Decision Support Systems

Much of this section is summarized from:
Barrett, J.R. and C.H. Castore. 1989. Decision making and decision support. In: Barrett, J.R. and D.D. Jones (eds.) Knowledge Engineering in Agriculture. ASAE Monograph No. 8, ASAE, St. Joseph, MI.

Introduction

Decision making involves the mind, the brain, sensory mechanisms, perception, cognition and the expression of results. An individual feels, perceives, thinks, remembers and reasons in an adaptive conscious and unconscious manner. Knowledge is reviewed, perhaps intuitively, ordered according to importance, and alternatives identified and ranked. Short- and long-term consequences are explored and choices made. Systems analysis often provides key inputs to the above process.

The human decision making process uses symbolic reasoning. It is not done by solving sets of equations as in LP or performing other complicated mathematical computations. Instead, symbols chosen to represent problem concepts are used and then various strategies are applied and heuristics are invoked to manipulate these symbols. A symbol is a string of characters that stands for some real-world entity or concept. Symbols may be likened to pictures that the mind sees or has fixed in memory. This is in contrast to values that are measured and quantified numerically. Of course, a number may be a symbol, but a symbol does not have to be a number. Much of the focus of artificial intelligence and expert systems has been on the definition of symbols and their manipulation for decision making purposes.

When an individual is confronted with a problem, several things happen. Some happen immediately while others are delayed. If the situation has not been experienced before, there are three possibilities that are explored to see if something might be done differently.

The circumstances may be comparable to others faced, but since dissatisfaction occurred with the results of actions in those cases, the decision is made to investigate other alternatives.
The problem is seen as somehow different enough from any dealt with in the past that examination of options available is warranted.
Portions of situations are recognized as having been dealt with either successfully or unsuccessfully, but there are additional new elements in the problem at hand; and it may not be clear if previously successful actions are appropriate, or how they might be effectively combined to produce success.

In all three cases, the individual will begin a sequence of activities involving:

defining the problem to be solved or establishing the parameters of the decision to be made
generating a list of possible alternative actions
evaluating, as best as possible, the relative risks, costs and benefits associated with the various options available
carrying out the alternative(s) chosen

At all stages along the way in this chain of actions, the individual is confronted with feedback leading to a revision of views on what actions are available, and their relative merits and drawbacks.

Experts and Their Decision-Making Process

Experts follow much of these same processes, only they have more experience and have more knowledge than those less expert. When experts make intelligent decisions they systematically consider knowledge that has been experientially acquired along with information that is sensed and observed. An expert might be defined as: An expert is a person who, because of training and experience, is able to do things the rest of us cannot; experts are not only proficient but smooth and efficient in the actions they take. Experts know a great many things and have tricks and caveats for applying what they know to problems and tasks; they are also good at plowing through irrelevant information to get at basic issues and they are good at recognizing problems they face as familiar instances. Underlying the behavior of experts is the body of operative knowledge we have termed expertise. It is reasonable to suppose, therefore, that experts are the ones to ask when we wish to represent the expertise that make their behavior possible.

How Decision Support Systems Work

The conceptual components of decision support expert systems are given in the figure below. No problem encompasses all the blocks, yet all fit within the structure.

User Dilemma

The end user determines the question that is addressed by the program. Different users will have different questions. The program developers' dilemma is to develop a program that makes sense to the end user, whomever they may be. Good advice is "know thy user for he is not thyself."

Problem Definition and Formulation

Development of a clearly defined problem and the establishment of objectives are discussed in Building Expert Systems. Two sets of information are needed to start a decision support program: one characterizes the situation and the other describes the general domain where the problem resides.

Description of Situation

This block represents the inclusion of information, through symbol definition, describing the specific conditions of the situation. For example, the condition of a diseased plant or animal, the specific field description, the farm status, the economic picture, etc.

Description of Domain

This represents the inclusion of information describing the more generalized domain of knowledge that will be needed to allow solution of the problem. This is descriptive information on a time basis.

Processing by Computer

This block represents the flow through the program as ordered by the logic of the expert and the programmer and accomplished in the machine.

Experts' Reasoning and Logic

The knowledge structure that represents the problem solving logic of the expert is usually kept separate from the coding that arranges the machine specific input and output.

Knowledge, Information, Data Sources

The facts that are used in the logic of the solution process may be in a variety of forms, including tables, algorithms, images, other programs, sensors, etc. Included are:

Description of Status. These symbolically describe the circumstances under consideration.
Data Sets. These are databases, assumed to be valid unless qualified.
Experientially Learned Information. This has been accumulated over time case-by-case and has associated levels of certainty and possibility.
Neighbors, Family Members, Associates. These facts and factors are confounding, critical and given much consideration.
Deterministic Factors. Clearly, without question, yes-no relationships.
Algorithmic Relationships. These are experimentally derived functions and have associated probabilistic accuracies.
Models and Simulations. These are descriptions of systems that may interface and can be mathematical or otherwise descriptive.

Knowledge, as assimilated information, is structured through knowledge engineering to help solve ill-structured, ill-defined agricultural problems.

Presentation of Alternatives

The usual way of bringing possibility and uncertainty into the process is to present ranked alternatives to the user for a choice to be made. The correctness of the proposed choice as evaluated by the expert and the end user, can be expressed relatively. This handling places the decisions that involve spending resources in the minds of the managers, or other users and not in the hands of the modelers!

Selection by User

Here the user selects an alternative for processing through the remainder of the program.

Projection of Consequences

The projection of the short- and long-term effects of a decision is an area that needs more development. Work is being done to bring the human considerations into the process that are necessary to make accurate, believable projections. These projections would allow the possibility of examining the effects of choices through sensitivity analysis.

Social and Economic Factors

The interaction of social and economic factors are inputs that are used in the projection of consequences.

Information for Future Use

This may be as hard copy of the output from the running of the program or as information stored in memory for use in initiating the next run of the program.