Rene F. Reitsma
University of Colorado at Boulder
Center for Advanced Decision Support for Water and Environmental Systems
(CADSWES)
Campus Box 421
Boulder CO 80309 - 0421
e-mail: reitsma@colorado.edu

Trivial Pursuit and a Few Suggestions

1. Trivial Pursuit?

Thinking about collaborative decision-support makes me nervous. I know, very few people are likely to care or to be even remotely interested in me being nervous, but the whole concept of collaborative decision support throws me for a loop. For crying out loud! I am still trying to find out what "decision support" in general is, and whether or not it is anything at all! This is indeed very frightening, especially since I pretend to make a living off of the design, development and implementation of (environmental) decision support systems (DSS). But then again, I've always been a little slow.

Collaborative spatial decision support. Ok. So there's three aspects to this thing:

Perhaps the three aspects are related in some sort of set-theoretical way. Something like: collaborative spatial decision support is decision support for spatial problems which require collaboration. Hmm, not bad. (with that I mean that I can still follow; you see, a child's hand is easily filled). Now the "spatial" part does not seem to be too hard. Although many decision problems have a spatial components, in most cases that spatial component is not very interesting. I mean, the fact that the Colorado river is located somewhere, is rather irrelevant when it comes to deciding how much water to release from Hoover Dam, except that the locational aspects are figured into the equations which describe the time the water needs to travel from its point of release to its point of consumption. Similarly, the fact that an in-stream flow right is established at a particular reach of a river is relevant in that it figures into the equations which express the flow over time at that and other locations on the river, but the spatial aspect of a problem such as this is at best one that functions as a constraint on the decision to be made. That is, if indeed the decision problem involves the determination of a release or diversion schedule, and the spatial aspects are given.

Spatially much more interesting are problems of a locational nature; for example, decisions as to where to establish that in-stream flow right, or whether to change temperature regimes below Lake Powell in order to sustain certain fish species.

Does the latter type of problem; the problem where space is to be considered a resource to be allocated among subjects or objectives associated with subjects, constitute an interesting, or better, "rich" enough set of problems to consider them "spatial" problems? Well, perhaps. Certain spatial location and allocation problems seem complex enough from a spatial point of view to warrant special attention, spatial modeling, support and consulting. However, in many cases the spatial aspects seem rather trivial when compared to the other, non-spatial aspects of even spatial problems. For instance, the allocation of a particular space to, say, residential versus open space objectives (hiking, biking, recreation, wildlife, etc.) could, under certain circumstances be considered spatially trivial. It's zoned A or B or, in the most complex case, some of it will be zoned A and some B. What really matters in those circumstances, is how a complex array of objectives, both momentarily and relative to the future (strategic, tactical) interact with a) (perceived) opportunities for satisfying these objectives and b) with each other.

2. Coordination Theory Perhaps?

Coordination theory (e.g. Mallone and Crowston, 1990) provides a framework of studying these interactions between objectives, actions, and resources. The theory (I really need to start cranking at this stuff, especially since I'm in the process of writing a paper about it(!)), provides an integration of elements from game theory, operations research, and various aspects of decision-making, negotiation, and collaboration. One of its central concepts is that of harmonious work, or harmonious action. Harmonious actions are defined as actions which jointly increase utility in terms of the satisfaction of objectives. The actions can be mapped in action or decision space. Different regions of the space have different values for various objectives. Trajectories through areas which have high utility values for joint objectives can be labeled harmonious. Clayton Lewis (University of Colorado at Boulder - Computer Science) and myself, are currently working on a framework which extends the basic notion of harmonious behavior to cases which involve conflicting objectives. We aren't quite sure yet whether or not we are trying to prove things by just changing the definition of harmony (something which would make Popper turn around in his grave (rest his soul)), or whether we are really on to something, but the idea is that even in cases where objectives are clearly conflicting, harmony is possible, albeit in terms of increase in efficiency only. The notion of conflicting objectives also necessitates the distinction between stable and unstable harmonies. Harmonies might be established through increased efficiency (dictatorships and swindles are good examples), but such harmony, although momentarily preferable, is unstable in that sooner or later the swindle is revealed, the dictatorship becomes unbearable, or, more interesting, the environment changes such that the values of resources allocated to the various objectives change. Examples of the latter are the re-evaluation of city-centers leading to such processes as gentrification, or the suburbanization of the Colorado Front Range, an area where not long ago not many people wanted to live, but where growth seems nowadays to be almost out of control.

Whether coordination theory is an applicable research topic for I17; we can talk about that. Spatially, applications could be possible in mapping conflict vs. harmony, or degrees of harmony in space. This would be particularly interesting if the utility functions which would compute the valuations of space as a consequence of a series of actions (policy alternatives) could be dynamically linked with their mappings into space. Note that in these cases harmony would not be computed as a simple compensatory model of momentary objectives. Harmony rests in the process and the valuations of its intermediate steps or moves relative to objectives, not the end-result.

From a non-spatial point of view, coordination theory offers at least a means of formalizing collaboration, in terms of harmonious or inharmonious actions and their valuations relative to one or more objectives. Although it remains to be seen how complex these formalizations must be in order to adequately represent the salient features of, for instance, environmental negotiations, it does at least provide a conceptual schema for an attempt to do so.

On this latter point, coordination theory could perhaps be seen as (yet?) another attempt at operationalization of the well-known Brown-Moore model of (spatial) decision-making (Brown and Moore, 1971). One of more fascinating aspects of that model was that valuations of empirical (spatial) situations is the result of a dynamic, complex interaction between

  1. objectives,

  2. means to satisfy these objectives, and

  3. the opportunity to modify the environment or context which provides these means and shapes the objectives.
I know, lots of folk thought the stuff was trivial and rather frivolous, but I've always been fascinated by the incredible difficulty of operationalizing such a simple model. Coordination theory, by nature of the way it tries to represent these three aspects, provides remarkable insights into, especially the third alternative, that of modifying the environment thereby opening up whole new areas of the (previously hidden or unattainable) decision space and hence opportunities for collaboration.

3. New Modeling Paradigms Perhaps?

Something which has been bugging me for a while now, is the use of mathematical models of the physics of environmental resources for environmental decision making. As part of a study into the role of simulation models in environmental negotiation, Ilze Zigurs (University of Colorado at Boulder - College of Business), Clayton Lewis and I have

conducted experiments were we had subjects resolve a water resources allocation problem supported by a simulation model of the resource under various policy alternatives. We experimentally manipulated the conditions under which that model could be used, both in terms of frequency and private vs. group access (Reitsma et al. 1996).

conducted a detailed study into the use and role of the Colorado River Simulation Model (CRSM - Schuster, 1987), in the determination of the Colorado River Annual Operating Plan (AOP).

Where the results of the former seem to indicate that making the model more and easier accessible did hardly contribute to the quality of the problem solutions and the resolution process (both objectively and as perceived by subjects), the latter revealed that although the model maintains a central place in the negotiations, the vast majority of negotiations are about aspects of river management which are not in the least represented by the model! Yet, all participants in the AOP process want increased accessibility to the model, and significant system development and modeling are underway to comply with these requests.

Assuming that these observations and generalizations are correct, the only way in which we can explain these results is in stressing:

1. the managerial role of models in environmental management; i.e. the model as a managerial tool, a tool for establishing due process in the negotiations, regardless of the model's correspondence, and

2. reliance of participants common knowledge and reasoning capabilities to evaluate the really relevant aspects of policies. In other words, although the model provides some sort of rational, impartial, formal core to arrange a negotiation around, many of the real objects of negotiation are not modeled at all and are dealt with on an ad hoc basis. Interestingly, in the paper on coordination theory (in preparation) we argue that the latter forms a nice example of harmonious behavior in that the instigation of the AOP process itself changed the context of collaboration such that more harmonious behavior was possible.

Based on these observations, the following research questions could be thought of:

1. Validating physical models in collaborative (or negotiation) settings. Our results show that the value of these models in collaborative and/or multi-objective settings is low, except from the point of view of providing (quasi) rationality to the process. Yet, vast amounts of resources are spent on the development of these models (yes, yes, mea culpa). Question: can this claim be upheld? In other words, what is the role of these models, and how does the rationality of the models correspond with the rationality of negotiations, disputes, or policy making?

2. If it is true that physical models represent so little of the true aspects of many decision problems in, for example, environmental decision making, could those aspects be modeled differently and ought they be modeled? What other modeling paradigms are there which could contribute to the construction of better decision-making models?

3. Is it possible to develop other types of environments (e.g. GDSS, computational planning games (see below) which better support complex decision making?

4. Perhaps Computational Gameboards?

For a while now, Ernesto Arias (University of Colorado, Dept. of Planning and Environmental Design), is trying something different. He develops planning games; board games with boards (space) and pieces (choice attributes) and rules (rules of process), and uses these games to support collaboration among policy and decision makers, especially in the context of physical planning. Recently, Ernest let me in on a plan to develop an interactive, computational gameboard for spatial planning. The idea is that of an interactive computer monitor/screen, tilted on its back, projecting upward onto a large "gameboard" (the technology would work a bit different, but that's the basic idea). Unlike a computer screen, the gameboard would allow people to walk around it, put game pieces on it (add a third dimension), use it as a working environment, etc. But unlike a traditional gameboard, it would be interactive and would dynamically adjust to the situation of the game. Although the former is approached in GDSS decision rooms with screens projected on the walls of the room, we expect that the simple availability of the "screen" as a flat space so that it can be used as a table or desk, or more important, that the space represented by the gameboard can be extended with three-dimensional pieces, may dramatically increase computers' propensity to serve as a collaborative decision-making tool.

When Ernest and I started to work out the idea some more, it appeared that one of the things that was more or less missing in all of this, was a theory of why these gameboard planning environments could work, and ideas on how to (experimentally) evaluate that theory and modify the games accordingly. Since we started working on this we have made progress in these areas and hope that soon we will be able to test some of the theory out in experimental settings.

5. Does it make a difference?

One of the things we badly need in environmental decision support (collaborative or not) is methods and techniques for the evaluation of the effectiveness and efficiency of DSS. Yes, we can easily keep track of system use (although it is often very hard to control these data for frivolous use and mistakes in use (clicking the wrong button, that sort of thing), and we can go back in after a system has been on-line for a while and ask people whether they like it, what else they want and whether or not they use the darn thing. But assessing real utility often is a matter of letting the market take its course. If they come back for more, than it must be good! Needless to say that although such evaluation can sometimes be handy (isn't it great to tell others that you just got another 900K in DSS development money?), success or failure of a DSS often rests in quite different measures. Unfortunately, little evaluation of at least environmental DSS is ever conducted. In most cases, when people are happy about the product they do come back and ask for more features, additional modeling, etc. But an assessment of how much the DSS has contributed to organizational efficiency, internal communication, or savings in or modified distribution of, for instance, water resources, often lacks.

Note that this issue has ties with the issue of the utility of models for decision support; namely the question: does it make a difference, and if so, how so and how much? It is interesting that as geographers or social scientists we spend so much effort evaluating the effects of policy and or the changing landscape of geographical patterns and processes. Yet, at the level of decision support we are often content with simply putting the thing together, throwing it over the wall, and hey presto, another branch on the tree of bleeding (!= leading) edge of technology.

I feel that if research initiatives are developed which are aimed at developing collaborative decision support tools, at least some amount of resources ought to be spent on the systematic evaluation of those tools in practice. This requires the development of evaluation techniques and metrics, as well as the development of data representation schemes which efficiently represent system use information; e.g. in the form of sequences of use. Much of this work has already been addressed in Group Decision Support Systems (GDSS), but it can be argued that in the context of dynamic, physically based systems such as environmental systems, this question needs to be looked at.

Such development of evaluation methodologies also applies to the use of system design methodologies. Although more an issue of software engineering in general, many of us who have developed systems before, have used different methods of designing, developing and implementing (fielding) systems. Some we liked, some we didn't. But as so many writers on software engineering point out, few of us can clearly articulate how we developed our systems, let alone provide clear statistics and measures of the system design and development process.

What we should not forget, however, is that software systems are complicated pieces of (logical) machinery, for which people often pay sizeable amounts of money and which are expected to play important roles in organizations. As such, they deserve scrutiny in design, development and implementation. Although in most engineering and sciences there is a clear distinction between the academic phases of new technology and those were the new technology goes into production; each with their own internal and external metrics of quality assurance and control, in decision support the distinction is often blurred. This leads to an often very messy situation, in which academics see themselves confronted with system development tasks that they really do not want to conduct, the more so since the relationship between effort required to build a software system well and the academic credit received for such efforts is such that many systems are "thrown over the wall" or abandoned in prototype state, a situation which is most annoying for all involved in the project.

One perspective of design and development methodologies is that they should represent the results of some of the research questions formulated here and elsewhere in the position papers. Design methodologies can be seen as the engineering aspects of building collaborative decision support tools. If the science behind is any good at all, it must be possible to translate the scientific results into concrete methodologies for system design and development. Seems like an excellent way "making a difference".

References

Brown, L.A., Moore, E.G. (1971) The Intra-Urban Migration Process, a Perspective; in: Bourne, L.S. (ed.) The Internal Structure of the City; Oxford University Press; New York; pp.200-209.

Malone, T.W., Crowston, K (1990) What is Coordination Theory and How Can It Help Design Cooperative Work Systems; ACM CSCW 90 Proceedings; pp.375 - 388.

Reitsma, R.F., Zigurs, I., Lewis, C., Sloane, T., Wilson, V (1996) Experiment With Simulation Models In Water Resources Negotiations: ASCE Journal of Water Resources Management and Planning; forthcoming.

Schuster, R.J. (1987) Colorado River Simulation System Documentation. System Overview; U.S. Department of the Interior, United States Bureau of Reclamation, Denver, CO.