SGML:
Process, Policy, and Politics
By Kurt Conrad, The Sagebrush Group
Abstract
Organizational decisionmaking patterns determine SGML investment strategies and potential benefits. A framework for understanding the primary policy objectives that can influence the selection of SGML (inherent policy effects) and application design (user-defined policy goals) will be presented. Competing and often contradictory goals and perceptions of value often make the development of a business case for SGML very difficult. Methods for integrating stakeholder principles, interests, and expectations in the early stages of application conceptualization and design will increase real and perceived benefits and de-fuse potential political problems before they develop.
Introduction
Those of you who have followed my writings and presentations over the past few years know that I have tended to focus on the more non-technical aspects of SGML and information management. In the past, I have talked about such topics as information policy, stakeholder values, organizational performance, politics, and how these things influence the design of business processes and their supporting information management systems (especially those based on SGML). Along the way, I started to make a distinction between engineered and organic information management approaches.
My goal in this paper is not to recover all of that ground. You can find most of my
writings at http://www.3-cities.com/~conrad
sagebrushgroup.com. The January/February 1996 issue of The Gilbane
Report contains a fairly recent synthesis of my writings and extends the argument
with a critique of reengineering. And my other paper in these proceedings, Tools for
Implementing SGML-Based Information Systems: Viewers and Browsers, Text Retrieval
Engines, and CD-ROMs, introduces many of the same concepts but deals with
them in a more introductory fashion. Instead, this paper attempts to bring together a
number of threads that have been bouncing around in my head for the past year.
At SGML'95, for example, while I was listening to Eric Severson's presentation, SGML on a Strictly Business Basis, I couldn't help but think that the argument that he presented was both compelling and contradictory. Eric described a wide range of potential benefits, but few of these benefits result simply from using SGML. Besides readability and conformance testing (which result from the use of text and validation process that is defined in the standard) most of the benefits and their policy equivalents are determined by the specific application design approaches and the tools which are selected. I was hoping that this isn't the case and that an important set of policies/benefits were inherent to the use of SGML and could be realized by using the standard in any form, but I don't think that such is the case. If you have any other ideas, please challenge me on this assertion.
Even more important, the benefits of SGML tend to align themselves at two distinct loci. Some of the benefits involve improving (often short-term) operational performance and are based on streamlining complex mechanical transformations. Others concern themselves with positioning the organization and its data holdings to become increasingly efficient over time. The potential conflict stems from the fact that the design approaches (and especially the data structures) which tend to maximize long-term value also tend to be considerably more expensive in the short term.
As the value of SGML isn't realized from only the decision to use it, it also requires the organization to identify which subset of the potential benefits will be the target of the technology initiative and how costs and benefits are to be reallocated in the organization. Because these decisions are likely to pit competing interests against each other they tend to be a difficult decisions for many organizations to make. Most organizations distribute decisionmaking down a hierarchy and lack the mechanisms necessary to re-integrate the resulting disintegration. As a result, information policy tends to be implicitly defined by technologists who create policy as they implement new technologies. Most information technologies are designed to support a fairly narrow set of individual and organizational behaviors, and through their selection, the organization effectively adopts these policies.
SGML, in contrast, is predominantly policy neutral. In and of itself, the standard does not imply much in the way of automated, individual, or organizational behaviors. It is extraordinarily powerful and flexible, precisely because it fails to make all of the necessary decisions. The price for this flexibility is complexity, and so it was during Michael Sperberg-McQueen's closing keynote speech at the conference that another idea hit me: to compare the SGML with the U.S. Constitution (the document, not the boat). In the Politics of Information Management, Paul Strassman makes the following observation:
it [the U.S. Constitution] represents a point of view that addresses the governance of complexity by concentrating only on the fundamentals, while leaving everything else for resolution by means of a due process wherever that is appropriate.
But while the Constitution describes some of the mechanisms which should be used to make decisions, the SGML standard does not. It only describes how to document and communicate the resulting decisions. It documents the contract between information producers and consumers, leaving them to negotiate the specifics among themselves. From this perspective, I look to our political processes for guidance on how to improve the quality of organizational decisionmaking, but return discouraged. The dysfunctionality that exists in so many organizations seems to be reflected not only in our country's political institutions but seems to pervade the culture, as a whole.
And finally, Yuri's death in January caused me to spend much of the last nine months thinking about just one thing: community. In reading the many remembrances which were posted on the internet and printed in a variety of publications, I was struck by the fact that virtually everyone not only could remember when they first met Yuri, but were compelled to tell the story. And what everyone seemed to get from Yuri at that moment was a sense of belonging. They walked away with a deeply held convictions that SGML could not only solve their problems, but by doing so, help solve the problems of humanity, as well; and that because their values were reflected in the values of the SGML community, this was a natural home for them to pursue their individual interests.
Yuri's message was, perhaps, the ultimate example of a big tent. Everyone should participate in this revolution, because if they do, all of us will be better able to communicate reliably and with the richness of understanding which is needed to solve the truly big problems that we face both individually and together.
Dysfunctionality: The Dilbert Principle
Nothing seems to make sense any more. Nothing seems rational. A Roper Survey just found, among other things, that 74% of Boomers think that the government is completely out of touch (not just has a few problems). If this is the case, the dysfunctionality which prevents individuals and organizations from seeing the obvious benefits of SGML (or more accurately, identifying the pertinent benefits) is not unique to SGML, or even organizations, but appears to exist at a more fundamental cultural level. More than a couple of times this year, I have found myself asking if dysfunctional decisionmaking is what really caused the dinosaurs to die out. I wouldn't have mentioned it, except that I just finished Michael Crichton's The Lost World and I found the argument repeated there. The idea is basically that in complex, interrelated communities, changes in behavior can lead to maladaption and, theoretically, extinction.
In The Dilbert Principle Scott Adams makes the point directly. Basically, we are all idiots. We can't help it. Life is too complex for us to understand everything. Advanced communications and recording technologies (like books) have allowed a relatively small number of very bright people to make the world too complex for everyone else. Complexity drives specialization. Pools of expertise and thus mankind's capacity for intelligent behaviors become both more decentralized and more highly focused. But with all of this fragmentation comes the issue of how to align and integrate. The decisions necessary to assure our individual survival are no longer just ours to make. An increasing variety of social, commercial, and political institutions are making more and more of these decisions for us every day.
At some point, and we may be there now, individual decisions no long remain rational in the larger context. The same communications models which allowed decisions to be decentralized seem incapable of aligning the distributed decisionmaking with common or shared interests, and as a result, nothing seems rational anymore. Adams goes on to argue that today, because everyone is an idiot, organizations no longer promote people beyond their level of competence. The idiot that exhibits competence in their current job is too valuable to be reassigned. Hence, the Peter Principle has given way to the Dilbert Principle. Only the incompetent are moved into the ranks of management.
Closely tied to increasing complexity, we have seen an explosion of choice, especially since the 1950s. The demand for choice and the options that are available to us are certainly more expansive than ever before in human history. These choices range from the media that we consume, to our physical appearance, to our definitions of family. The support structures that have made two-income households practical have also made marriage, two-parent families, and even single teen-age motherhood matters of choice. Free-agency has brought choice to professional sports, but neither players nor fans seem to be happy. In fact, there is an emerging market value in simplicity. After years of expanding their product offerings, Procter & Gamble is discovering that they can reduce costs and even increase market share by simplifying their product lines. Consumers are rewarding them for reducing the number of available options.
But with these choices come not only complexity, but also instability. The world of limited choices that defined life in the 1950s is gone and so, too, appear to be the lasting relationships. Team owners can choose to move to another town. Our employers can choose whether to employ us next week. Our insurance companies can choose to reduce coverage. Our spouses can choose to divorce us. The present is increasingly complex, the future is increasingly uncertain, and paradoxically, we have become both more and less interdependent. We have become less dependent on specific individuals, but more dependent on others for meeting our individual needs. Often, those that we depend on are nameless and faceless and part of a bureaucracy, where our individual interests tend to be marginalized. Hmmm, now we're back in the Dilbert Zone.
The Quest for Community
We may be idiots, but we're not stupid. We seem to understand these fundamental threats to our individual survival very well, even if they are not well articulated. The solution to the problem of complexity seems to be to get more idiots. There's safety in numbers. Or perhaps more importantly, there is wisdom in numbers. The more diverse the group of idiots, the better. And this takes us back to issues of community, coordination, and communication... the stuff of SGML dreams.
Communities, groups which are defined by their shared interests and goals for the future, are coalescing at an explosive pace. In fact, the primary challenge in many organizations is to create communities, not the narrow ones which naturally develop, but broader, more expansive communities that integrate a number of diverse and potentially conflicting interests.
A recent Harvard Business Review echoes this phenomenon. It talks about
How the Internet's open standards are leading to networked manufacturing and those that cling to proprietary systems will be shut out from these emerging communities.
Chrysler's efforts to modify the keiretsu concept to create an extended enterprise from suppliers that were previously distrustful adversaries.
The need to develop long-lived relationships with customers instead of just selling product.
How the shift from processing resources to processing information has shifted an increasing portion of the economy from one of diminishing returns to increasing returns and rewards cooperation over competition.
But perhaps the most important article in this particular issue takes us back to the
realm of decisionmaking and describes how the traditional hierarchical chain of command is
counterproductive and both threatens and undermines the value of communities. Strategic
planning processes which start with the visions of senior managers are proving to be
elitist and sterile. True strategy can only be realized through a truly democratic
process that can give voice to the revolutionaries that exist in every company.
Concurrent Engineering seeks to create communities that are tied together by the entire product lifecycle, not just a portion of it. SGML initiatives that seek to improve an information lifecycle can only succeed if they create a healthy community that integrates the diverse and potentially conflicting interests that exist throughout the information lifecycle.
Who Sets the Course?
Here I echo an all too familiar cry: The problem is politics.
All too often,
technology initiatives, especially the type of radical initiatives that are common with
SGML, are undermined by office politics.
Politics, in one sense, is simply the set
of mechanisms that a group uses to make decisions. If stakeholders aren't given the proper
channels for integrating their interests into the decisionmaking process, it is only
natural that they put their energies into damage and destruction. A political process that
doesn't engage is very likely to alienate and create enemies.
Going back to the Dilbert Principle. The problem probably isn't just that we are getting the wrong idiots as managers, but that any organizational model that centralizes decisionmaking in the hands of managers is likely to fail to respond to the complex situations that information and knowledge-intensive organizations now face.
Strassman's history of governance models (and their parallels in the evolution of
computing technologies) proves useful here. Information technologies have moved through
the many stages in a relatively short period of time, and are now in what Strassman calls
The Age of Cooperative Alliances.
Many organizations, however, remain locked in
the strict hierarchy of decisionmaking which was more typical of feudal states and warring
armies. In fact, the recent (and not yet finished) phenomenon of downsizing, rightsizing,
dumbsizing, etc. has much more in common with the use of fear and force than the pursuit
of shared interests.
Is it any wonder that the workforce is concerned about the future? Computers are becoming the most important tools of business. They are enabling and driving a model of distributed decisionmaking that is generations ahead of the average organization's own methods of governance, and the mismatch is becoming glaringly obvious.
Active engagement of the widest possible set of stakeholders in the earliest stages of an SGML project is important to align project planning and design with the political realities that will either make or break the project. The emphasis here shouldn't be on technology, but on formally identifying the values (principles, interests, and expectations) that should be used as design parameters and will be used by each stakeholder to privately measure the success of the project. Attempts to integrate these stakeholder values often follows one of two models: consensus or accumulation.
The integration process shouldn't focus on functional requirements of the new system, but on the behavioral requirements of the organization, itself. I have been involved in a large number of requirement gathering exercises. Fixation on functional requirements tends to drive the discussion into technology areas that the stakeholders have limited expertise in and leads to a consensus approach, where the group needs to reach consensus before a given function is to be included in the specification. This does a fairly good job of limiting the number of requirements, but it often does so at great cost. Individuals and groups are compelled to argue for their interests to try to win support. This often creates an overall atmosphere of competition and conflict and damages the development of a sense of community. From a political perspective, many stakeholders often fail to get what they need to meet their individual requirements and thus have little or no incentive to support the initia tive. From an organizational development perspective, the resulting designs are likely to be incremental changes from the status quo that fail to significantly improve organizational health or competitiveness.
Focusing on behaviors and values in the earliest stages of the analysis keeps the
analysis at a policy level and provides a neutral forum for stakeholders to begin the
process of integrating and prioritizing their needs. The process is largely cumulative,
and what is perhaps most important is that by being accepting of individual stakeholder
values, integration occurs without creating a climate of conflict. Recently, I was
involved in facilitating one of these planning sessions for an international trade group.
Going into the analysis, the clients were quite concerned that long-standing conflicts
with a state agency would poison the process. At the end of the first day, the values of
over 50 stakeholders representing more than 40 organizations had been successfully
integrated and a course of action had been defined. The client, however was concerned that
nothing had been accomplished because it was too easy,
and a representative of the
state agency in question commented that the session wasn't exciting.
By focusing the discussion on values and instructing each stakeholder that their role is
not to compromise but defend their interests (with facilitator enforcement, if necessary)
much of the need
to argue a position is removed. Also, the addition of each value
to the list of planning parameters modifies and frames the other parameters. The result is
a very clear picture, both of a mutually-desired end state and the safest, most reliable
path to get there. And because each stakeholder can see their needs reflected in the
results, they are far more likely to work to realize the shared vision, either as part of
the formal initiative or in the execution of independent activities.
Finally, it is the process, not the outcome that produces the political viability. Only by participating, can stakeholders feel confident that their interests were incorporated in the results. Only by sharing their thought processes, can stakeholders understand and appreciate the importance of each other's values. Only by experiencing the process, can the results be put into a context that has meaning.
Designing Data to
Drive Intelligent Behaviors
Just as the quality of group decisions can be improved by establishing an understanding of the context that motivates individual stakeholders, the quality of information can be improved by documenting the context that drove its development. Specifically, intelligent behaviors are based not on simply believing something, but on knowing and understanding and having confidence in the information on which the behaviors are based.
Just as metadata is what drives behaviors, a very special class of metadata is useful to drive truly intelligent behaviors. The approach to information management that I am about to describe is based on the work of one of my associates, Bo Newman. Bo has put together a knowledge management model which describes the relationships among data, information, and knowledge by focusing on the transformations or processes that are used to create information from data and knowledge from information. Capturing and retaining information about these transformations (metaknowledge) is critical to preserving the behavioral value of the information and knowledge artifacts and helps to facilitate both understanding and the confidence needed to act intelligently on the information.
In previous drafts, I tried a number of approaches to explaining this knowledge management concept. None of them satisfied me (or the people that saw them). Communicating the concept is difficult enough that many knowledge management practitioners don't even use the terms knowledge management and metaknowledge. But just as one of my favorite Goldfarbisms is the message that there isn't anything new about multimedia, I don't think that there is really anything new about metaknowledge. Metaknowledge is simply a class of metadata that answers the questions: who, what, where, why, and how. Being able to reliably answer these questions is a key issue in many industries. In the airline industry, for example, they use the term effectivity to describe the need to be able to easily identify which pieces of technical data relate to a given airplane. Much of the information that controls that relationship involves knowing past activities and projecting future activities.
First, we need to know what the information is about. Some people even use the term
aboutness
to describe this characteristic. Good novelists are expert at
communicating aboutness and providing a context of descriptions that are rich enough to
make the worlds seem real and the actions of the characters rational. In addition,
answering the what question often involves answering it from a number of different
perspectives that reflect different ways of looking at the situation and employ different
conceptual models. We need to know why a piece of information was produced, to reach a
conclusion about whether it has relevance to our current focus. This is often called the
focus question. Until we know where the information comes from, who produced it, and how
it was produced, it tends to be suspect.
These are issues that initially involve documenting the production process. An SGML application that seeks to better preserve the long-term usability of information artifacts should provide mechanisms for managing these types of metadata (metaknowledge). Including appropriate element types in the DTD is one way to capture this information, but probably won't be adequate for all of it. Links to workflow systems, policies and procedures, status reports, strategic plans, and other documents and databases that help provide context should be considered. Likewise, mechanisms should be provided to layer metaknowledge over document sets to allow individuals to:
Capture tacit knowledge and metaknowledge for legacy documents.
Document and communicate their own analysis, interpretations, and understandings.
Describe the application of the knowledge artifact in new contexts.
Conclusions
Organizations that centralize decisionmaking in the hands of a few are incapable of responding wisely to the increasingly complex world that they face. Groups need to integrate the knowledge of many individuals in order to make wise decisions. Techniques that are based on the identification and sharing of stakeholder values have been proven to create the communities of shared interest that are needed for wise and politically viable decisions.
For SGML practitioners, the process of mapping stakeholder values uncovers the critical, non-technical policy parameters that ensure that new and innovative initiatives provide the right benefits to the right people. Also, by focusing at the policy level, it is possible to involve stakeholders much earlier in the process than would be practical if the focus were on functional requirements and their associated technical issues. In fact, because such a process can identify the types of benefits that are most compelling to the target community, it can be used to build the business case for an SGML initiative and justify the investments needed for the initial wave of training, analysis, and design.
Without such an integration process, building a business case can be much more difficult and risky. Difficult, because the public goals of an organization rarely reflect the reality of private values, and SGML's policy neutrality prevents it from providing a definitive set of benefits that a constituency can rally around. Risky, because many of us live in the Dilbert Zone, and non-participatory decisionmaking processes are often suspect, usually for good reason.
Finally, the same factors which are driving an increasing number of businesses to
democratize their workforce are building interest in a shift from information management
to knowledge management. Capturing the rationale and context for individual information
and knowledge artifacts can shift interpretation from That's stupid!
to Oh,
that's what they were thinking
and by doing so, improve understanding and
reutilization of valuable knowledge assets.
Keywords
community, decisionmaking, Dilbert, dysfunctionality, document lifecycles, information economics, information politics, knowledge management, metadata, SGML, stakeholder interests, values