# Some concepts and distinctions of Making and Decision Aids in system perspective

In order to understand the nature of decision-making (and the need for computer support) a description of the characteristics of a decision situation must first be given. It can be defined with the items below.

• A problem exists.
• At least two alternatives for action remain.
• Knowledge exists of the objective and its relationship to the problem.
• The consequences of the decision can be estimated and sometimes quantified.

Generally, different aspects such as economic, environmental and political fitness create the background of a decision. In this setting the decision maker first has to find existing alternatives and then make a choice between a set of them (normally small). In turn, every alternative has consequences, connected to the aspects via the alternatives. Given these sets of aspects and consequences the decision maker has to choose the best alternative.

All decision situations belong to one of the following three classes.

• Decisions under strict uncertainty. Here the decision maker is unable to know anything about the situation. Quantification of the uncertainty is not possible.
• Decisions under certainty. In this case the decision maker has full knowledge of the situation, and the consequences of the decision can be predicted. The alternative which has a value not less than the value of any other alternative is chosen.
• Decision with risk. In this situation, the decision maker is able to quantify the uncertainty by assigning probabilities, generally known in advance, to each alternative. Note that the level of risk increases exponentially as more data are left out. A decision-maker must also be aware that unlikely events is likely to happen because there are so many unlikely event that could happen!

A special problem with decisions under strict uncertainty is the famous “Prisoner’s Dilemma”. This problem was first formulated in 1950 by two American mathematicians, Flood and Dresher.  It presents the very essence of personal decision-making related to both cooperation and defection and is a typical game theory problem.

In the Prisoner’s Dilemma, we assume that the reader and his companion has taken part in a crime with a stolen car and now are in custody, held for questioning in separate cells. They cannot communicate with each other. The procecutor offers each in turn the following deal: “We have evidence enough to convict both of you for the crime. If neither of you confess, you will probably be sentenced to one year each (for the stolen car). If you confess and testify against your partner, you will be par donned and he will get ten years”. “But if we both confess?” you ask and the reply is that “you will both get four years” (confessions are mitigating). The best for you and your companion would be to stay mum, both getting off with light sentences. This, however, demands cooperation and communication which is not possible for the moment. If you are a scoundrel, why should you care for your partner? And if he gives way to the same temptation? Then you will be the sucker who spends ten years in jail.

The existing alternatives say that if your partner has decided to stay mum, it will be best for you to confess and avoid all punishment. If he confesses it will no doubt be best for you to confess too, because then you get four years instead of ten. So either way, it is best for you to confess and by doing this you know that your partner will do the same thing. Apparently it seems inevitable that both of you will spend four years in jail, although there were alternatives which both of you knew about resulting in one year each. The Prisoner’s Dilemma gives a good example of how a rational, selfish analysis will have consequences not within the real interests of the those involved.

In the real world, many situations are like the Prisoner’s Dilemma. The difference is that companions can communicate and the process is repeated. Statistical analysis and computer simulations have shown that the most successful strategy will be to use either a randomized or a conditional response and use the memory of previous dealings with the same partner. Called TIT FOR TAT, this strategy tells us to cooperate on the first trial with a new partner. On every subequent trial, one has to do what the other player did on the preceding trial. It can be found that those who never defect before their partners were, in the long run, the more successful. Totally random strategies, fixed strategies or those totally unresponsive to the actions of a partner can never succeed as they invite defection which will go unpunished.

The secret of TIT FOR TAT is simply that it spends more of its time to cooperate than defect with partners. That this also does the partners some good makes it the best available strategy. Interesting is that TIT FOR TAT- strategies lead to cooperation in the natural world even without intelligence.

Furthermore, a problem can be solved, resolved or dissolved. A solution occurs when one of the alternatives in the existing repertoire is chosen and then implemented. To solve the problem is to select the means which yields the best possible outcome. To resolve a problem is to select a means that gives an outcome good enought. If the whole view of what is defined to be a problem is changed, or if a completely new alternative is generated to handle the situation, the problem is said to be dissolved.

For the decision maker, the ultimate goal is considered to be to arrive at as effective and precise decisions as possible. But in reality, different qualities of a decision must be accepted as inevitable. R. Ackoff (1970) has categorized the quality of a decision into three levels. Optimizing implies finding the best possible existing solution. The tools available in such situations, apart from the decision maker’s own intuition and experience, are different types of models to support the decision maker. The majority of these models are, however, either of a mathematical or statistical origin and of a rather complex nature. With the aid of a computer, these models can be handled more effectively and efficiently, often by use of an algorithm. An algorithm is a step-by-step procedure (often of a mathematical nature) that guarantees that an optimum solution is achieved.

Satisficing is attaining a certain minimum quality level for the decision, enough to solve the problem but not necessarily more. The satisfier seldom evaluates the existing alternatives, because the first acceptable solution is considered to be as good as all the others. To satisfy is to use the principle of least effort. Most satisficing problemsolving strategies are based on heuristics — rules of thumb that are good enough for most decisions.

The main reason why decision-making in most cases appears to be satisficing is the following limiting circumstances:

• Limited time. The decision has to be done in a finite amount of time.
• Limited information. It is impossible to gather all necessary data relevant to the problem due to finite resources.
• Limited information-processing capability. Most human can only handle about seven items of information at any one time.

Idealizing is to change the whole system or its environment in order to bring it closer to an ultimately desired state, where the actual problem does not arise. Ackoff himself called this alternative for the design approach.

Ackoff is also the originator of an often cited typology of decision making and planning (Ackoff 1981). This includes four basic orientations with different temporal adaptation. Ackoff, however, emphasizes that these orientations are like the primary colours; seldom to appear in pure form.

If the dominant orientation belongs to the past it is said to be reactive. The reactivist seeks to return to a previous state in his decision-making. He is nostalgic about the past and has a better view of where he has been than of where he is going — that is, to drive into the future while looking in the rear-view mirror.

The inactivist is satisfied with things as they are. He does not want to return to a previous state, is not fond of the way things are going and tries to prevent change. Survival and stability is his primary aim. The strategy could be defined as satisficing by muddling through. Ackoff thinks the best example of an inactivist organization is the typical university which is as ‘difficult to change as a cemetery and for the same reason’.

Preactivists believe that the future will be better than the present or the past and subsequently try to accelerate change. New technique and technology is considered as a general panacea and experience is not considered very valuable. Generally, errors of commission are less costly and easier to correct than errors of omission. The preactivist style is normally to optimize.

The proactivist is not willing to return to a previous state, nor neither to accept thing as they are or to accept the future that appears to confront him. The future is largely a subject of his own creation. Learning and adaptation are lodestars as no problems stay solved for long. The style is to idealize and to develop.

A metaphorical summary of the different typologies is that the inactivist tries to hold a fixed position in a moving stream; the reactivist tries to swim against it; and the preactivist tries to ride with it along its leading edge. The proactivist in his turn, tries to change the reaches of the river.

Interesting arguments questioning the need for optimization have been published by C. Holling (1977). He notes that ecological systems strive to maximize options rather than to limit them by selection of an optimal alternative. From a human point of view, the possibility of a bad choice and pertinent failure is not rejected. Instead, a strategy which minimizes the cost of such a choice is applied. In that way, efficiency is sacrificed for adaptability.

A theory of decision-making has  long existed  in economics,  being associated with the idea of homo economicus, the strictly rational decision maker. This ideal human being has the following qualities:

1. He can always make a decision if faced with a number of alternatives.
2. He ranks the consequences on a scale of preferred results (a value- scale).
3. His order of preference is always transitive (first A then B, not C then A).
4. The first alternative is always chosen (utility maximizing).
5. The same choice is always made if the situation is repeated.

According to Herbert Simon (1976) the process of rational decision- making is an act of choosing among alternatives which have been assigned different valuations. It involves the following process:

1. Listing all of the alternative strategies.
2. Determining all the consequences that follow upon each of these strategies.
3. Comparatively evaluating these sets of consequences.

Simon, however, admits that total rationality is an unattainable idealization in real decision-making — who can be aware of all existing alternatives?

The task of making decisions can generally be seen as an (iterative) procedure of information gathering and processing, summed up by the following keywords.

• Intelligence: Find raw data to be processed
• Design: Evaluate different alternatives of action
• Choice: Choose one of the alternatives
• Implement: Firmly establish the chosen alternative
• Control: Check that orders are obeyed and make necessary adjustments

The decision maker can only obtain information and thus only have real knowledge about future development within a rather short time frame. This frame is defined by the decision maker’s and his assistant’s specific knowledge about the expected and agreed consequences of already known projects. The first step in the above process will inevitably suffer from an inherent discrepancy called ‘the information dilemma’ or even the ‘uncertainty relation of decisionmaking’. It is associated to the need for explicit and actual information and states: ‘the precise information is not timely, and the timely information is not precise’.

It is important that the whole procedure is understood to be cyclical; repetition and feedback of certain steps are virtually always indispensable. Even redefinition of the original problem and the existing alternatives may be necessary.

Using the above keywords, the following more detailed steps in the decision-making process can be elaborated.

1. Identify the problem (recognize a situation that requires decision/ action)
2. Gather the facts which will affect the decision
3. Generate possible alternative solutions
4. Specify the alternatives
5. Select the best one
6. Gain acceptance by motivating/explaining the basis of the decision to other members of the decision-making group
7. Communicate the decision to all those affected
8. Put the decision into action
9. Supervise the execution

The decision maker thus has to choose the best option given the existing set of consequences. Note that a non-decision exists as an ever- present alternative (unfortunately most often the worst!). A classic reminder to the decision maker is ever present: ‘You may be so preoccupied with doing things right that you forget to do the  right things.’

When ready to make his choice, the decision maker will meet four basic types of difficulty:

• How to compare the alternatives with regards to different aspects of the decision.
• How to compare the alternatives within each aspect.
• How to estimate the probability that the given consequence will occur if a certain action is taken.
• How to estimate the value of the consequences.

It is, however, obvious that in many critical situations where rapid decisions are necessary, the decision maker has to act in a less analytical and more instinctive manner. There is no time for careful calculations when the sabre-toothed tiger is attacking. As evolutionary survivors, we have brought with us mental decision tools based on very little information and simple rules or decision heuristics. Although applying to different kinds of problems, these rules have a common structure and use a mixture of probability and chance as basis for decision. First, we search the environment for cues or information, upon which to base a choice. The heuristics contain rules which direct the search and a stopping rule after a few cues have been analyzed. After that, the choice is made — to run, to stop and fight or to attack. But of course all the benefits of a rapid decision is wasted if the wrong decision is made.

In most situations, the choice of what you recognize works better than a choice at random. But heuristics does not work well when you know too much. Regarding emotions, they are part of each heuristic rule and help us to make correct decisions. Outmost fear will certainly reduce the options to only one; to run away. Unpredictability can also be a part of heuristics, especially in social decision-making. Sometimes it can be rational to be inconsistent.

A short look at the internal nature of problems which have to be solved by different kinds of decision reveals that they may be structured, unstructured or semi-structured. Structured problems are those for which we can define an explicit procedure to solve the problem. An example is the construction of a schedule for the use of existing classrooms in a school. To solve unstructured problems the decision maker must show judgement, evaluating capacity and insight into the problem-definition. Political decisions are often unstructured as their success is dependent upon the changing opinions and hidden beliefs of the people. Semi- structured problems are partly structured and partly unstructured.

If the structure of a problem is related to the operational, tactical or strategical decision levels identified in most major organizations, the examples in Table 9.1 will be typical.

Not apparent from the table is the general tendency to find the majority of structured problems in the operational level. Also, most semi- structured problems are to be found at the tactical level, while unstructured problems are most common at the strategical level.

Structured decisions  seldom  involve  managers  and  can  hence  be made by lower level personnel or by a computer. Semi-structured decisions are by their nature appropriate for managers with computer support. Computational complexity, problem size and  the precision of the solution often make strictly managerial judgement insufficient. Unstructured problems are not able to be formalized in a technical sense and hence are impossible to feed into a computer. The nature of the problem, the volume of data, or the lack of an appropriate method make any decision entirely dependent upon human experience and intuition.

Source: Skyttner Lars (2006), General Systems Theory: Problems, Perspectives, Practice, Wspc, 2nd Edition.