During the last 60 years we have seen the beginning of a major technological revolution, the Information Revolution. IT has spanned large new economic sectors and has, over a long period, doubled the rate of increase in labor productivity in the U.S.1,16 Over two-thirds of job openings in science and engineering in the coming decade are in IT.12 Intellectual property, rather than physical assets, has become the main means of production: control over intangibles (such as patents and copyrights) are at the forefront of the national and international business agenda;6,23 investment by industry in intangible assets has overtaken investment in tangible means of production.7,19
The information revolution is far from having run its course: “machine-thought” has not yet replaced “brain-thought,” to the extent that “machine-made” has replaced “hand-made.” One can be confident that the use of digital technologies will continue to spread; that more and more workers will move from the physical economy to the information economy; and that people will spend more and more of their work and leisure time creating, manipulating, and communicating information.
The fast evolution of IT motivates a periodic reexamination and reorganization of computing and information (C&I) research and education in academia. We seem to be in one such period. Many universities have established or expanded schools and programs that integrate a broad range of subdisciplines in C&I; and NSF is affecting the scope of research and education in C&I through the creation of programs such as the Cyber-Enabled Discovery and Innovation (CDI) and Pathways to Revitalized Undergraduate Computing Education (CPATH) programs.21,22
The fast evolution of IT motivates a periodic reexamination and reorganization of computing and information (C&I) research and education in academia.
I strongly believe that C&I is one broad discipline, with strong interactions between its various components. A coherent view of the whole must precede any discussion of the best ways of dividing it into subdisciplines. The dominant discourse in our community should be about building a coherent view of the broad discipline, building bridges between its constituents, and building bridges to other disciplines as we engage in interdisciplinary research. I hope this column will contribute to these goals.
C&I Is a Use-Driven Research Discipline
I am discussing in this column the broad field of Computing and Information Science and Engineering (CISE): the study of the design and use of digital systems that support storing, processing, accessing and communicating information. To prevent possibly misleading connotations, I shall call this broad field Computing and Information (C&I).
We still seem to be debating whether computer science is science, engineering, or something unlike any other academic discipline (see, for example9,11). The debate is often rooted in a linear view of science and engineering: Scientists seek knowledge, for knowledge sake; through a mysterious process, this knowledge turns out to have practical consequences and is picked up by applied scientists, next engineers, and then used to develop better technologies. This view encourages an implicit value system whereby science is seen a higher call than engineering.
Donald Stokes, in his book Pasteur’s Quadrant,25 leads a powerful attack against this simplistic view of science. He points out that, over the centuries, fundamental research has been often motivated by considerations of use—by the desire to implement certain processes and achieve certain goals—not (not only) by the desire to acquire knowledge for knowledge’s sake. His paradigmatic example is Pasteur, who founded modern microbiology, driven by the practical goal of preserving food.
According to Stokes, research should be described as a two-dimensional space, as shown in Figure 1. Stokes further argues that “Pasteur’s Quadrant,” namely use-inspired basic research, is increasingly prevalent in modern research institutes. The argument of Stokes strongly resonates with schools of engineering, or computer science. Most of their faculty members pursue scientific research that has a utilitarian justification; their research is in “Pasteur’s Quadrant.”
Any engineering department in a modern research university is a science and engineering department. This is often indicated by the department’s name: Material Science and Engineering, Nuclear Science and Engineering, or even Engineering Science (at Oxford University). Figure 2 describes the research activities in such a department.
Faculty members perform basic use-inspired or applied research related to the applications of their discipline. The foundations guiding this research and constraining the engineering design space are natural sciences—mostly physics.a The practical goal of their research is to enable the production of better artifacts or better processes. The design of and experimentation with prototypes often is an essential step in the transfer of knowledge from research to practice, as they provide a proof of concept, a test and validation for theories, and a platform to experiment with design alternatives. I believe it is the richness of the feedback loops between research and practice and between basic research and applied research that best characterizes top engineering departments.
The diagram in Figure 2 describes not only engineering departments, but also other use-oriented disciplines such as medicine or agriculture. Furthermore, concern about impact and use, and research in “Pasteur’s Quadrant,” are increasingly prevalent in science departments, be it life sciences, social sciences, or physical sciences. Only a few purists would claim that departments are weakened by such concerns.
The diagram in Figure 2 clearly applies to C&I. Our discipline is use-inspired: We want to build better computing, communication, and information systems. This occasionally motivates use-inspired basic research (for example, complexity, cryptography), and often involves applied research (such as architecture, databases, graphics). The design and experimentation with prototypes is essential in system research. C&I scientists use scientific methods in their research;8,10 and there is a continued back and forth between basic and applied research and between academic research and the development of digital products and services by industry.
C&I Needs Broader Foundations
C&I was lucky to develop early on mathematical abstractions that represented important constraints on computing devices, such as time and space complexity; this enabled C&I to develop useful artifacts while being fully contained within the confines of mathematics: The early development of algorithms, programming languages, compilers or operating systems required no knowledge beyond C&I and its mathematical foundations.
Mathematics continues to be the most important foundation for C&I: The artifacts produced by C&I researchers and practitioners are algorithms, programs, protocols, and schemes for organizing information; these are mathematical or logical objects, not physical objects. Algorithms, programs or protocols are useful once realized, executed or embodied in a physical digital device; but they are mostly studied as mathematical objects and the properties studied do not depend on their physical embodiment. Indeed, one might call much of C&I “mathematical engineering”b: it is focused on the creation of new mathematical objects under constraints, such as low time and space complexity for discrete algorithms, good numerical convergence for numerical algorithms, or good precision and recall for classifiers; the difference between mathematics and “mathematical engineering” is precisely the emphasis on such constraints.
As technology progresses, new constraints need to be considered. For example, time complexity is increasingly irrelevant when communication (to memory, disk, and network) replaces computation as the main performance bottleneck, and when energy consumption becomes the critical constraint. New technologies that will take us “Beyond Moore’s Law” (quantum computing, molecular computing) will require new mathematical abstractions.
Part of C&I, namely computer engineering, has always been concerned with the interplay between the mathematical abstractions and their physical embodiment. In addition to mathematics, physics is foundational for this specialty, and will continue to be so. Physics is also important for cyber-physical systems that directly interact with their physical environment.
I believe, however, that physical constraints are a small fraction of the constraints relevant to the design of C&I systems. For example, software engineering research has strived for decades to define code metrics that represent how complex a code is (hence, what effort is required to program or debug it)—with limited success. Such a code metric would measure how difficult it is for a programmer to comprehend a code. But this is a cognitive issue: It is highly unlikely that one can develop successful theories on this subject without using empirically validated cognitive models that are based on our best understanding of human cognition. Unfortunately, traditional software engineering research has not been rooted in cognitive sciences.
Cognitive, cultural, social, organizational, and legal issues are increasingly important to engineering, in general.5 This is a fortiori true for C&I. In the early days of computing, only few people interacted directly with computers: the psychology of programmers or users could be ignored without too much inconvenience: these few people would adapt to the computer. Today, the situation is vastly changed: Billions of people interact daily with digital devices and C&I systems become intimately involved in many cognitive and social processes; it is not possible anymore to ignore the human in the loop. Indeed, interesting research increasingly occurs at the intersection of the social and the technical: One may well argue that the essential insight that enabled efficient Web search and led to the creation of companies such as Google is that the structure of the Web carries information about the usefulness of Web pages—a socio-technical insight. Progress in graphics and animation increasingly requires an understanding of human vision: otherwise, one makes progress in quality metrics that have low correlation to the subjective quality of an image; examples can be easily multiplied.
Another important aspect of the evolution of our field is the increasing importance of applications. Precisely because software is so malleable and universal, one can develop very specialized systems to handle the needs of various disciplines: computer-aided design, medical imaging, DNA matching, Web auctions—these are but a few examples of application areas that have motivated significant specialized C&I research. Such research cannot be successful without a good understanding of the application area.
This suggests a new view for the organization of C&I that is described in Figure 3: Mathematics is no longer the only foundation. For those working close to hardware or working on cyber-physical systems, a good foundation in physics continues to be important. An increasing number of C&I research areas (such as human-computer interaction, social computing, graphics and visualization, and information retrieval) require insights from the social sciences (cognitive psychology, sociology, anthropology, economics, law, and so forth); human subject experiments become increasingly important for such research. At a more fundamental level, the development of artificial cognitive systems provides a better understanding of natural cognitive systems—of the brain and its function; and paradigms borrowed from C&I become foundational in biology. Insights from neuroscience provide a better way of building artificial intelligent systems and biology may become the source of future computing devices. Finally, research in C&I is strongly affected by the multiple application areas where information technology is used (such as science, humanities, art, and business), and profoundly affects these areas.
Organizational Implications
Similar to a school of medicine, a college of agriculture, or an engineering department, I believe the correct organizational principle for a use-driven research area such as C&I is not common foundations, but shared concerns about the use of C&I systems. The view illustrated in Figure 3 does not imply that each C&I researcher needs to be an expert in all core sciences or application areas. Rather, it implies that C&I researchers with different foundational knowledge and knowledge of different application domains will often need to work together in order to design, implement, and evaluate C&I systems and provide students with the education needed to do so.
The broad, integrated view of C&I is reflected at the NSF in the name of the Directorate for Computer and Information Science and Engineering (CISE). It is no surprise these days to find a linguist, anthropologist, or economist in a research lab at Microsoft or Yahoo. Some U.S. universities (including Carnegie-Mellon, Cornell, Georgia Tech, Indiana, Michigan, and the University of California at Irvine) are establishing or expanding schools or colleges that bring under one roof computer science, information science, applied informatics (C&I research that is application domain specific) as well as interdisciplinary research and education programs. These universities are still a minority. The broad, inclusive model is common in Japan (University of Tokyo, Kyoto University, Tokyo Institute of Technology, Osaka University), and is becoming more prevalent in the U.K. (Edinburgh, Manchester).
We can and should develop an environment where no scientist has an incentive to withhold information.
While organization models will differ from university to university, it is essential that all C&I units on a campus develop an integrative view of their field, and jointly develop coordinated research and education programs. This may require a change of attitude from all involved. Many cognitive and social aspects of system design are not amenable to quantitative studies; however, the engineering culture is often suspicious of social sciences and dismissive of qualitative sciences. Conversely, the importance of prototypes and artifacts is not always well appreciated outside engineering.
Undergraduate Curriculum
I discussed in the previous section the increasing variety of C&I research. In addition, there is a tremendous diversification of the professional careers in IT. Less than half of students who graduated in computer science in 19921993 were employed in traditional computer science professions 10 years after graduation (compared for 57% in engineering and 69% in health sciences).4 In many computer science departments, more than half of the students graduating with bachelor’s degrees are hired by companies in finance, services, or manufacturing, not by IT companies; this is where most of the growth in IT jobs is expected to be.12 The Bureau of Labor Statistics tracks a dozen different occupations within computing12 (although its categories are somewhat obsolete). A recent Gartner report20 suggests the IT profession will split into four distinct professions: technology infrastructure and services, information design and management, process design and management, and relationship and sourcing management.
These trends imply an increasing diversification of C&I education. Currently, ABET accredits three different types of computing programs; ACM has developed recommendations for five curricula. Many schools experiment with more varied majors and interdisciplinary programs—in particular, Georgia Tech.17 This evolution could lead to an increasing balkanization of our discipline: It is fair to assert that we are still more concerned with differentiating the various programs than defining their common content. In particular, should there be a core common to all programs in C&I?
To clarify: A common core is not about what every student in C&I must know: most of the specific knowledge we teach will be obsolete long before our students reach retirement age. A common core is about C&I “education,”c not about C&I know-how. It is about educating students in ways of thinking and problem solving that characterize our community and differentiate us from other communities: a system view of the world, a focus on mathematical and computational representations of systems, information representation and transformation, and so forth. The selection of courses for the core will not be based (only or mostly) on the usefulness of the facts taught, but on the skills and concepts that are acquired by the students.
I believe such a common core is extremely important: It is, to a large extent, what defines a discipline: You can expect a student of physics to take a sequence of physics courses that start with mechanics and end with quantum physics. This is not necessarily what those students will need in their future careers; but those courses define the physics canon. If we take ourselves seriously as a discipline, we should be able to define the C&I canon. Like physics, this core should be concise—say four courses: A common core does not preclude variety and specialization in junior and senior years.
Eating Our Own Dog Food
IT has a profound impact on the way the information economy works. It can and should have a profound impact on the operation of universities that are information enterprises par excellence. The C&I academic community can and should have a major role in pioneering this change. We should be ahead of the curve in using advanced IT in our professional life, and using it in ways that can revolutionize our enterprise. I illustrate the possibilities with a few examples here.
William J. Baumol famously observed that labor productivity of musicians has not increased for centuries: it still takes four musicians to play a string quartet.2 This has become known as “Baumol’s cost disease:” Some sectors are labor intensive, require highly qualified personnel, and see no increases in labor productivity, due to improved technology. This is true for higher education: As long as a main measure of the quality of higher education is the student/faculty ratio, teaching productivity of faculty cannot increase; as long as faculty salaries keep up with inflation, the cost of higher education will keep up with inflation.d Such a situation will lead to the same pressures we see now in the health sector, and will force major changes. IT is, in many service sectors, the cure for Baumol’s cost disease;27 can it be in higher education?
IT often cures Baumol’s cost disease not by increasing labor productivity, but by enabling a cheaper, replacement service. It still takes four musicians to play a string quartet, but digital recording enables us to enjoy the music where and when we want to hear it. ATMs replace bank tellers, Internet shopping replaces sales clerks. The convenience of getting a service where and when we want it, and the lower cost of self-service, compensate for the loss of personal touch. To many of our students, the idea that one must attend a lecture at a particular place and time in order to obtain a piece of information chosen by the lecturer is as antiquated as pre-Web shopping. Increasingly, students will want to obtain the information they need when and where they want it. An increasing shift to “self-service” education that is “student pull” based, rather than “lecturer push” based, may well be the cure to Baumol’s cost disease in higher education, as well as the cure to the depressing passivity of many students.
“Self-service” education need not imply a lack of social interaction. The study of Richard J. Light, at Harvard, indicated that participation in a small student study group is a stronger determinant of success in a course than the teaching style of the instructor.20 Rather than focusing on the use of IT to improve the lecture experience, we should probably focus on the use of IT and social networking tools to make individual and group self-study more productive by multiplying the interaction channels between students and between students and faculty.15
As the half-life of knowledge grows shorter, it becomes less important to impart specific knowledge to students (and to test them on this knowledge) and more important to teach them how to learn, how to identify and leverage sources of knowledge and expertise, and how to collaborate with experts in other areas, creating collective knowledge. Yet our education is still strongly focused on acquiring domain-specific individual knowledge; and students mostly collaborate with other students that have similar expertise. Projects and practicums that involve teams of students from different programs, with different backgrounds, could refocus education so as to train more foxes and fewer hedgehogse—a change I believe will benefit many of our students. Such collaborative learning-by-doing empowers students, increases motivation, improves retention and teaches skills that are essential for success in the information society. A skillful use of IT technology, both for supporting course activities and for assessing teaching and learning, can facilitate this education style.f
IT changes the way research is pursued: For example, it enables citizen science projects where many volunteers collect data. Such projects have become prevalent in environmental sciences24 and are likely to have a large impact on health sciences. They not only provide researchers with data that cannot be obtained otherwise but also change in fundamental ways the relation of the scientist to the object of study. The volunteers are unlikely to be motivated by pure scientific curiosity; they want the research they participate in to have an impact—save the environment or cure cancer. The researcher that uses their data has an implicit or explicit obligation to use the data collected for that common purpose and not use it for other purposes. Research becomes engaged and obligated to a large community.17
IT enables the fast dissemination of scientific observations and results. Research progresses faster if observational data and preliminary results are shared as quickly and as broadly as possible. One obstacle to such unimpeded sharing is that academic careers are fostered by the publication of polished analyses, not by the publication of raw data or partial results: Research groups tend to hold on to their data until they can analyze it and obtain conclusive results. Better ways of tracking the provenance of data used by researchers and the web of mutual influences among researchers would enable to track the impact of contributions other than polished publications and develop a merit system that encourage more information sharing. We can and should develop an environment where no scientist has an incentive to withhold information.
C&I has been, for years, an amazingly vibrant, continuously renewing intellectual pursuit that has had a profound impact on our society. It has succeeded being so by continuously pursuing new uses of IT and continuously adjusting disciplinary focus in research and education so as to address the new problems. This fast evolution must continue for our discipline to stay vital. IT will continue to be a powerful agent of change in our society and, to drive this change, we must continuously change and strive to change our academic environment.
Join the Discussion (0)
Become a Member or Sign In to Post a Comment