Front page   Table of  Contents   Abstract Chapters  1   2   3   4   5   6   7   8   9   Bibliography

1. Salient findings from the three cases

Chapters 4, 5, and 6 presented three cases of information infrastructures being built to facilitate collaboration among government agencies: the Great Lakes Information Network (GLIN), the Gulf of Maine Environmental Data and Information Management System (EDIMS), and the interrelated Northwest Environmental Database (NED), Coordinated Information System (CIS), and StreamNet. All three chapters used the same set of criteria, grouped as follows:

their history and context;

choices made in building the infrastructure;

characteristics of the information and infrastructure;

impacts of the infrastructure; and

challenges and lessons encountered.

From a policy viewpoint, all three infrastructures were intended to facilitate joint protection of a water body shared by several states or provinces; but this goal varied in clarity and urgency between the cases. In the Great Lakes case, several related agreements and policies dating back several decades had articulated and instilled norms of Great Lakes stewardship and community among many people and organizations in the region. Furthermore, the US Environmental Protection Agency adopted an emphasis on ecosystem management and partnerships in the early 1990s. In addition to these norms, however, two additional elements were key to getting GLIN started: a few people at the Great Lakes Commission and CICnet who articulated the relationship of these norms to shared information; and the deep pockets of the Ameritech Foundation.

In contrast, Gulf of Maine regional policies consisted only of a 1989 Agreement between three governors and two premiers; joint stewardship of the Gulf of Maine was not yet a strong or clearly articulated norm among most of the region’s agencies or inhabitants. It’s not too surprising, then, that participants’ commitments to Gulf-wide concerns were smaller, including their ability to articulate shared information goals or to obtain and keep funds for information-related activities.

In the Pacific Northwest, the 1980 Northwest Power Act provided a clear set of norms (state-federal cooperation, and the electric power industry’s responsibility to mitigate the impacts of dams) and allocated large resources from the region’s electric power sales. These norms and resources grew even more significant after the Endangered Species Act was brought to bear in 1990. The twin crises of balancing electric power against fish and wildlife and restoring threatened salmon species provided a strong impetus for information-based cooperation between states, tribes and federal agencies; but the information infrastructure choices still depended on the 1985 GIS task force to articulate them, and on a few regional coordinators to keep them on the budget year after year.

All three cases, then, illustrate the importance not only of a shared ecosystem, but of a convergence between (i) shared norms about stewardship of the ecosystem, (ii) significant fiscal resources available, and (iii) someone to articulate the relationship of those norms to shared information, and to harness available resources.

On a technological note, all three cases featured an infrastructure trying to define its standards and strategies amidst a field of other, competing choices. This theme is evident in the Great Lakes case, in which GLIN was trying to define its relationship to CIESIN, IMAGIN, and others; it also appears in the Pacific Northwest case, in which NED and CIS were funded simultaneously by the same Bonneville division, and several state systems had their own data standards and ongoing GIS projects. In the Gulf of Maine, as the Web suddenly appeared on the "radar screens" of many in the Council on the Marine Environment, it suggested many alternatives for communication and information sharing, and thus not all saw EDIMS as the Council’s principal mechanism for networked information sharing. These examples suggest that multiple competing alternatives to information sharing in a region may not be unusual; and that they may or may not be harmful to the effective growth of information sharing infrastructures. In a cross-section of the three cases, proponents of GLIN, NED, and EDIMS were all concerned about alternative efforts grabbing the attention and resources (both money and information) of their participants; but actual outcomes may be mixed. The many alternatives may diminish funding and institutional support (as shown by EDIMS in the Gulf of Maine). However, long-term benefits may include learning from the alternative efforts (StreamNet shows what NED’s proponents learned from CIS’ focus on detailed data management), or a more focused view of the infrastructure’s audience and purpose (GLIN sharpened its focus on training agencies to maintain online servers).

b. Infrastructure choices

In the three cases, the primary institutional choice was an interagency committee or board to coordinate infrastructure-building efforts. The Great Lakes cases featured the largest and most diverse board, with participants from two dozen government, industry, academia, and non-profit organizations. Coordinating committees in the Northwest case featured only government agencies; NED mixed energy specialists with fish and wildlife specialists; CIS’ and StreamNet’s committees narrowed the focus to fish and wildlife-related agencies (StreamNet brought in the US Geological Survey to oversee Reach File coordination). In the Gulf of Maine, the committee that oversaw EDIMS was loosely defined, with several non-participating members: participants belonged to state and province agencies and universities, and came together based largely on their personal interest in coastal and ocean data management. The level of functioning of these committees (their meeting frequency, consistency of attendance, etc.) correlated with the effectiveness of the infrastructures they oversaw; it was hard to distinguish cause from effect.

Certain technological choices were also quite similar across the three cases. All three infrastructures were aimed at linking distributed information across the region without imposing too many changes on source data sites. In the Great Lakes and Gulf of Maine cases, the goal of not intruding on data sources led to a "laissez-faire" evolution with few or no standards and a diminished emphasis on detailed, dynamic data. In contrast, the Northwest Environmental Database (NED) team was more data-centered: it invested in building state cross-reference tables, and based its infrastructure on the River Reach standard. Recognizing the difficulty of maintaining data for regional use, GLIN and NED both began by equipping agencies in their regions: GLIN staff conducted training on using and providing networked data services; in the Pacific Northwest, Bonneville purchased GIS hardware and software for state and tribal agencies, and funded several full-time staff for data collection and coordination. (The Gulf of Maine data team, with a much smaller budget, simply deferred its distributed data plan until more agencies put servers online.)

The three infrastructures show quite different choices, however, in their emphasis on geographic information: GLIN waited until its third year to begin building interactive map interfaces on the Web; whereas EDIMS hadn’t yet taken the GIS plunge after some five years of existence. As for the Northwest case, NED was inherently geographic from its start in 1985, but CIS handled only tabular data to avoid the complexities of geographic data; and StreamNet featured an online map library, later joined by an interactive "map builder."

c. Infrastructure characteristics

One salient point in common between the cases: from quite different beginnings, all three infrastructures converged onto a Web-based architecture, in which a central "hub" server was intended as an intermediate "scaffolding" for building a future distributed web of data servers in the region. However, evidence suggests that moving beyond this "scaffolding" may be difficult. The Gulf of Maine EDIMS has remained at the centralized stage for much longer than expected. The Great Lakes network was almost entirely centralized for a long time, but saw other agencies’ servers come online in 1995-96. As for StreamNet, it’s too early to tell how easily this so-called "Aquatic Information Network" will evolve beyond its current centralized form.

The size and traffic levels of the three infrastructures are quite different, in keeping with their vastly different funding levels, and reflect the length of time they’ve been in service, the size and breadth of their audiences, and the usefulness of their information to their audiences.

The quality of the information distributed by these infrastructures is difficult to compare, given the quite different forms of information shared and the different uses targeted. Nonetheless, usability and encapsulation are one clear point of comparison: much of EDIMS’ data was raw ASCII tables, whose use required considerable processing and interpretation. StreamNet (and its predecessors NED and CIS) presented information to users in more usable form: database queries as well as maps (and an interactive "map builder" on the Web). Some of GLIN’s partners feature similar query interfaces; GLIN staff also began to experiment in late 1996 with map query tools on the Web.

The quality of the three infrastructures is a bit easier to compare. Reciprocity (two-way information flow) was part of all three designs, but carrying out that vision on the Web has proven difficult given that few sources of data could or would serve it online. All three infrastructures are scalable in that they can link to new Web sites as needed; but none of the three has achieved a scalable data management strategy. As for non-intrusiveness, all three infrastructures recognized the need for it, but GLIN and EDIMS went so far as to define no standards at all. GLIN eventually defined design guidelines and a keyword list, and began to comply with the Global Information Locator Service (GILS) standard to facilitate document search and retrieval. The three infrastructures in the Pacific Northwest were more data-centered, and adopted the River Reach files as their common standard from the outset. NED was an early implementation of non-intrusive design; as StreamNet moves out of its centralized "scaffolding" form, it may need to re-visit NED’s non-intrusive approach and to adapt it to the networked environment.

d. Infrastructure impacts

Users of all three infrastructures report learning experiences and improved collaborative relationships. The Great Lakes has the most tangible evidence of this, with a great many servers going online, and large traffic volumes on GLIN’s core server. In the Great Lakes, the Regional Air Pollutant Inventory Development System (RAPIDS) is claimed as a GLIN impact—but it was really a separate EPA-funded project, directed by GLIN’s director, that used GLIN’s main server to exchange electronic mail and to interchange software under development. Only NED in the Pacific Northwest seems to have already had clear, tangible impacts on environmental planning and policy: impacts in the other two cases are mostly in the future tense, or in less specific, measurable terms.

e. Challenges and lessons

One challenge encountered in all three cases is the complexity of building and maintaining networks of distributed, dynamic information. Building the "scaffolding," a temporary structure to get started, was relatively easy; but moving on to a distributed data architecture, or to dynamic data rather than fixed "brochures," or maintaining an infrastructure over time, was more difficult. Related to this is the importance of large commitments of resources, considerable technical expertise, and especially a technical and institutional growth path. Indeed, in none of the three cases was there a comprehensive blueprint that could be followed to build the mix of network technologies, data structures, and organizational support required for an effective infrastructure; GIS and Web-based software tools were subject to especially unpredictable change. Therefore, the process of improvised learning and growth was more important than any set of factors that could have been established in advance. Yet at the same time, a completely improvised, "laissez-faire" approach, with too little attention to the design of organizational, technological and policy systems, proved less than effective as well. Section 4 below explores in more detail the dilemma of balancing the planned and improvised approaches.

2. Differing views of information and sharing

One way to generalize beyond these cases towards a more general pattern of infrastructure design, growth, use, and impact, is to examine what has motivated these people to act in their various ways—that is, their perspectives on the problems to be solved by information-sharing infrastructures and apparent solutions they provide. These multiple perspectives are possible because information-sharing infrastructures are currently in a state where many choices are possible, and many definitions of the problem coexist: a state of interpretive flexibility (Pinch and Bijker (1984). What follows is a sketch of the salient views of information sharing, the designs they engendered, and what these different views have to offer both separately and in concert.

a. "We will serve no data before its time!": Information sharing as data management

The tongue-in-cheek quote above was from a hydrologist at the U.S. Army Corps of Engineers in Detroit, which maintains several hundred gigabytes of data on Great Lakes hydrology, bathymetry, and meteorology. While this quip was a deliberate exaggeration, many keepers of large geographic data collections, across the three cases, did hold the view that lengthy preparation and documentation of information resources was both necessary and sufficient for inter-agency data sharing: at Michigan’s Resource Information System, technicians worked night and day shifts to manage some 40 Gb of vector data sets, and distribute them through a service bureau. A GIS manager at the U.S. Environmental Protection Agency (EPA) in Chicago painstakingly documented all geographic information before releasing it, and compared himself to a "sea anchor" preventing EPA staff from being swept away by a data-hungry public. The state of Washington’s Department of Natural Resources (DNR) put a lot of emphasis on data management, license agreements, and pricing procedures, in keeping with their view of data as a state asset alongside timber or minerals. A chief concern in all these settings was to protect the data investment by releasing only high-quality, extensively documented data and ensuring that it was not misused. The resulting infrastructures tended to consist of quite formalized administrative procedures (price sheets, Memoranda of Understanding, and signatory partnerships). Also, maintainers of such infrastructures were slow to use data networks: to them, the Internet brought mostly security holes, and chances to "get in over one’s head." For instance, Michigan’s sophisticated Resource Information System (MIRIS) awaited a statewide networking plan for months before experimenting with any kind of Internet data distribution: "otherwise, we’d be alone in figuring it out," said one MIRIS administrator. (Noticeably absent from these characterizations is the Gulf of Maine EDIMS, which hasn’t yet tried to tap into the sizable repositories of geographic data in the states and provinces.)

b. "Democracy in action": information sharing as public disclosure

Through the above quote, an editor for the Great Lakes Information Network (GLIN) emphasized the goal of making information freely available to the public—supporting inter-agency collaboration was not the primary goal, but an immediate and fortunate consequence of communicating with the public. In keeping with this view, a former chair of GLIN’s advisory committee argued that "information is the only commodity worth anything today; withholding it results in costs to society." In the Pacific Northwest, a tribal fisheries coordinator had concluded from that region’s experience that it was foolish to keep information "behind closed doors." And although the Gulf of Maine data team hadn’t made public disclosure a high priority (their focus was state and province agencies), they were getting pulled in that direction by the Gulf of Maine Council’s committee on public education and outreach. These groups all held high the public disclosure of data, above concerns for privacy, documentation, expense, or direct personal or agency benefit. The resulting information-sharing infrastructures were informally structured, and targeted a broad audience and simpler forms of data. They did much to turn public agencies outward towards interaction with the public. But their emphasis on public channels risked overlooking collaborative opportunities that involved more confidential data; or excluding partner agencies that didn’t have the same openness towards public information and participation.

c. If we build it they will come: information sharing as a networking project

To some participants in the cases, information sharing was chiefly a matter of building networked data access tools. In the Great Lakes case, two foremost goals of the Consortium for International Earth Sciences Information Network (CIESIN) were to test its novel "Gateway" software and to grow its Socio-Economic Data Archive Center. This meant an emphasis on building prototype data structures, query tools, and data catalogs. On a smaller scale, the Gulf of Maine’s Environmental Data and Information Management System (EDIMS) had a similar emphasis on putting data on a website and counting on people to fetch them: the information itself was not packaged for any particular use, and its relevance to users’ tasks or needs was not yet a primary concern. In the Northwest case, NED and CIS were set up long before Internet connectivity became widespread among the region’s government agencies: without the distractions of a networked environment, their designers could focus on data-centered design choices. Once these systems were in place, StreamNet’s network-intensive approach proved slow to implement, and required going beyond incremental changes (such as the CIS "distributed system"). But thanks to its data-intensive predecessors, StreamNet could draw on a long technological and organizational foundation of joint data collection and management throughout the region, and quickly go beyond a networking project to serve its audience with useful, valuable data.

d. "We learn from each other": information sharing as a meeting of the minds

For many participants in these cases, the clearest payoff of the information-sharing infrastructures was fresh ideas from people with similar professional interests in distant organizations, whether through electronic mail or Web pages. Granovetter (1972) emphasizes this "strength of weak ties" in diffusing information through a social network. In the cases, informal learning was a frequently-cited benefit of all three information infrastructures. Members of Great Lakes agencies, for instance, could keep in touch with regional events, funding opportunities, and the activities of their counterparts in other states or across the Canadian border. In the Gulf of Maine, coastal-resource managers could learn about environmental conditions and different regulations in other parts of that region. In the Pacific Northwest, several participants report learning from people on other sides of that region’s environmental debates: fisheries managers, electric-power planners, foresters, and so on. Some in that region also report learning about information management and GIS through NED and other inter-state data sharing efforts. Within this view, networked infrastructures exchanged primarily knowledge and expertise, rather than specific sets of data. Building professional communities is an important part of supporting inter-agency collaboration through information-sharing infrastructures. However, absent a common base of shared information (as in the Gulf of Maine case in particular), each participant may learn new things but only the most informal of joint actions is possible.

e. Getting it right: creatively solving new problems, organizational change, integrated choices

The above are some of many views of the "information sharing problem" observed in the cases: (i) painstaking data management, (ii) public disclosure and participation, (iii) networked experimentation, and (iv) professional cross-fertilization. These views yielded an equally wide array of technological solutions: (i) data standards and procedures, (ii) open-access sites, (iii) custom networking tools, and (iv) electronic forums.

All of these views are important and have something to offer to the design of shared information systems; and most of the time, people can work on these different areas in isolation from each other. For instance, rigorous data management creates a trusted information resource, one that people can count on to be available and reliable; public disclosure encourages organizations to be more open to new relationships and opportunities; networked experimentation is needed to stay abreast of a rapidly-changing set of technologies; and forming professional communities is important to keep everyone’s choices mutually beneficial. However, at certain decision points in the growth of the infrastructure, too sharp a focus on any one of these may obscure choices that could support more than one goal. For instance, an exclusive focus on data management, through robust data structures, security, or extensive documentation, may overlook information structures that would serve both internal and external needs. Conversely, focusing purely on communication and networking may overlook the importance of robust and secure data management, or may even omit structured alphanumeric or geographic data in designing the infrastructures. The cases showed several examples of a strong dichotomy between the data management and communication functions: few instances featured data structures suitable for external communication, or communication media suitable for detailed structured data. Thus, even though these different views of information sharing, and the design choices they lead to, can happily coexist most of the time, there are times when someone needs to bring these views together, and consider multiple goals and multiple development paths at once. In the Great Lakes and Gulf of Maine cases, the slow move away from temporary "scaffolding" systems may have been due to the difficulty in bringing these different areas of expertise to bear.

A second limitation of these partial views is that trying to use a new, still-malleable tool to solve existing problems may miss an important potential of information sharing infrastructures. Perhaps the "problem" of inter-organizational information sharing is better described as a situation in which several problems and solutions can be creatively invented and reconciled (cf. Carlson, 1992). If so, then successful cases of information sharing infrastructures will be those where important new problems are identified and resolved through information sharing—rather than those where networked information is simply used to address existing problems. As Weick (1990) suggests, "one of the many streams found in organized anarchies is a steady stream of solutions in search of problems." As far back as 1980, researchers on telecommunications and other information technologies have suggested that "new computing systems are often applied not only to existing organizational problems but to qualitatively new organizational activities." (King and Kraemer (1980), cited in Attewell and Rule (1984)).

As a consequence, these views (i-iv) are also limited in that they presuppose an essentially unchanged organizational structure: this assumption may be incompatible with information sharing infrastructures. The "qualitatively new" activities supported by information sharing infrastructures will often require deep changes in the priorities, dependencies, and identity of participating agencies: this possibility tends to be overlooked as various participants set their expectations. Integrating these different views together into a single information-sharing infrastructure holds a lot of promise. The following section suggests a number of ways to leverage technology, organizations, and policy in order to move towards such an integrated view.

3. Mutual influence of technology, organizations, and policy/planning

As another way to interpret the case study findings, Chapter 2 suggested that sharing geographic information ought to be viewed as an interaction between organizations and technology (Markus and Robey, 1988). As detailed in that chapter, a structuration perspective highlights the "intertwined" nature of technological and behavioral choices (DeSanctis and Poole, 1994), and allows technology to be defined as a malleable element of organizational structure (Orlikowski, 1992)—that is, a set of rules and resources that enable some actions, while constraining others, and that are in turn shaped by those actions over time. Within this perspective, the technical design of an information sharing infrastructure is ineluctably tied to its implementation and use within an organizational context. This provides a fruitful model for understanding the processes of ongoing learning and growth evidenced in the case studies. Orlikowski (1992) traces a three-way cycle of mutual influence between institutional properties, technology, and the human agents who use and build the tools within the institution. The cases in Chapters 4, 5, and 6 suggest a similar cycle of influence between organizational, technological, and policy/planning structures, and the actions people perform within those structures (Figure 7-1).

Figure 7-1. Mutual influence of technology, organization, and policy/planning on people and actions

a. Technology influences actions

Tangible impacts of technology on action (arrow a above) were often less than had been hoped for in many agencies reviewed in the Great Lakes and Gulf of Maine cases. Except for those who actually ran these infrastructures, most members of these agencies had kept much the same tasks, priorities, and identity they had prior to using information sharing infrastructures. However, in the Pacific Northwest, even without the grand technologies of wide-area electronic networks, NED’s use of a simple regional information standard did much to influence people’s actions: many who had been loath to disclose information changed their minds when they saw the political strength to be gained by helping to build a regional resource inventory based on the River Reach Files. Furthermore, NED’s intuitive data categories and colorful maps proved instrumental in helping participants to articulate their priorities and (in this case) to find common ground on which to act. On the other hand, technological difficulties may also hinder changed actions: despite NED’s early success in the Pacific Northwest, lingering data incompatibilities (nominal data, identifiers, etc.) and the absence of data networks for many years, this group of people had difficulty making a broader set of joint decisions in the ensuing years. In the Great Lakes, air-pollution specialists used widespread Internet access for regional pollution assessment through RAPIDS; but information differences, primarily language (English vs. French) and measurement units (feet vs. meters), did complicate other trans-boundary projects.

b. Actions influence organizations

The cases show several examples of people’s actions affecting organizational structures (arrow b above). In the Gulf of Maine case, the first committee chair to oversee EDIMS invested time and effort to obtain federal funding, thus making information-sharing systems highly visible in the early years of the Gulf of Maine Council. In the Pacific Northwest, the efforts of NED’s regional coordinator did much to encourage continued familiarity and cooperation between agencies. In the Great Lakes, GLIN’s director at the Great Lakes Commission played much the same role, by keeping in touch informally with many GLIN partners and assisting many others to put information online and join the GLIN partnership. As a result, many GLIN participants felt quite strongly that exchanging networked information about programs or personnel was leading to improved coordination and responsiveness between agencies. Conversely, some felt this benefit sometimes brought with it greater outside scrutiny and a potential loss of privacy—especially when using public channels like the World Wide Web. This, too, was a (perhaps unintended) organizational effect of people’s actions.

c. Organizations influence actions

The organization’s influence on people’s choices (arrow c above) takes a variety of different forms. First, the organization’s size and experience clearly affect the ambition of information-sharing initiatives. It’s no great surprise, for instance, that the Pacific Northwest StreamNet took on broader objectives than the Gulf of Maine EDIMS: it had twice as many participating agencies, a budget more than ten times as large, technical input from some of the region’s top GIS experts, and a fifteen-year history of grappling with inter-organizational information. Organizational properties also show their influence when group "insurgencies" (Thomas, 1993) enable individual "champions" to lead others in making key choices. One such insurgency was the so-called "users’ revolt" by state GIS centers in the Pacific Northwest when the first version of the upgraded River Reach files failed to meet expectations. This movement was led by two individuals, but it gained prominence because the state’s GIS centers had reached a level of organizational maturity and technological expertise that enabled them to see beyond the "marching orders" of their federal funding sources. In the Great Lakes, the Great Lakes Commission’s long-standing role as an inter-state compact, and CICnet’s position as regional Internet provider, prompted and empowered a few people at the Commission and CICnet to launch the GLIN concept and follow it through. In the Gulf of Maine, the organizational context had a dampening effect on EDIMS development and support: due to the Gulf of Maine Council’s relatively weak role in relation to the region’s coastal and marine management agencies, EDIMS could not count on many agencies in the region to support its growth.

d. Actions affect technology

The cases provide many examples of people’s actions affecting technology (arrow d above). In the Northwest example just cited above, although organizational maturity allowed two individuals to lead a "user’s revolt" with important consequences on regional standards, nonetheless what actually started the change was these individuals’ dissatisfaction with the product they were given, and their choice to do something about it. Similarly, in the Great Lakes case, GLIN’s open-ended, decentralized design were due largely to one person’s vision and enthusiasm for a regional network, combined with another person’s technical abilities and shared enthusiasm. The Gulf of Maine’s decentralized design owed much to the strong preferences of its hosts at the University of New Hampshire, who believed strongly in decentralized systems with few standards. Also, the slow growth of EDIMS beyond its centralized form can be attributed to the hesitancy of some members of the EDIMS committee to plunge into Internet connectivity at the expense of other networking alternatives. Although in retrospect, organizational or policy conditions might seem to have dictated these choices, nonetheless these and other individuals were key to starting and promoting (or slowing) technological change.

e. Policy and planning influences action

An agency’s policy and planning environment can have important impacts (whether deliberate or not) on what its members do (arrow e above). Ecosystem-based policies in particular often require or encourage agency personnel to coordinate and collaborate across jurisdictional boundaries (and thus to form interagency committees and to use information-sharing technology). The Northwest Power Act is a good example of a policy that has shifted public investments strongly in favor of inter-agency relationships and shared information systems. Policy and planning goals also affect budget levels, and thus the magnitude of investments in organizational or technological change. The Northwest Power Planning Council, for instance, has a modest size (as measured by its personnel and budget), but the Northwest Power Act endows it with a lot of power to leverage Bonneville funds; its members and partners have used this position to influence the region’s choices of data collection and alliance formation. The discussion of "norms" in Section 1 above discusses other examples of how the policy context has influenced people’s choices in the Great Lakes and Gulf of Maine cases.

f. Actions influence policy and planning

Finally, people using new technologies within these emerging organizational structures made a few region-wide policy and planning choices (arrow f above)—and sometimes, the difficulty in changing technology or structures kept people from making effective policies. In the Pacific Northwest, for instance, a handful of NED coordinators using a regional information standard played a major role in drawing up the region’s first system-wide regional hydroelectric and environmental protection plan. This was the ultimate goal of the information-sharing infrastructure in all three cases, although NED is the only one to have already had a clear policy impact.

g. Technological and organizational change

Besides tracing cycles of influence between structure and action, the structuration perspective also posits the potential for ongoing change in social structures. It assumes that Spackman’s (1990) "organic, yet systematic" change over time is normal and ongoing, rather than surprising and temporary. Thus, as detailed in Chapter 2, information sharing infrastructures are seen not as fixed sets of interlocking components, but as a chosen direction, or even a style of evolution through an uncertain future. Change is usually a response to new technology, funding, relationships, or policies; but (whether intended or not) it can also serve to develop and maintain an organization’s agility. The case studies illustrate each kind of change.

Changes in the technological or organizational context can threaten the survival of information sharing infrastructures. Recent years have brought an acceleration in the turnover of networked information technologies: barely a year after the Great Lakes Information Network got underway, for instance, it had to make major changes to its information services, switching to the World Wide Web after having put a few hundred documents on Gopher servers (recall that Gopher was the Internet technology of choice in 1993). The World Wide Web also caught the mostly telnet- and ftp-based Gulf of Maine EDIMS by surprise in 1994, just as it was being mothballed for lack of funds. When it awoke from its 18-month slumber, the "cyberspace" phenomenon was in full swing, and several committees in the Gulf of Maine Council were clamoring for a sophisticated Internet presence. The organizational front is just as unpredictable: new powerful partners or competitors may arise unexpectedly; others may go dormant; alliances may wax and wane. In the Pacific Northwest, congressional actions on regional energy supply, salmon restoration, and endangered species led to an ebb and flow of federal funds over the years; with each wave of funding came new webs of inter-agency partnerships and new attempts at data standards.

To respond to an unpredictably changing world, information-sharing infrastructures need to be thought of, and designed, as living, growing organisms rather than fixed artifacts. Management guru Tom Peters has written extensively about deliberate organizational change, "necessary disorganization" in response to a unpredictable, rapidly evolving business context (Peters, 1992). One example of this "design-for-change" was the Pacific Northwest River Reach Files: even before the first version was deployed at the 1:250,000 scale, plans were underway for a dramatically upgraded second version at a 1:100,000 a few years down the road.

In summary, then, the perspective depicted in Figure 7-1 highlights the many influences at work between technologies, organizations, and policies, and the people who live and work within these structures. Separating these different influences may seem a bit artificial in places (for in stance, it’s through using networked technologies that people may be inspired to create inter-agency relationships). Nonetheless, it’s important to highlight the role of (ultimately unpredictable) individual actions in building, maintaining, or changing technology, organizations, and policy. By separating out the different influences of technology, organizations, and policy, it becomes a bit easier to anticipate the effect of adjustments to these structures, either separate or in concert. Even so, none of the three cases played out a single predefined strategy, but rather evolved in response to a number of dilemmas and a shifting set of policies, organizations and technologies. More generally, then, it would seem that no strategy can have fully predictable impacts: instead strategic choices are likely to evolve with the changing context.

This uncertainty may be especially important when dealing with geographic information systems, for which the technical tools are still evolving rapidly, and which (as detailed in Chapter 2) tend to encourage relationships that cut across traditional hierarchies and boundaries and thus encourage either official or de facto organizational change.

Keeping these uncertainties in mind, the next two sections consider the possible role of new technologies and organizational forms.

4. Technology choices

a. From datasets to data services

Findings from the case studies suggest a number of promising directions for the design of information-sharing infrastructures. In particular, certain new technologies promise useful perturbations in the "influence cycle" described earlier. In particular, one frequent theme in the three cases was a dichotomy between data management and inter-agency communications. Each of the cases featured several Web servers, but most of these delivered only static pages and graphics; few provided a gateway into an organization’s larger (especially geographic) information resources. These resources was generally kept behind database managers, GIS packages, or other software intended for interactive use by a data custodian. Accordingly, while one set of people focused on inter-agency electronic communications, another worked independently to manage the agency’s information resource and distribute it on compact discs, magnetic tape, or paper maps. Quite possibly, merging these two functions would reap greater benefits from information-sharing infrastructures. Two important requirements for this are communications systems geared for queries against structured data, data management systems geared for online access, and a commitment to serve both internal and external needs in concert where possible. Newer client-server architectures may help with these requirements: the next chapter uses a simple example of a geographic data service to explore the implications of such emerging technologies.

b. From standards to metadata

Chapter 2 contrasted traditional data standards with newer "minimal" standards, in which metadata are used to help users make sense of information without altering the information itself. The cases showed many settings where this approach might be useful: independent agencies that had an interest in sharing information, but were unlikely ever to agree on a common set of definitions, formats, or coordinate systems for their data. The Great Lakes Information Management Resource provided an example of metadata for information sharing: a set of common keywords allowed a single query at any one site to retrieve relevant documents from several Web sites at once. The cases also showed an interesting early example of geographic metadata, the cross-reference tables that linked state stream identifiers in the Pacific Northwest to the regional River Reach Files. The next chapter looks more closely at the benefits and challenges of maintaining and using geographic metadata.

c. From compatibility to interoperability

If information is shared through an interaction between software systems, then getting the software programs to understand each other becomes more important than reducing data to a common format, structure, or coordinate system. This requires generic reference models of the geographic entities and operations to be shared, and program interfaces between those reference models and individual information systems—that is, interoperable data services rather than compatible datasets. Interoperability is a long-term goal of many information-sharing infrastructures such as those in the cases. However, the difficulty of defining a reference model to suit widely-varying applications, has made progress slow thus far. In addition, as illustrated by the Pacific Northwest River Reach Files, geographic information presents special challenges, such as scale, not raised by the alphanumeric and administrative information behind most Information Systems research. The next chapter evaluates a few small steps towards interoperability between geographic data services, and towards scale-independent geographic referencing.

5. Organizational choices

a. From autonomy to interdependence

Finally, a more "online" view of information, with data services, metadata, and interoperability, implies several organizational shifts for the agencies involved. First, it requires data managers to take responsibility for providing reliable data services to others outside. Conversely, it expects analysts and decision-makers in these organizations to depend on outside data for essential tasks. Third, learning to live with different data conventions, and to negotiate them through metadata, implies informal, yet trustworthy cooperative relationships between organizations. In fact, reaping the intended planning and policy benefits of information-sharing infrastructures may require deeper-than-expected changes in organizations and inter-organizational relationships. This may explain the difficulties GLIN proponents had in seeing tangible impacts of information sharing: most of the region’s organizations had not yet undergone these deeper changes. The same explanation may have been behind EDIMS’ lack of relevance to Gulf of Maine organizations. By comparison, in the last decade, many more agencies in the Pacific Northwest began (whether willingly or forcibly) to see inter-organizational coordination and information interchange as fundamental to their work, and made appropriate organizational changes: this may have helped them to reap greater benefits from their information-sharing infrastructure.

From observing the business world, Moore (1996) suggests that increasingly complex relations among organizations have made traditional market models obsolete: in their place, he proposes a "business ecosystem" ruled not by autonomy, authority, or competition, but by interdependence, persuasion, and "co-evolution" with competitor/partners in a constantly changing business context. This perspective seems at least as convincing in the world of public management as in the business world. In particular, in the complex trans-boundary contexts of the three cases, many organizations are trying to look beyond a traditional emphasis on leadership on a policy issue, and to embrace teamwork within a shifting pattern of alliances and resources. Chapter 9 explores the organizational and strategic implications of "government ecosystems" linked by information-sharing infrastructures.

b. Growing complexity over the long term

Finally, the cases suggest that building effective geographic information infrastructures may require balancing simple, tangible results early on with a much more complex permanent structure, attained through a long term path of intentional learning and growth. As mentioned earlier, none of the three cases followed a single strategic "blueprint" throughout their evolution: the technical and organizational requirements are too complex, and too context-specific for that—especially where they have attempted to handle geographic information. Chapter 9 explores this question of planned and improvised development, and the tension between incremental and more radical change.

c. Sharing costs and benefits reliably

The balancing act between simplicity (now) and complexity (later) raises another dilemma: who pays? The costs are often immediate, ongoing, and sizable; the benefits, though potentially great, are often distant and uncertain. Given this equation, the apparent "bureaucratic inertia" often encountered in geographic information sharing should come as no surprise. This theme is explored further in Chapter 9 as well.

 Front page   Table of  Contents   Abstract Chapters  1   2   3   4   5   6   7   8   9   Bibliography