The challenges of achieving a sustainable society are truly global, with complex interdependencies that affect risk assessments, technical and social opportunities for solutions, and economic and political feasibility. No one field or discipline on its own could possibly be expected to “solve” even one aspect of this problem. However, information is central to making progress on many fronts. Thus computer science (CS)—which couples information and innovation—is vital to sustainability. For computer science to play its part in meeting global sustainability challenges, priority should be given to research that addresses one or more important sustainability challenges (examples were described in Chapter 1) and that offers significant impact. This impact may be direct, or it may be through game-changing contributions that offer significant leveraging opportunities for other domains. In either case, priority should be placed on opportunities to address the sustainability challenge to a tangible degree.
This chapter explores some of the potential impediments within the field of computer science to making significant progress on issues pertinent to sustainability. It considers how to bridge the gap between the traditional research quest for universality1 and the imperative to have a specific impact on sustainability challenges. The chapter is aimed primarily
1The committee uses the term “universality” to encompass the related notions of generalizability (solutions that are amenable to relatively straightforward abstractions in order to address more general versions of a given problem) and breadth (solutions that can be revised to be applicable to broad problem domains and spaces). The quest for universality captures the traditional CS research goals of abstractability and broad applicability.
at the CS research community—including both researchers and funders. First, it discusses some of the fundamental aspects of CS research and computational thinking and how these aspects are also critical to the sustainability problem space. Then it explores the challenge of universality and emphasizes that a bottom-up approach is not only necessary in the sustainability space but also has precedent in many other areas of deep computer science. It describes the connection between universality, bottom-up approaches, and sustainability. It then offers suggestions on how to structure research to promote meaningful impact on sustainability. Finally, the chapter identifies methodological opportunities for optimizing research outcomes and impacts.
Chapter 2 highlighted the centrality of data and information to sustainability. Given this centrality, computer science and information technology (IT) are essential to meeting sustainability challenges. The challenge for IT experts and CS researchers is in ensuring that technologies and approaches represent usable, appropriate solutions; that they are highly effective; and that they take advantage of the deepest and most powerful insights that can be brought to bear. IT has been and continues to be a critical enabler of progress in vast arenas of society. Sustainability is no exception: IT offers a powerful tool to assist in addressing sustainability challenges.
Moreover, fundamentals of the computer science field itself offer unique and important contributions to sustainability. To name just a few such fundamentals, consider abstraction design, algorithms, operating systems and layering, real-time systems, machine learning, human computer interaction (HCI), and databases. For instance, the very notion of queryable structured data is at the heart of much of computer science; at the same time strides are being made to cope with the vast amounts of unstructured data now available. Given the scope and scale of sustainability challenges along with the vast amounts of relevant data, the structuring and understanding of these data present many challenges. The lens of computational thinking is essential to solving many complex problems,2 and there are key opportunities within computer science that are clearly
2See National Research Council, Report of a Workshop on the Scope and Nature of Computational Thinking, Washington, D.C.: The National Academies Press (2010); and National Research Council, Report of a Workshop on the Pedagogical Aspects of Computational Thinking, Washington, D.C.: The National Academies Press (2011).
Additional Areas of Promising Computer Science
and Related Research for Sustainability
In addition to the research areas discussed in Chapter 2 of this report, following is a list presenting a sampling of topics and areas that arise in computer science and information technology more generally that are likely opportunities for making progress in sustainability.
• Science of resilience and adaptive systems. This is a likely area of opportunity especially as it applies to self-regulating processes, biodiversity, and metrics of adaptability.
• Design for robustness, resilience, graceful degradation, and the decoupling of abstractions from implementation (for instance, designing for average-use cases and building in techniques for degradation, as opposed to designing for peaks with safety margin).
• Mass customization, especially in the role of programming languages. This is a likely area of opportunity for many levels of programming, at many stages of the life cycle.
• Understanding technology in context. One cannot understand how technology will affect sustainability without understanding what people will do with it. The emphasis in computer science on extensibility in system design takes into account the fact that technology as used matters, not just as designed.
• Design thinking. This area is involved with the invention of things that people will use and engage with, which is crosscutting for multiple domains.
• Search—a profound advance resulting from decades of research and innovation in multiple areas. What is the equivalent of search in the physical world? How do we deal with unstructured search, taxonomy, structured query processing—search for data relevant to scientific discovery?
• Computer vision. This field offers likely opportunity as a modality for searching and understanding the physical world.
• Representation for purposes of discovery. This is an area of opportunity in terms of representation of the physical world and of sustainability problems.
• Social media. This area of opportunity relates to information support and sharing, building community, structured argumentation, sensing, modeling, and observation.
• Tools for the automated design of very large scale systems. This is an area of opportunity that includes the development of capabilities to cope with challenges where there is functional decomposition
As one example, many sustainability challenges, particularly those related to infrastructure, make salient the importance of architecture. Architecture encompasses not just structural connections among subsystems, but expectations regarding what a system will do, how its performance
will scale, what behaviors are within bounds, and how subsystems (or external actors) should interact with the system as a whole. This type of challenge can be seen perhaps most clearly in the smart grid example discussed in Chapter 1. However, the other illustrative examples in Chapter 1—food systems and the development of sustainable and resilient infrastructure—also make clear that early architectural choices (such as the expected role and behavior of individual farms in the global food system, or the anticipated communications capabilities of first responders in a crisis) can have long-lasting repercussions.
A system’s architecture instantiates early design decisions and has a significant effect on the uses, behaviors, and effects of the system long past the time when those decisions were made. Moreover, requirements inevitably change over time, necessitating flexible or evolvable designs. Because of this large effect of a system’s architecture on almost all aspects of the system over its life cycle, the architecture of larger-scale systems of necessity merits significant attention and resources. As systems have become global in scale, the disciplines of computer science and software engineering have grappled with the challenges of architecture as they pertain to large-scale systems working over large geographic areas, with countless inputs and millions of users. Lessons from architecting hardware, software, network, and information systems thus have broader applicability to the processes of structuring, designing, maintaining, updating, and evolving of infrastructure in pursuit of sustainability.3
One question not yet addressed in this report is how well the CS research community is poised to play its part in meeting global sustainability
3For an in-depth examination of the importance of architecture in software-intensive systems, see Chapter 3 of the following report: National Research Council, Critical Code: Software Producibility for Defense, Washington, D.C.: The National Academies Press (2010). It describes the importance of architecture as follows (pp. 68-69): “Architecture represents the earliest and often most important design decisions—those that are the hardest to change and the most critical to get right. Architecture makes it possible to structure requirements based on an understanding of what is actually possible from an engineering standpoint—and what is infeasible in the present state of technology. It provides a mechanism for communications among the stakeholders, including the infrastructure providers, and managers of other systems with requirements for interoperation. It is also the first design artifact that addresses the so-called non-functional attributes, such as performance, modifiability, reliability, and security that in turn drive the ultimate quality and capability of the system. Architecture is an important enabler of reuse and the key to system evolution, enabling management of future uncertainty. In this regard, architecture is the primary determiner of modularity and thus the nature and degree to which multiple design decisions can be decoupled from each other. Thus, when there are areas of likely or potential change, whether it be in system functionality, performance, infrastructure, or other areas, architecture decisions can be made to encapsulate them and so increase the extent to which the overall engineering activity is insulated from the uncertainties associated with these localized changes.”
challenges. In the committee’s view, one perceived barrier to its doing so is the sense that aiming research toward sustainability challenges may conflict with ultimate scientific aims of universality.4 This question is explored below in more detail.
The most powerful and important computer science innovations to date share the characteristic of universality. Indeed, it is their utility across a wide range of domains that makes their aggregate impact so great. See Box 3.2 for a short list of just some of computer science’s most significant achievements. Universities, research laboratories, departments, and funding agencies increasingly recognize the value of multidisciplinary research. However, as the field has matured, there has been comparatively less emphasis on domain-driven approaches to innovation in favor of research that attempts to go directly to universality—that is, abstractability and breadth. Nevertheless, it is a strength of computer science that the field can, and does, ground its advances in real-world problems. As described in Chapters 1 and 2, CS can contribute significantly and critically to sustainability. In this section it is argued that to have the biggest impact on the pressing challenges facing the world today, CS research must be informed with deep knowledge, input, and context from domain experts.
In some areas of computer science, universality is built into the problem definition. Much of theoretical computer science, of course, begins by representing the target problem in abstract, symbolic language. Other examples of research with universality as the focus from the start include the von Neumann computer model itself, programming languages such as Fortran and ALGOL, and early human-factors research (such as that of Doug Englebart and Alan Kay) that created new modes of humancomputer interaction. In other equally consequential areas, however, broad applicability has only emerged years or even decades later, as researchers began with domain-specific problems and developed solutions and then later were able to generalize and understand deeper truths from this panoply of specific contributions. Examples of important contributions that began as highly specific projects include the World Wide Web (originally conceived as a means to share research papers and scientific information) and object-oriented programming (early object-oriented languages were developed to address specific problems such as discrete event simulation or graphical interaction).
In this chapter it is argued that CS research on sustainability is best approached from the bottom up: that is, by developing well-structured
4The integration of computer science with domain sciences was a central tenet of a 1992 National Research Council report: National Research Council, Computing the Future: A Broader Agenda for Computer Science and Engineering, Juris Hartmanis and Herbert Lin (eds.), Washington, D.C.: National Academy Press (1992).
Universality and Computer Science’s Greatest Achievements
The list below offers one view of some of computer science’s most significant achievements over the years. “Computer science” is construed broadly here, to encompass information and communications technologies.1 Most of the achievements listed below were accomplished by focusing on developing domain-specific solutions with an eye toward eventual abstraction and universality.
- Universality—the Turing machine
- Computability and hardness—P, NP, and complexity classes
- Stored program computer—execute anything
- Operating system (versus operator)
- The Internet and Internet Protocol—move anything anywhere
- File storage—move anything forward in time
- Coding, decoding, and cryptography
- Programming language and its translation
- Layered design (versus vertical integration)—abstraction as a foundation
- Microprocessor, the personal computer
- Audio/video representation
- Distributed systems
- Self-describing schemas, documents
- Computer-aided design, computer-aided manufacturing, computer-aided optimization, computer-aided engineering (design, simulate, build, test, measure, use)
- Database query languages and management systems
- Continuous user input (pointers, clickers)
- Graphical user interfaces in various forms
- Parallel execution
- Numerical simulation of physical phenomena
- The web, hypertext, and distributed markup in a global namespace
- Massive keyword search
- Scheduling, planning, optimization
- Very-large-scale integration, design rules, synthesis, verification, massive systems production
- Search techniques, heuristics
- Machine learning
- Structure of graphs
1There are, of course, other takes on this question. The Computer Science and Telecommunications Board’s (CSTB’s) well-known “tire tracks” diagram (as seen in National Research Council, Innovation in Information Technology, Washington, D.C.: The National Academies Press ) explores innovations in computer science and information technology from the perspective of spawning billion-dollar industries. That list includes time-sharing, client/server computing, entertainment, Internet, local area networks, workstations, graphical user interfaces, very-large-scale integration design, and reduced instruction set computing processors. A CSTB report published in 2004 articulated the essential character of computer science, and focused on seven key themes: computer science (1) involves symbols and their manipulation, (2) involves the creation and manipulation of abstractions, (3) creates and studies algorithms, (4) creates artificial constructs, (5) exploits and addresses exponential growth, (6) seeks the fundamental limits on what can be computed, and (7) focuses on the complex, analytic, rational action that is associated with human intelligence. (National Research Council, Computer Science: Reflections on the Field, Reflections from the Field, Washington, D.C.: The National Academies Press .) The differences in these lists are less important than noting the power of focused problem solving in a field whose core strengths include abstraction and adaptability.
solutions to particular, critical problems in sustainability and later seeking to generalize these solutions, as opposed to striving for universality from the start. Many advances will require CS research for progress, as described earlier, but those advances may not be immediately evident as universal approaches. The committee believes that demanding evidence of clear universality from the start is likely to inhibit close interdisciplinary collaboration and, ultimately, major advances. Moreover, many sustainability challenges need to be addressed sooner rather than later, even if imperfectly. At the same time, the fact that the problems associated with sustainability are complex, multifaceted, and in some cases poorly defined means that close attention to experimental robustness and underlying mathematical rigor will be essential.
Previous chapters illustrate this overall approach. For example, the hypothesized research on the smart grid takes an approach to the problem that is fundamentally a computer networks perspective (inspired by the success of the Internet), but it is not initially intended to make universal contributions to the theory of networks. Rather, by focusing on the problem at hand—that of integrated control of power generation, distribution, and use—there is the potential for breakthrough advances on this critical issue at the same time that new computational techniques are being developed. There are clearly general lessons to be learned about these issues. The committee suggests that the best way to learn them is to start with the particular sustainability challenge at hand, make progress on that, and later seek to generalize.
This approach does not mean, however, that any application of computation or IT to problems in sustainability should automatically be seen as computer science research for sustainability. Rather, to be judged as a significant contribution within the intersection of CS research and sustainability, the contribution first must have the potential to make a real difference in moving toward a more sustainable future. Second, the contribution must have the potential, if it is successful, to add to generalizable knowledge about sustainability, and the contribution or proposed solution should, at the same time, require new computational techniques or thinking beyond the current state of the art in computing.
The specific criteria for judging research success should of course evolve over time, with members of the community themselves proposing and debating what constitutes the most worthy research. The committee emphasizes, however, the criterion of having the potential to make a real difference. An open research question in its own right is how best to assess and evaluate impacts and how to isolate the effects of any particular sets of interventions, given the scales and time frames of many sustainability challenges.
PRINCIPLE: Encourage research at and across disciplinary boundaries, well informed by specifics and well structured to handle scale, data, integration, architecture, simulation, optimization, iteration, and human and systems aspects. CS research in sustainability should be an interdisciplinary effort, with experts in the various fields of sustainability being equal partners in the research.
Although the committee emphasizes that a premature focus on universality would be detrimental to the kind of high-impact sustainability solutions so desperately needed, universality should not be ignored. Indeed, domain-specific research can lead toward universality. A challenge, though, is how to pursue the universality that contributes so much to the power of computer science. In this section, it is argued that the purposes of domain specificity and contextualization are not at odds with ultimately producing universality in results, and that universality is not achieved directly in most cases in any event. Consider the development of important advances in CS and IT. Achieving universality typically involves developing well-structured innovative solutions, applying them to the problem at hand, evaluating their efficacy, and using this evaluation to guide further improvement, enhancement, and new directions. Successful approaches are then refined and applied in other areas, perhaps similar to the original problem domain, perhaps more remote. As the iterations of application proceed, the universality of the approach is discovered and refined.
Why has this approach worked so well in computer science? Despite the fact that computer science has information at its heart, tools and methods are ultimately instantiated in software. Software is malleable and well disposed to iteration. Software technology is developed, deployed, used, and modified in continuous iterative cycles. Developing modern software is not done through implementing a perfect software system once, at the start. Instead, the state of the art in software engineering urges iteration and architectural flexibility. Software is designed to be updated on a frequent basis over its entire life cycle.
This approach to the creation of software systems has developed for many reasons. For one reason, the work required to discover all or even most of the bugs before release in non-critical systems far exceeds the value of that approach. Similarly, feature sets are expanded through use. The range and number of possible features of any particular target system are larger than what is implemented—if that were not the case, the systems would be even more complex and difficult to use and would take even longer to roll out. Thus systems are rolled out with a modest feature set,
and new features are then added over time. This approach allows products to get to market sooner and helps to avoid or reduce the development of features that are not really needed, as well as revealing what actually works in practice. In short, after initial deployment, reality (instead of anticipated or modeled reality) guides the evolution of the feature set. The understanding of the human systems into which all computing systems are deployed is highly limited (not least because the understanding and modeling of humans and organizations are highly incomplete and flawed at best). Thus one cannot generally anticipate and simulate all uses. The world of computer systems has grown to evolve features that are adapted to what the users of those systems demand. A virtuous cycle has resulted—users have come to expect flexibility and malleability, thus ensuring that feedback loops occur. Systems have versions that are rolled out as available and as features are demanded. Change management is a basic fabric of these systems, which are designed for ongoing change. These systems have innovation and expectations of innovation literally encoded within them.
Another way to think about the inherent adaptability of computer science is to consider an “end to end” argument for the inclusion of authentic applications in systems research. The original end-to-end argument put forward by Saltzer, Reed, and Clark was as follows:
[F]unctions placed at low levels of a system may be redundant or of little value when compared with the cost of providing them at that low level…. The argument appeals to application requirements, and provides a rationale for moving function upward in a layered system, closer to the application that uses the function.5
This same logic has implications for systems research and innovation. Authentic applications should be included as part of systems research exploration at as high a level as possible in order to keep functional and performance requirements on a purposeful track. For the purposes of this report, notions of authenticity must encompass high-impact applicability to sustainability challenges.
These fundamentals in computer science are relevant to the way that CS research is done and the way in which research and development investment is approached. This “built for change” characteristic also facilitates the transition from one application to another. Algorithms and their instantiations can be adaptive, iteratively modified to fit a new
5J.H. Saltzer, D.P. Reed, and D.D. Clark, “End-to-End Arguments in System Design,” Proceedings of the Second International Conference on Distributed Computing Systems, Paris, France, April 8-10, 1981.
context. More importantly, such iterations are not a significant departure from what occurred in their initial creation; the expectation of iteration is part of the core of the technology. Building for change is done through modularity, through system designs resulting from hard thinking about where to place functionality, through the isolation of errors and details. Change to software happens as the software is developed, and as it is deployed, debugged, and iteratively improved; and it happens as it is applied to a new problem. As a given technique is applied to a new problem, and yet another new problem, and so on, the universality of the technique emerges. For each new application, the characteristics of generality are exposed, and the possibility for further abstraction and broad applicability grows. In the best case, an “exponential” process emerges in which techniques that are broadly applicable are exposed as each successful reapplication enables multiple new adaptations to come to light. Not all potential new applications are developed, of course, but those that are find their ultimate universality through bottom-up cycles of change and through the iterative process of design which promotes that process. Past successful examples of this approach include language translation, Internet protocols, machine learning, object-oriented languages, and databases.
The approaches discussed in Chapter 2 were not described in their most general terms. The committee does not suggest that they be pursued generally. But universality is often seen as the ultimate win of computer science techniques. Although universality is important and must be the goal because some big wins are needed in order to attack the unprecedented challenges of sustainability, the challenges should be approached through the concrete. There are opportunities for CS research to take on the key challenges in sustainability, learn about them, and design focused solutions that work. The design of those solutions should embed the best of CS design and systems learnings—modularity, isolation, simplicity, and so on. Then CS researchers and practitioners should experiment with, apply, and pilot solutions to specific problems; look for the successes and reapply and adapt them to other applications; and develop universality while seeking to increase applicability and impact. If the concrete is embraced across the range of infrastructure, ecosystems, and human systems, reality will help hone and filter possible approaches, and multiple and adapted applications will emerge.
FINDING: Fast-moving iterative, incrementally evolving approaches to problem solving in computer science, which were critical to building the Internet and web search engines, will be useful in solving sustainability challenges.
The vision described above implies a broadening of what it means to be a computer scientist. A significant opportunity for change is in the area of education. This change should include educating computer science students to achieve impact with computing, computational methods, and systems approaches in important domain-specific areas. Such a shift in culture would encourage these students to develop domain expertise and to collaborate directly with domain experts while in graduate school or in preparing for graduate work6 and to address such topics as modeling and predicting energy use and designing for reuse.
Making such a shift successfully will also require a culture of experimentation and innovation in the application of computer science. Further, it will require a research infrastructure in order to make progress. That infrastructure should include the following: (1) available standard data sets, models, and challenge problems to the community in order to assist in developing a common discourse and target for innovation, analogous to Grand Challenges in robotics, speech, vision, and so on; and (2) the building of shared infrastructure through open architectures and testbeds that allow for grounded iterative experimentation in the context of real components, both human and technical. Such architectures could go a long way to increasing the feasibility and impact of experimental research in academia and to creating an ecosystem that supports iterative innovation.
Education and training within the target domains constitute an equally important goal. One challenge is in the translation of problems from one domain or field to another—for instance, describing the power and electric grid systems as a dynamical system and control problem—and then translating sometimes newly exposed assumptions back to the problem domain. Information and data are critical to understanding the challenges, formulating solutions, deploying solutions, communicating results, and facilitating learning and new behaviors that are based on results of the work. Thus a significant component of meeting virtually all sustainability challenges is to infuse computational thinking and computer science- and information-rich approaches into the deploying industry and the research and mission agencies.
PRINCIPLE: Undergraduate and graduate education in computer science should provide experience in working across disciplinary boundaries. Graduate training grants and postdoctoral fellowships should support training in multiple disciplines. Undergraduate and
6These shifts are already underway in various fields—for example, biocomputing.
graduate programs should include tracks that offer introductory and intermediate course work in such sustainability areas as lifecycle analysis, agriculture, ecology, natural resource management, economics, and urban planning.
Research institutions—both universities and the funding organizations—could better address the needs of authentic multidisciplinary research, in terms of publication venues, funding, criteria for promotion, research infrastructures that help enable sustained collaboration, and cross-training. The latter include the cross-training of students in multiple fields to enable them to bring a computer science perspective into other arenas. Authentic multidisciplinary work is challenging.7 Work will need to be done across disciplinary boundaries and incorporating experts from many disciplines, as well as individuals with deep expertise themselves in more than one discipline. Examples of opportunities to enhance multidisciplinary approaches are described below:
• The creation of certificate programs, extension programs, and online programs for professionals in the target industries and agencies through professional societies and lifelong learning and training;
• Scholarships and fellowships both for computer science graduate students and for early-career professors that provide financial support for taking the time to develop expertise in a complementary discipline;
• The development of cross-agency initiatives (such as the collaboration of the National Science Foundation [NSF] with the Environmental Protection Agency8) that encourage interdisciplinary collaboration in relevant fields;
• Support for the development of new, cross-discipline structures (perhaps departments or institutes) between computer science and other fields that can create a new generation of students who are agile both in computer science and in fields relevant to sustainability;
7For a discussion of some of the challenges, see Sean Eddy’s essay on “antedisciplinary” science (Public Library of Science, Computational Biology 1(1), doi: 10.1371/journal.pcbi.0010006), in which he notes: “Focusing on interdisciplinary teams instead of interdisciplinary people reinforces standard disciplinary boundaries rather than breaking them down. An interdisciplinary team is a committee in which members identify themselves as an expert in something else besides the actual scientific problem at hand, and abdicate responsibility for the majority of the work because it’s not their field.”
8An example is the joint National Science Foundation/Environmental Protection Agency establishment of two centers to study the environmental implications of nanotechnology, described in a 2008 press release: see http://www.nsf.gov/news/news_summ.jsp?cntn_id=112234.
• Institutional structures that support multidisciplinary and interdisciplinary teams focused on a problem or set of problems over an appropriately long period of time;9
• Internships and career paths and placement programs that encourage computer science students and postdoctoral researchers to work in relevant government agencies, non-governmental organizations, and industries;
• Coordination between academic research in computer science and non-traditional industrial partners—that is, beyond the large IT companies—to scope problems, help train students, and cross-fertilize ideas; and
• Regular, high-level summits involving computer science and sustainability experts—practitioners and researchers—to inform shared research design, assess progress, and identify gaps and opportunities.
The conceptualization of the bottom-up emergence of universality is relevant to researchers, university systems, and funding agencies in the following respects:
• Researchers—emphasizing a bottom-up approach affects how researchers select and approach their problems and how they approach the training of their students. Cross-training, learning other languages and vocabularies, immersion, and intensive and sustained collaboration are all important aspects of how research will need to be done. Although it may be essential to making real progress with respect to sustainability challenges, longterm commitment to a specific domain area is typically inherently risky for a CS researcher, because specific problems in sustainability may be addressed successfully but universal ideas or techniques may not necessarily materialize.
• University systems—a focus on bottom-up approaches affects how universities incentivize and create the infrastructure for faculty to pursue sustained multidisciplinary efforts. The computer science community has made progress in tenure and in the promotion of individuals who straddle disciplinary boundaries. Is such boundary crossing sufficiently encouraged and explicitly incentivized? Publication rates and pressures are higher than ever. Publication is essential to a successful R&D ecosystem, but does an emphasis on frequent publication have a negative effect on the pursuit of R&D which tackles difficult application domains that have not previously been processed and translated into computer science problems?
9One successful example of such an effort was the collaboration between computer scientist James Gray and astrophysicist Alexander Shalay on the multiyear effort (which, of course, involved many others) to develop the Sloan Digital Sky Survey (http://www.sdss.org/).
Moreover, much truly multidisciplinary work will require large teams and collaborations; is there appropriate recognition of CS contributions to large, multiauthor publications? Also, an evaluation of the productivity of junior faculty may need to extend to evaluating the impact of the researcher in the realm of sustainability in addition to the field of computer science. Related to promotion is the question of appointments—in what departments are multidisciplinary researchers appointed, and how can such appointments be handled so that the multidisciplinary nature of the researchers’ work does not count against them in their home departments?
• Funding agencies—emphasizing bottom-up approaches may affect how agencies structure multidisciplinary programs. The National Science Foundation is a primary funder of research in computer science in the United States. The former Information Technology Research Programs at NSF and its current Cyber-enabled Discovery and Innovation Program have demonstrated the feasibility of programs with significant multidisciplinary aspects and the impacts that can result. But such programs provide for a minority of CS research, and in the committee’s view, the sense of the community, as seen in review panels, program structures, review criteria, and so on, is not generally favorable toward funding domain-specific projects. One challenge is that typical reviewers of prospective research in computer science tend to want to see universality from the start, which presents a fundamental problem for the CS research community; funding agencies such as NSF can do little about this matter if the community does not adapt.
The committee is encouraged by the establishment of Science, Engineering, and Education for Sustainability (SEES) as an NSF-wide area of investment. SEES aims for a systems-based approach to “promote the research and education needed to address the challenges of creating a sustainable human future” and places an emphasis on interdisciplinary efforts. With its emphasis on interdisciplinarity and the involvement of NSF’s Directorate for Computer and Information Science and Engineering, the SEES program offers an opportunity to demonstrate the depth of IT and CS innovation that the core discipline can offer and the rich and globally important problem space of sustainability.
There are ongoing opportunities for NSF to take advantage of the significant domain expertise in other agencies in order to pursue a strategy broader than programs that are crosscutting with other research directorates within NSF. Such a broadened strategy would involve programs that connect researchers with domain experts, practitioners, and projects in relevant mission agencies (such as the Department of Energy, Department of Transportation, Department of the Interior, and Department of Health
and Human Services).10 Furthermore, it is critical to ensure that funding structures support data sharing, encourage citability, and so on, but at the same time, an emphasis on sharing should be balanced with the need for researchers actually to collect data and to begin solving problems.
Another role that funding agencies can play is to fund longer-term projects and to be tolerant of risk, particularly in these multidisciplinary and cross-disciplinary, potentially high-impact research areas. It may be useful to target special funds that encourage the switching of focus to sustainability challenges and to incentivize grounded, domain-specific collaboration and training.11
PRINCIPLE: Refine funding and programmatic options to reinforce and provide incentives for the necessary boundary crossing and integration in CS research to address sustainability challenges. In particular, funding, promotion, and review and assessment (peer review) models should emphasize in-depth integration with data and deployments from the constituent domains.
PRINCIPLE: There should be strong incentives at all stages of research for focusing on solving real problems whose solution can make a substantial contribution to sustainability challenges, along with in-depth metrics and evaluative criteria to assess progress.
Another critical issue for structuring research is to build in evaluation tools for prioritizing efforts and evaluating meaningful impact. The committee offers an evaluative framework below.
One of the greatest challenges in multidisciplinary research is to establish evaluation metrics that are both actionable and meaningful across the constituent disciplines. This chapter concludes by identifying methodological opportunities for optimizing research outcomes and impacts. Each of the recommended areas for evaluation necessarily incorporates
10An example of such a program is NSF’s Industry/University Cooperative Research Centers Program.
11An illustrative example is the National Institutes of Health Mentored Quantitative Research Development Award (K25), which serves to fund quantitatively trained researchers so that they can learn about an area of biomedical science (see http://grants.nih.gov/grants/guide/pa-files/PA-06-087.html). These awards require the identification of a mentor in the substantive area and a plan for training; they provide funding for a commitment of at least 75 percent time.
interdisciplinary team members to assess accurately and to guide the potential for positive and significant impact.
One set of validation metrics would ensure that sustainability-oriented efforts related directly to key components of sustainability and were making significant progress on one or more of the legs of social, economic, and environmental impacts.12 Specific metrics might include the following: metrics for analysis of environmental impact (life cycle, energy needed to create and execute a solution, as well as energy saved, and so on); metrics for analysis of economic impact (cost of implementation, resources used to create intervention, externalities); and metrics for equity and engagement across different stakeholder groups, taking into account their interests, values, and concerns, during both design and execution processes.
Many of these metrics involve humans, and so the assessment of research using such metrics will often involve techniques from the social sciences.13 They also involve techniques drawn from other fields, such as life-cycle analysis from civil and environmental engineering. Any solution that claims to engage with and increase sustainability should be able to make an argument that addresses possible ways in which it may also negatively impact sustainability in any of its three overarching components (social, environmental, and economic). Moreover, proposals for projects that aim to make an impact should include a life-cycle analysis, and such analyses should be accounted for in the projects’ budgets, as such analyses are a non-trivial effort to do well.
Scale analysis is critical to most information technology designs; it includes spatial scaling, temporal scaling, location scaling, and computational scaling. As demonstrated by the classic computer science talk
12As an example, although the challenges of sustainability are much broader than those related to climate change alone, in that domain a measurable impact is the goal of decreasing emissions so that carbon dioxide levels in the atmosphere are at or below a certain threshold. Being as specific as possible in terms of the anticipated impact on sustainability is critical. More broadly, one might consider the areas addressed in life-cycle analysis: global warming, stratospheric ozone depletion, acidification, eutrophication, smog, terrestrial toxicity, aquatic toxicity, human health, resource depletion, and land use. See U.S. Environmental Protection Agency, “Life Cycle Assessment: Principles and Practice,” available at http://www.epa.gov/nrmrl/lcaccess/lca101.html.
13This is a characteristic of much of the evaluation done in human-computer interaction research and practice. Also very relevant here are design methodologies and approaches that seek to account for human values: for example, Value Sensitive Design, as discussed in B. Friedman, P.H. Kahn, Jr., and A. Borning, “Value Sensitive Design and Information Systems,” in P. Zhang and D. Galletta (eds.), Human-Computer Interaction in Management Information Systems: Foundations, Armonk, N.Y.: M.E. Sharpe (2006), pp. 348-372, reprinted in K.E. Himma and H.T. Tavani (eds.), The Handbook of Information and Computer Ethics, Hoboken, N.J.: Wiley (2008), pp. 69-101.
question “But does it scale?,” an early measure of the potential for the impact of a solution or system relates to scale. As sensors become more powerful (being able to measure many parameters at high frequency) and cheaper, a massive increase in the amount of data is expected. Each step or component of a proposed solution should be designed to work at scale. However, scale has new connotations with respect to sustainability that must be considered. In particular, many of the most problematic phenomena involved in sustainability challenges play out over timescales that are difficult for humans to comprehend, and many of the solutions make sense only when applied at scale. Here are some of the scalability questions that should be considered when evaluating a project:
• Spatial scaling requires articulating the minimal scale that can have impact (for example, touch points or density, geographic coverage, and so on).
• Temporal scaling incorporates real-time, human-time, and planning-time considerations; duration; timescale of relevance (onset, persistence, resilience); and the very challenging issue of addressing multiple-human lifespan timescales.14
• Location scaling addresses the applicability of the approach across a range of location-related contexts.
• Computational scaling refers to the tractability of the problem in terms of data generated and in terms of the anticipated footprint of the approach with respect to energy and dollars expended.
For each of these dimensions, the impact of a targeted innovation may well be difficult to quantify precisely. But it is the responsibility of the researcher at least to estimate and justify the anticipated, first-order measurable savings and efficiency improvements, or mitigated damages, from its realization. Is the proposed approach fundamental infrastructure? A game changer? The introduction of new computing technologies and concepts should be coupled with impact assessment (positive and negative) and follow-up study/assessment along with plans to integrate and iterate learning. Quantifying sustainability results as contributed by computational methods is a daunting challenge, especially given the current lack of data for real-world systems. Focused efforts toward creating publicly available data repositories that could be used to compare the effect of methods on the performance metrics chosen may prove useful in some domains. Without such data, there is little common ground for
14B. Friedman and L.P. Nathan, Multi-lifespan information system design: A research initiative for the HCI community, Proceedings of the 2010 ACM Conference on Human Factors in Computing Systems, New York: ACM Press (2010), pp. 2234-2246.
systematically comparing different methods and their potential benefits. It is particularly important for researchers to estimate outcomes and for the community to develop ways to assess impact, as these steps may have a dramatic impact on whether or not an approach can be appropriated and applied in other contexts.
Meeting the challenges of sustainability, as noted in Chapter 1, will require more than information technology, applications of clever technology, and computer science research. Indeed, at the heart of many global sustainability challenges are questions of resource consumption and standards of living. (See Box 3.3.) Nevertheless, the committee believes that
Toward an Information-Rich, Sustainable Future
Numerous analyses make clear that resource consumption is at the heart of many global sustainability challenges. At the same time, populations around the world are striving to improve their standard of living—and despite the efficiency improvements that also accompany development, that has inevitably meant increased resource consumption.
Efforts to improve efficiencies and substitute more sustainable for less sustainable materials and methods are what underlie much of the discussion in this report. However, there may be a broader sense in which information technologies and computational approaches can alleviate or mitigate the problem. Efforts to shift standard-of-living metrics from resource-intensive to information-intensive have the potential to be a significant lever in addressing global sustainability, although such shifts will increase the need for ever-“greener” information technology solutions themselves. In an increasingly information-rich and carbon-restricted world, finding ways to use information so that it both enhances perceptions and realities of standard of living and reduces resource consumption will be critical.
Examples of shifting to information-rich, less resource-intensive lifestyles include adjustments to transportation practices such as: information infrastructures that transform the convenience and trust of shared and alternative transportation modalities instead of private automobiles; improved technology in vehicles; transportation displacement such as telecommuting, social media, and e-commerce; and so on.
Although such examples emphasize opportunities to shift what counts as improvements in the standard of living for individuals, ultimately it is the policy choices and decisions, at local, regional, and federal levels, that will determine how many, if any, of these shifts are possible. Thus, organizational and governmental actions and decisions will have significant impact on whether a shift to information-intensive choices can happen in order to produce a shift in the way that society operates, to engender more sustainable outcomes.
CS and IT research has deep and fundamental contributions to make to these challenges. This chapter has argued for a bottom-up approach to research that values application-driven results while also supporting the iterative process that eventually leads to more universally useful contributions. The committee has argued for a series of validation metrics that explicitly explore the true impact of a piece of work in the arena of sustainability. Such validation metrics should include those that deal directly with humans, economics, and ecosystems and those metrics that engage with the concept of scale (a good first-order proxy for the universality that may not yet be present).
Information technology is at the heart of nearly every large-scale socioeconomic system—financial systems, manufacturing systems, energy systems, and so on. One important consequence, which has been the focus of this report, is that advances in IT have become critical enablers of change in these systems. The goal of this report has been to shine a spotlight on areas where information technology innovation and computer science research can help, and to urge the computer research community to bring its approaches and methodologies to bear on these pressing global challenges.