate, maintain, and upgrade large-scale systems and social applications of IT. Research based on government systems would undoubtedly improve the knowledge base for private-sector systems as well. Designing a program for agencies that lack the funding, personnel, and orientation to research will be more difficult than design-
ing conventional research programs, but again, the Digital Government program, which has forged links between many agencies, should be leveraged, at least to explore what is possible and to initiate experiments. The coordinating structure of Digital Government, the Federal Information Systems Advisory Committee (FISAC), is charged with building bridges between the IT research community and government agencies, and its activities should be encouraged and strengthened. Experimentation will be necessary, because differences among agencies will demand different mechanisms for research linkages. Some activities under FISAC's purview (e.g., those associated with universal access) hold great potential for social applications research.
An important element of efforts to better understand the social applications of IT—which are based on large-scale systems—is researcher access to operational IT systems for purposes of observation, data collection, and analysis. Many research proposals should include plans for gaining such access, which could come through the participation of end-user organizations, as described above. Access to operational systems for testing and demonstrating proposed solutions would also be desirable but may not be feasible because of the need to keep many operational systems running almost continuously and the uncertainties inherent in introducing new elements into an existing system. Accordingly, researchers will need to develop plans for testing new research results in more limited testbed systems, in which outcomes can be evaluated before the solutions are deployed in operational systems.
Although end users and systems integrators must be encouraged to participate in research on social applications, organizations more experienced in managing IT research (such as DARPA and NSF within the federal government) will need to oversee such programs, as is being done in the Digital Government initiative. In the committee 's judgment, these organizations are the best qualified to ensure that the research maintains a focus on long-term, fundamental results rather than devolving into applied research and development and targeting the needs of specific end users. Early experience with the Digital Government program demonstrates that what end users perceive as research is often seen by the IT research community as development. A strong focus on fundamental research is necessary if leading IT researchers are expected to participate in the research program and if meaningful, broadly applicable insight is to be gained.
Finally, the work funded in this area should pursue a fundamental knowledge of the interaction between IT and the context in which it is deployed. A central element of the overall program must be to inform IT research as well as understand the social and economic implications of IT applications. One challenge will be to establish effective peer review
mechanisms to ensure the quality of the research. Because reviewers will need to be drawn from the range of disciplines encompassed by an individual proposal, programs in specific areas need to be announced beforehand (such as digital libraries) so that appropriate sets of reviewers can be assembled to evaluate groups of related proposals.
Recommendation 4. The Bureau of the Census should work with the National Science Foundation to develop more effective procedures for classifying data on federal and industry investments in information technology R&D that better account for the dynamic nature of the industry.
Better data on industry investments in R&D would allow for better-informed policymaking about research support, especially levels of funding. Such data are currently gathered for the NSF by the Census Bureau, but they are highly inconsistent from one year to the next, owing largely to frequent reclassifications of companies into, out of, and among the industry sectors most closely allied with IT.7 Significant improvements could probably made without increasing the cost of data collection by simply aggregating the data in a more consistent way from one year to the next. Admittedly, the dynamic nature of the IT industry and waves of mergers and acquisitions among major players can make company classification difficult, yet existing procedures for classification appear to give priority to accurate reporting for a given year rather than across years. Moreover, they tend to categorize a firm according to the industry classification that best describes the composition of its domestic payroll rather than its main source of revenues. As firms move production overseas or change the workforce mix through mergers, acquisitions, spin-offs, and outsourcing, their industry classifications can change dramatically, producing equally dramatic changes in reported research.8 Large, diversified IT companies would appear to be most prone to reclassification because their lines of business span several industry sectors. They also tend to have the largest R&D budgets, so reclassifications can have a large effect on the reported, aggregate statistics.
Efforts are under way to replace the Standard Industrial Classification codes with a set of North American Industrial Classification System codes that will provide better coverage of the information and information technology industries.9 This change is unlikely to have a significant effect on the quality of industrial R&D data in the IT industries unless improved methods are developed for classifying firms into the new categories and ensuring greater consistency in such classifications over time. Attempts to improve the collection and reporting of IT R&D in industry should include efforts to develop more robust procedures for classifica-
tion that are less sensitive to small changes in company structure. The goal should be to facilitate the compilation and reporting of more consistent sets of data series describing past, present, and future investments in IT R&D. These issues will become even more important as the nation continues its course toward an information economy and IT becomes more pervasive.
RECOMMENDATIONS FOR UNIVERSITIES
Whereas government and industry will be the primary sources of funding for an expanded IT research agenda, universities will be primary sites for conducting that research. They will also be the primary institutions for educating the next generation of researchers, developers, and users of large-scale systems and social applications. Research is closely tied to education, forming part of the educational process of graduate and some undergraduate students and generating additional knowledge that cannot be conveyed through more conventional course work. Universities have a long tradition of conducting fundamental research on IT that has contributed to innovation in industry. This tradition must be extended to the problems of large-scale systems and social applications. This will not be an easy task in today' s environment. Universities are not presently set up to address these important areas, and they will need to change if they are to help make IT better. Many computer science and engineering departments are caught up in Internet-related technologies, which will make it hard to shift attention to issues of large-scale systems and social applications when so much can be done using a traditional component-oriented approach. Nevertheless, there may be pockets of interest that can be motivated in the near term. In the longer term it may become more apparent that many of the most successful Internet-based innovations are those that pursue social applications—which face the challenges of large-scale IT systems. Several steps can be taken to shift attention to large-scale systems and social applications.
Recommendation 5. Universities should take steps to increase the ability of faculty members and students to participate in interdisciplinary research related to information technology and research on large-scale systems.
Given the tendency of faculty members and other researchers to orient their research to conventional disciplinary pursuits, they will not be able to pursue interdisciplinary research focused on social applications of IT without additional incentives and the removal—or relaxation —of existing barriers to interdepartmental collaboration. Increases in the amount
of research funding available for interdisciplinary work (as recommended above) would be an important means of achieving both objectives, but universities should also implement changes that will enable their faculties to work at the intersection of multiple disciplines and help research staffs to win interdisciplinary research funding. The steps they take will vary from one university to another, reflecting the relative strengths and weakness of different departments and the relationships among them, but several areas are ripe for examination and should be addressed, as noted below. Efforts to enable university research in large-scale systems and the social applications of IT will also enrich educational opportunities in these areas, enhancing the nation's ability to conduct research on IT and make better use of it.
Recommendation 5.1. Universities should ensure that their hiring, review, and tenure processes are aligned with the interdisciplinary nature of the research that this report recommends.
As noted in Chapter 4, one barrier to interdisciplinary research in universities is the difficulty of hiring, promoting, and granting tenure to faculty members who pursue interdisciplinary work. Their work is often viewed through a disciplinary lens that does not properly appreciate work that crosses disciplinary boundaries. Within the computer science community, work on applications is seldom seen as a valuable or respectable target of research. Any of a number of mechanisms could be used to address this problem, depending on the university. Some universities have established interdisciplinary schools, divisions, or departments that can hire and promote faculty members who specialize in multiple interdisciplinary areas, but this is not necessarily the only solution. These formalized structures also contribute to the university's educational missions that help train future generations of students and IT-related workers. Other universities have found ways to establish tenure committees and review processes that more accurately assess interdisciplinary research, even if that research takes place within a traditional academic department. Computer science departments or university administrators could also promulgate policies stating that research in interdisciplinary or applications-oriented areas will be given full consideration in the promotion and tenure process.10 Any one of these approaches—or a combination of them—could be appropriate. The key is to ensure that disciplinary-based review procedures do not disfavor work at the intersection of disciplines while simultaneously ensuring the quality of research. All interdisciplinary research cannot be of high quality, and quality control is especially important in fields with growing budgets.
Recommendation 5.2. Universities should encourage closer ties between faculty and student researchers and their counterparts in industry, especially in companies with pressing needs to resolve problems of large-scale systems and social applications of informa tion technology.
The forging of stronger ties between universities and companies engaged in building or operating large-scale systems or social applications of IT could help university researchers to gain (1) greater exposure to, and insight into, the challenges faced by these types of organizations and (2) access to some elements of the research infrastructure. One way to strengthen such ties is to appoint representatives from such companies to departmental review committees and advisory boards. Another mechanism, which should be strongly encouraged, is to promote sabbaticals and internships to facilitate direct interaction between university researchers and industry.
One of the most effective mechanisms for transferring knowledge and expertise is to have people work together. Indeed, a number of leading researchers require all their Ph.D. students without industrial experience to spend one summer or semester in industry to gain what they consider invaluable experience.11 The problem with pursuing such an approach in the areas of large-scale systems and social applications is that few IT end-user companies have internship programs for IT researchers (most IT companies with research labs have such programs). University administrators should work with such companies to establish trial internship programs that could be expanded if they prove successful. Officials will need to work with funding agencies and foundations to secure funds for such programs, at least until the end-user organizations become convinced of their value and provide additional support themselves. University administrators should also promote interest in these programs among faculty and students. Some faculty members may be hesitant to encourage students to take time away from their university laboratory, especially if their absence would create staffing problems for critical research projects. But if the internship programs are designed to benefit the students, and spending time in industry is important for students, then such issues as time away from professors and staffing research projects should be planned for in advance. If the internships are as successful as this committee envisions, their contribution to the quality and effectiveness of IT research will quickly become apparent.
Recommendation 6. Senior faculty members should take the lead in pioneering research on large-scale systems and social applications of information technology.
Universities need to find ways in which junior faculty can work on important, interesting problems. In most cases, junior faculty members are at a disadvantage in pioneering new research in interdisciplinary areas. Concerns about gaining tenure within the existing disciplinary structure of most universities can dissuade them from proposing revolutionary research ideas early in their careers, as can the difficulties inherent in securing federal funding for interdisciplinary research. The recommendations above are intended to alleviate some of these problems, but they do not address them all. The trend away from large grants to university research labs and toward smaller grants to individual investigators has further limited the ability of senior faculty to support innovative work by junior faculty. The expanded research programs outlined above on large-scale IT systems and social applications of IT could address part of this problem by making additional funding available, but the leadership of senior faculty members will also be important in legitimizing new research areas. By building on the vision of research in large-scale systems and social applications that is laid out in this report and communicating that vision to the research community and to funding agencies and universities, senior faculty will help create more opportunities for junior faculty to strike out in these new directions.
RECOMMENDATIONS FOR INDUSTRY
Industry is an important partner in any attempt to expand the scale and scope of IT research. It both funds and conducts IT research, and, ultimately, it must harvest the fruits of research to develop new products, processes, and services for clients. In the eyes of some, industry is already the leader in work on large-scale systems and social applications because it is intimately involved in developing systems to support innovative applications in commerce, publishing, health care, education, and many other fields. Yet, industry's development activities may have exceeded its research capacity, contributing to the deployment of systems that are not well understood. As noted in Chapter 2, most industry research is concentrated in the vendor community—the companies that produce IT components (e.g., hardware, software, and devices). These companies must continue to invest in research that enables continued progress in IT, but they and other organizations that use IT need to become more involved in research that addresses large-scale systems and social applications.
Recommendation 7. Organizations that are significant end users of information technology systems should actively seek opportunities to engage in IT research.
As IT becomes increasingly intertwined in the operations of end-user organizations in the public and private sectors, there is an increased need to bring such organizations into the research process. These organizations can contribute to the IT research base through any of a variety of mechanisms, some of which were outlined above: by funding research in other organizations (such as universities), by conducting research themselves (although most end users currently lack this capability), by forming industry consortia, or by providing input into ongoing research initiatives to ensure that the right problems are being addressed and that the solutions are viable. Participation in the IT research process will make them better-informed consumers of IT goods and services and will help guide IT research into areas that are well matched to particular end-user needs and problems. Incentives for such organizations to invest in IT research are growing as they become increasingly dependent on IT systems to carry out their missions, yet most such organizations lack the resources and expertise needed to manage IT research programs conducted either internally or externally.
The committee recognizes the difficulty of engaging end-user organizations in a productive way but believes that greater collaboration between them and IT researchers is critical to ensuring that IT evolves in a way that will meet real-world needs and address the problems faced by organizations reliant on such systems. The federally sponsored programs recommended above are intended to help bridge the gap and introduce end users to the processes of IT research, but these programs should not be the only mechanisms through which end users engage IT researchers. End users should consider other activities, such as supporting university research that is applicable to their needs, serving on the advisory boards of IT research groups in universities and industry, supporting internships for students in IT-related academic programs, and forming external research groups that monitor work in the IT research community and serve as liaisons between companies and IT researchers.
This process needs to be viewed as a long-term, evolutionary one that can grow into more active participation in research over time. Eventually, end-user organizations may fund research or engage in research programs with vendors and academic research groups. Before this can happen, end users need to become more familiar with the IT research community. Although the incentives for such activities may not be immediately obvious, the committee believes that such engagement will ultimately prove beneficial to end-user organizations in terms of an improved ability to make IT systems perform the needed functions on time, within budget, and with high reliability.
Recommendation 8. Information technology companies with established R&D organizations should develop mechanisms for engaging end users more actively in the research process.
Vendors need to be involved in any effort to expand the scale and scope of IT research. They have an established tradition of conducting problem-oriented research and have a better understanding of customer needs than do most university researchers or government program managers. For the foreseeable future, vendors will remain the main funders of IT-related R&D. In recent years, industry research has become more applied as companies attempt to link research efforts to more clearly defined areas of corporate interest. Work on large-scale systems and social applications may give them an opportunity to increase the amount of fundamental research conducted while maintaining—or even increasing—the applicability of their research to user needs. The better their understanding of the fundamentals of large-scale IT systems and social applications, the better they should be able to develop systems and applications for their clients.
IT companies should explore a range of options for pursuing more interdisciplinary and systems-related research while continuing their research on IT components. Several mechanisms could be used. Companies could establish programs similar to IBM Corporation's First-of-a Kind program to link their researchers more closely to cutting-edge end users with interesting problems that call for research. Doing so will not only align research more closely with customer needs, making it more valuable in the long term, but will also create better-educated customers who can interact more effectively with researchers. Or they could fund joint research with end users in universities or industry, as Microsoft Corporation is doing in its educational venture with the Massachusetts Institute of Technology (MIT).12 Another approach is to increase company contacts with, and support for, university researchers by allocating more resources for students and faculty to spend time in industrial facilities and for industrial researchers to visit university laboratories. Money is a significant obstacle to such interactions, as is the argument that staff members have more compelling things to do. Research managers and corporate executives must also provide leadership to ensure that these relationships are established and to maintain a commitment to making them work.
Expanding IT research in scale and scope will be essential to ensuring that society captures the full benefits of the investments it has already made in IT. Continued progress is needed in the areas of research that
enabled the IT revolution unfolding today and that are the foundation for the nation's transition to an information economy. Additional work on large-scale systems and the social applications of IT will allow society to transform a range of interactions in all walks of life. The recommendations in this report emphasize putting in place the processes that will expand the research agenda and bring the needed range of expertise to bear on problems that have plagued large-scale systems and social applications. They reflect the uncertainties surrounding the course that future IT development and deployment will take, as well as the differences in the abilities of the various stakeholders to participate productively in the process. These initial, if tentative, steps need to be taken so that IT can better serve society's growing range of needs as it enters the twenty-first century.
Committee for Economic Development (CED). 1998. America's Basic Research: Prosperity Through Discovery. Committee for Economic Development, New York.
Computer Science and Technology Board (CSTB), National Research Council. 1989. Scaling Up: A Research Agenda for Software Engineering. National Academy Press, Washington, D.C.
Computer Science and Telecommunications Board (CSTB), National Research Council. 1992. Computing the Future: A Broader Agenda for Computer Science and Engineering. National Academy Press, Washington, D.C.
Computer Science and Telecommunications Board (CSTB), National Research Council. 1995. Evolving the High-Performance Computing and Communications Initiative to Support the Nation's Information Infrastructure. National Academy Press, Washington, D.C.
Computer Science and Telecommunications Board (CSTB), National Research Council. 1996a. Continued Review of the Tax Systems Modernization of the Internal Revenue Service. Final Report. National Academy Press, ashington, D.C.
Computer Science and Telecommunications Board (CSTB), National Research Council. 1996b. Computing and Communications in the Extreme: Research for Crisis Management and Other Applications. National Academy Press, Washington, D.C.
Computer Science and Telecommunications Board (CSTB), National Research Council. 1999a. Funding a Revolution: Government Support for Computing Research. National Academy Press, Washington, D.C.
Computer Science and Telecommunications Board (CSTB), National Research Council. 1999b. Trust in Cyberspace. National Academy Press, Washington, D.C.
Council on Competitiveness. 1996. Endless Frontier, Limited Resources: U.S. R&D Policy for Competitiveness. Council on Competitiveness, Washington, D.C., April.
President's Information Technology Advisory Committee (PITAC). 1999. Information Technology Research: Investing in Our Future. National Coordination Office for Computing, Information, and Communications , Arlington, Va., February. Available online at <http://www.ccic.gov/ac/report/>.
1. See, for example CSTB (1992, 1995, 1999a), Council on Competitiveness (1996), and CED (1998).
2. Several of the more high-profile failures have resulted in hundreds of millions of dollars being spent over several years. For instance, the Federal Aviation Administration has spent $42 billion to modernize the air traffic control system over the past two decades, and the system has still not been completed (see Chapter 3).
3. See Chapter 3 for a more in-depth discussion of these topics.
4. The Next Generation Internet program encompasses three related efforts: research on, and development of, new networking technologies; development of revolutionary applications that take advantage of enhanced networking capabilities; and deployment of several testbed networks across which new technologies can be deployed and revolutionary applications can be run. Additional information is available online at <www.ngi.gov>.
5. The Digital Libraries Initiative is a multiagency initiative that, in its second phase, will pursue research related to the development of the next generation of digital libraries, both to advance the use and usability of globally distributed, networked information resources and to encourage existing and new communities to focus on innovative applications areas. The initiative attempts to stimulate the partnering arrangements needed to create next-generation operational systems in areas such as education, engineering and design, Earth and space sciences, biosciences, geography, economics, and the arts and humanities. Its sponsors include the NSF, DARPA, the National Library of Medicine, the Library of Congress, the National Endowment for the Humanities, the National Aeronautics and Space Administration, and the Federal Bureau of Investigation. The research centers on topics such as human-centered computing, content, and systems, as well as on testbeds and applications. Support is provided for both individual investigator grants and multidisciplinary research groups. Additional information about the program is available online at <www.dli2.nsf.gov>.
6. It is also true that federal agencies have had many successes in creating new computer systems, successes that do not get as much publicity as the problems. But these successes are more a testimonial to the skill and perseverance of federal IT managers than a reason to praise the available knowledge base.
7. These sectors include those defined by the following Standard Industrial Classification (SIC) codes: 357, office, computing, and accounting machines; 366, communications equipment; 367, electronic components (including semiconductor devices); 737, computer and data processing services; and 48, communications (e.g., telephone and other communications services).
8. Linda Cohen, a member of the study committee, and Jerry Sheehan, a member of the CSTB staff, are examining this issue in greater detail. They will produce a summary paper on their findings late in the year 2000.
9. On April 9, 1997, the North American Industry Classification System (NAICS) became the new standard code system to describe business establishments and industries, replacing the Standard Industrial Classification (SIC) codes. This new system will be used by the U.S., Mexican, and Canadian governments to collect and distribute statistical information.
10. Recommendation 5.1 echoes a recommendation from an earlier CSTB report that also called for an expanded research agenda for computer science. See CSTB (1992).
11. David Patterson, University of California at Berkeley, personal communication, April 6, 2000.
12. As noted in Chapter 4, Microsoft announced a partnership with MIT in 1999 to pursue educational technologies. Microsoft is investing $25 million in the effort, and projects will be managed by a steering committee consisting of equal numbers of members from Microsoft Research and MIT.