3
Interactions for Information Technology Innovation in Federal Statistical Work
The workshop discussed the information technology (IT) requirements of the federal statistical agencies and the research questions motivated by those needs. In addition to articulating a sizable list of research topics, workshop participants made a number of observations about the nature of the relationship and interactions between the two communities. These observations are offered to illustrate the sorts of issues that arise in considering how to foster collaboration and interaction between the federal statistical agencies and the IT research community aimed at innovation in the work of the agencies.1
One obstacle discussed in the course of the workshop is that despite interest in innovation, there are insufficient connections between those who operate and develop government information systems or who run agency programs and those who conduct IT research. In particular, federal agencies, like most procurers of IT systems, tend to rely on what is available from commercial technology vendors or system integrators (or, in some cases, what can be developed or built in-house). A program aimed at bridging this gap, the National Science Foundation 's (NSF's) Digital Government program, was launched in June 1998 to support research aimed at stimulating IT innovation in government. The premise of
1 |
These observations should not be viewed as necessarily being conclusions of the study committee that organized the workshop. The committee 's conclusions will be presented in the study's final report, to be published later in 2000. |
this program is that by promoting interactions between innovators in government and those performing computing and communications research, it may be possible both to accelerate innovation in pertinent technical areas and to hasten the adoption of those innovations into agency infrastructure.
Building connections that address the needs and interests of both communities entails the establishment of appropriate mechanisms for collaboration between the IT research community and government IT managers. In principle, the right mechanisms can help federal program and IT acquisition managers interact with the IT research community without exposing operational users to unacceptable levels of risk. Also, incorporating new research ideas and technology into the operations of government agencies frequently requires spanning a gulf between the culture and practices of commercial systems integration and the research community.
Also relevant to the issue of innovation and risk in the context of government in general, and the federal statistical system in particular, is the value attached to the integrity of the federal statistics community and the trustworthiness of the results (relevant principles are summarized in Box 1.1). These are attributes that the agencies value highly and wish to preserve and that have led to a strong tradition of careful management. Such considerations could constrain efforts that experiment with new technologies in these activities.
Experience suggests that despite these potential constraints and inhibitors, both research and application communities stand to benefit from interaction. Introduction of new IT can enable organizations to optimize the delivery of existing capabilities. The full benefits of IT innovation extend further, as such innovation can enable organizations to do things in new ways or attain entirely new capabilities. Advances in IT research represent opportunities not only for increased efficiency but also for changes in the way government works, including the delivery of new kinds of services and new ways of interacting with citizens. Collaboration with government agencies also represents a significant opportunity for IT researchers to test new ideas—government applications are real and have texture, richness, and veracity that are not available in laboratory studies. Frequently, these applications are also of a much larger scale than that found in most research work.
While the workshop focused primarily on long-term issues, another benefit was the shedding of light on some short-term problems. Indeed, it is natural for people in an operational setting to focus on problems that need to be solved in the next year rather than on long-term possibilities. This suggests that focus on and investment in long-term challenges may be difficult. But in some respects, the near-term focus may be appropri-
ate, since some of the information technologies and IT practices of the federal statistical agencies lag behind best industry practices. In an example illustrating the short-term, mundane challenges that consume considerable time and resources, one workshop presenter described the challenges posed by the need to install new software on several hundred laptop computers. In later discussions, it was pointed out that this was a problem that had already been solved in the marketplace; there are wellknown techniques for disk imaging that allow initialization of thousands of computers. Underscoring the potential value of such interactions, informal feedback following the workshop suggested that the exposure to some cutting-edge computer science thinking stimulated subsequent discussion among some statistical agencies about their need for further modernization.
One factor that may be exacerbating many of the short- and long-term IT-related challenges is the decentralized nature of the federal statistical agencies, which makes it harder to establish a critical mass of expertise, investment, and experimental infrastructure. Another difficulty arises from the specialized requirements of federal statistical agencies. The market is limited for software for authoring and administering survey interviews of the complexity found in federal statistical surveys, which are quite expensive and are conducted only by government and a few other players. Workshop participants discussed how the federal government might consolidate its research and development efforts for this class of software. Several IT applications in this category were cited, including survey software, easy-to-use interfaces for displaying complex data sets, and techniques for limiting the disclosure of confidential information in databases.
Collaborative research, even within a discipline, is not always easy, and interdisciplinary work is harder still. Researchers at the workshop argued that in order for such collaboration to take place, both IT and statistics researchers would need to explore ways of tapping existing research programs or establishing new funding mechanisms. 2 Computer scientists do not typically think of going to one of the statistical agencies, and statisticians do not typically think about teaming with a computer scientist for their fellowship research. Both computer scientists and statisticians will find it easier to obtain funding for work in more traditional
2 |
Workshop participants pointed to two NSF programs that could facilitate such collaborations if they were explicitly targeted toward such interactions. One is a fellows program in the Methodology, Measurement, and Statistics program that sends statisticians to various federal statistical agencies. The second is a Computer and Information Science and Engineering (CISE) directorate program that provides support for computer scientists to take temporary positions in federal agencies. |
research directions. So, given all the additional difficulties associated with interdisciplinary work, particularly in academia, it is unlikely to occur without funding directed at collaborative work.3 This, of course, was part of the impetus for the NSF Digital Government program.
More generally, a number of workshop participants acknowledged that involvement in application areas related to federal statistics offers significant opportunities for IT researchers. Each of the areas described in Chapter 2 was identified by participants as one where considerable benefits would be obtained from direct collaboration between IT and statistics researchers. A leading example is the area of information security. While some segments of the computer science community may be ambivalent about doing application-focused research, it is difficult to make real progress in information security without a specific application focus. A similarly large challenge is building easy-to-use systems that enable nonexpert users, who have diverse needs and capabilities, to access, view, and analyze data. Both the magnitude of the challenge itself and the opportunity to conduct research on systems used by a large pool of diverse users make these systems an attractive focus for research. Another particularly interesting issue discussed by workshop participants was the development of techniques to protect the confidentiality of spatial data.
3 |
Participants in a workshop convened by CSTB that explored ways to foster interdisciplinary research on the economic and social impacts of information technology made similar observations. See Computer Science and Telecommunications Board, National Research Council. 1998. Fostering Research on the Economic and Social Impacts of Information Technology: Report of a Workshop. National Academy Press, Washington, D.C. |