Articles
ADVANCES AND BOTTLENECKS IN MODELS FOR FARM MANAGEMENT AND DECISION SUPPORT SYSTEMS: SUMMARY OF A GROUP DISCUSSION
There are a number of DSS types, each with different purposes and different approaches.
The main commonality among DSSs is that their purposes are to provide information to help improve management or control decisions (not to make the decisions per se). They may be aimed at tactical of strategic planning decisions or at operational control.
We usually think of DSSs as model-based, meaning that one or more models are contained in the system to help interpret data and suggest courses of action for the user.
However, data are also very important for all DSSs, and some may depend almost entirely on databases.
DSSs may include complex process models of crops/greenhouses, simplified models derived from complex ones or from data, expert systems, decision trees, etc.
They may be useful for researchers, advisors, or farmers, but the content, operation, and information provided by each will be different for different users.
Other common characteristics of DSSs are that they must be simple to operate, easy to interpret, and provide relevant information to users.
Complexity of models for use in DSSs is an important issue; some attempts at developing model-based DSSs have failed because of the complexity of models used and the numerous inputs required for their operation.
Models to be used by grower DSSs should not require many parameters.
Otherwise, DSS operation may not be justified in terms of time and money relative to the value it provides.
Generally, models in DSSs should be simple, although the scope of the problem to be addressed may require more complex models, particularly if the decisions are not structured.
In this case, users may want a wide range of information, which may only be possible with complex models.
One very important consideration in the development of DSSs is that users must be involved from the very start of the effort.
They should be involved in the definition of the problems to be addressed by a DSS as well as its design, evaluation, and refinement.
If users participate throughout the process, including supplying some of their own data for evaluation, they will help ensure that its operation is practical and will gain confidence in the system for their own uses.
Confidentiality of data must be respected when this is attempted.
At times, researchers may have their own goals for modeling for scientific purposes, then expect their work to be adopted by users.
Having end users involved from the start will help eliminate this problem.
Scientific modeling should be justified on its own merit, to produce scientific articles and new knowledge, and may not directly contribute to a product for direct use by advisors, growers, or even other researchers.
DSSs may be more acceptable for some uses if they address qualitative as well as quantitative information.
This has two aspects.
First, quantitative models are not available to address all aspects of decisions.
Thus, qualitative information may be required, and models such as expert systems may be needed to fill gaps in quantitative knowledge.
Secondly, not all users are the same in terms of attitudes to risk, preferences, and level of understanding.
Thus, practical DSS may benefit from components that provide options with interpretations that take into account these differences.
This is a