Decision Making

If we want to train people in decision making, it is important that we have some idea of what type decision making we want to practise. Analysing the characteristics of the decisions that arise in the task is therefore important. Basic questions that we may ask are: What is a decision? When, how and why are they taken? These are hard questions, but attempts have been made both in classical decision theory (Edward 1954; North 1968; Hammond 1980), and in later research as naturalistic decision making (Klein 1993, Hammond 1993). One problem for researchers in decision making is that in real life decision making is often characterised by the fact that the decision problem is not usually well-formulated; it can involve goals such as ”I want to feel good”. The Gestalt psychologist Karl Dunker (1945) states: “a problem arises when a living creature has a goal but does not know how this goal is to be reached”. When the concept of decision making is connected to problem solving, a decision can be viewed as a process where we select some actions that hopefully will lead us to a satisfactory result with respect to the goal.

Edward (1954) was one of the first classical decision theory researchers to describe the process of decision-making. In its simplest form decision making can be viewed as generating all (action) alternatives, validating the alternatives and selecting one alternative. This perspective has been the basis of classical decision theory. According to Bidgoli (1989), Herbert Simon describes the decision-making process as consisting of three phases wish he calls intelligence, design and choice. See figure 1.

Figure 1. Stages involved in the decision making process.

This view of decision making is probably insufficient if we want to train decision making in emergency organisations. What we need to do is to look at decision making in natural settings. An example of a natural decision-making model (described later in this chapter) is the recognition-primed decision model that is developed during studies of urban-fire ground commanders. We should also look at how distributed decision making is influenced by dynamic behaviour in complex emergency organisations. All this should be done in a context of real work procedures that a staff in an emergency organisation use.

Example Here follows an example of a work task for a commander and staff when they are commanding and controlling an organisation that exists in a large complex dynamic system. The example involves the work of a forest fire-fighting staff. A forest fire-fighting staff has directed four fire-fighting units to a location near a forest fire. The staff has been informed that a fire has started in the forest north of the position where the fire-fighting units are. The doctrine for the staff is as follows: one, save lives; two, save valuable objects; and three, extinguish the fire. The staff have a map that says that there are houses on both the west side and the east side of the fire. They have been informed that the wind is blowing from north to south.

The task for the staff is to decide how to handle the situation. Important questions include: Are human lives or houses in danger? What are the main features of the geographic environment? What type of vegetation is there in the area? What are the possible critical future situations? Based on answers from questions such as these, the staff make a risk analysis.

No people or houses seem to be in a directly critical situation so the fire-fighting unit does not need to make an offensive attack on the fire in any critical place. They choose to use a defence strategy where they fight the fire from two fronts, south-west and south-east, and press the fire inwards so that the southern part of the fronts meets. To check for future critical risk situations, they also command some reconnaissance persons to check the geography and vegetation of the area. Some of the decisions about proper strategies to solve the tasks are distributed to the subordinate fire fighting units.

As time progresses the fire-fighting units start to do their work. After some time the staff get information that tells them that the wind is getting stronger and that the geographic area east of the fire is a critical area where the fire can spread fast. They realise that the house east of the fire can be in a critical situation if the fire develops in that direction. They discuss the situation with the fire-fighting unit personnel and decide that they should make an offensive attack against some parts of the fire on the east side. This will result in more forest being burned down, but the house would not be in danger any more. To inform the subordinate units about the new situation all fire-fighting units get information on the new status and some get new orders for the situation.

Some time later they get a telephone call from a civilian who says that there are some houses in the forest west of the fire that are not on the map. They also get a message from a fire-fighting unit that says that they can’t fulfil their orders. Some time later one fire-fighting unit says that they see smoke from a certain position, but the staff have also got information from a fire-fighting unit near the position that says that there is no fire there.

The example above shows a typical decision situation in real life. First and foremost, decision making is not a task; the emergency management commander does not set out to ‘make a decision’, he sets out to control the resources they have available. Achieving this goal requires making many decisions, but decision making is part of a broader activity, not an end in itself. What the decision maker does is to size up the situation to a situation picture, based on the fact that he or she makes diagnoses, hypotheses and decides on a course of action. We can say that the decision maker reacts to the input information and processes it in a context of his own expertise. He or she is often not aware of any complex analytic thinking. This view of decision making forms the basis for the views taken in our research, and today this view is generally held in modern decision making research.

Decision Making In Natural Settings
Recognition-Primed Decision Model (RPD)
Distributed Decision Making
Tactical Reasoning

Decision Making In Natural Settings

Recently researchers have started examining natural decision making which can be viewed as decision making in natural settings. The researcher’s goal in naturalistic decision making is to study the decision making that occurs in real-world situations. Orasanu & Connolly (1993) have characterised natural decision making with these eight factors:
Ill-structured problems.
Uncertain dynamic environment.
Shifting, ill-defined, or competing goals.
Action / feedback loops.
Time stress.
High stakes.
Multiple players.
Organisational goals and norms.

This research field has resulted in a number of decision models. The models tend to have different views but they have all been generated from studies in natural settings. Two examples of models are:

Recognition-Prime Decisions (RPD) Klein (1993) has studied decision makers in operational settings and how they manage to be effective under high stress and time pressure. The primary conclusion is that the decision makers seldom analytically reach a decision by comparing different alternatives. Instead they assess the situation and select an appropriate strategy based on the recognition of the situation. More about the recognition-prime decision model follows below.
Search for Dominance Structure Montgomery (1993) suggests that in situations where people have a number of alternatives to choose from, they tend to look for a dominant alternative. A dominant alternative is an alternative that is perceived to be at least as good as all the other alternatives on all relevant attributes and better that each of them on at least one attribute.

Situation Assessment

One important step in natural decision making is situation assessment, as suggested by Noble (1986). Situation assessment is the process of ‘sizing up’ the received information into a mental picture of the situation (situation awareness). The process of situation assessment or ‘sizing up’ the situation unfolds as: First, concrete information on the situation is combined with additional background or ‘context’ information and general knowledge retrieved from the d ecision maker’s memory to form a tentative interpretation of the situation. This initial representation implies certain expectations of future events. These expectations are tested by additional information and the representation is refined or rejected in favour of a new representation, Lipshitz (1993). There seem to be four important aspects of situation assessment:

Understanding the types of goals that can be reasonably accomplished in the situation.
Increasing the salience of cues of information that are important within the context of the situation.
Forming expectations that can serve as a check on the accuracy of the situation assessment.
Identifying the typical actions to take.

Information Cues

The information that is used in the situation assessment comes from different cues; we can say that we have different sources to get information from. In the decision process a selection is made as to which cues are important and information that describes how to select the proper relations between the cues is used. Experienced people know which cues are important and how they should use them.

Recognition-Primed Decision Model (RPD)

One important model is the RPD model by Gary A Klein (1993). The RPD model asserts that people use situation assessment to generate a plausible course of action and use mental simulation to evaluate that course of action.

For several years Klein has been studying command and control performance. On the basis of his studies he has generated the RPD model for naturalistic decision making. Klein’s principal conclusion is that the decision makers assess the nature of the situation and, based on the situation assessment, select an action appropriate to it. Klein believes that the RPD model describes how decision making is usually carried out in real-world settings. The RPD process consists of three phases - situation recognition, serial option evaluation, and mental simulation - and he asserts that people use situation assessment to generate a plausible course of action and use mental simulation to evaluate this course of action. See figure 2.

Figure 2. The Recognition-Primed Decision Model.

Situation recognition At this stage the decision maker recognises the situation as typical or novel. Typical situations lead to typical actions, whereas novel situations pose new challenges that cannot be countered effectively by employing old routines. To recognise the situation and guide the selection of proper actions, the decision maker identifies critical cues that mark the type of the situation and causal factors that explain what is happening and what is going to happened. Based on these, he or she sets plausible goals.
Serial option evaluation In this phase the decision maker evaluates action alternatives one at a time until a satisfactory one is found. The actions are selected from an action queue, sorted according to typicality.
Mental simulation To evaluate whether an action is satisfactory, the decision maker acts it out in his or her imagination. He or she mentally simulates the successive steps to be taken.
This type of process let the experienced decision maker focus on critical cues and identify causal factors which focus the decisions on the important issues and reduce the information overload, Lipshitz (1993).

When Klein and his associates were constructing the RPD model a study was made where they observed and obtained protocols from urban-fire ground commanders (Klein, Calderwood & Clinton-Cirocco, 1986). The ground commanders' decision task was focused on initial search and rescue, whether to initiate an offensive attack or concentrate on defensive precautions, and where to allocate resources. The ground commanders argued that they were not making choices, considering alternatives or assessing prob­abilities. They saw themselves as acting and reacting, and modifying plans to meet the needs of the situation. Clearly, the fire ground commanders were encountering choice points during each incident. The commanders react to a situation in terms of highly familiar patterns associated with certain actions. If time permits, commanders used an apparently imagery-based consideration of the implications of the planned action, ‘watching’ it unfold in the present context to watch for any complications. During the interviews the fire ground commanders could describe alternative courses of action that were possible, but insisted that during the incident they did not think about alternatives or deliberate on the advantages and disadvantages of the different options. Once the fire ground commanders knew what type of case it was they had to deal with, they usually also knew the typical way of reacting to it. The used their time on a mental simulation of the current course of action. The explicit consideration of more than one alternative at a time was more characteristic of novices than of experts.

Distributed Decision Making

The decision making in emergency organisations and military systems has been classified by Brehmer (1995) as distributed decision making or team decision making. This means that decision making is distributed among the actors in the organi­sation. Emergency systems are also often based on a hierarchical organisation where the decision makers work on different time scales. The members of the staff work on a higher time level than the subordinate decision makers, and are respon­sible for the strategic decisions. It is important to understand that these systems are not totally controlled by the decisions made by the staff; decisions are made throughout the whole organisation and they are all involved in the success of the operation. Distributed decision making in an organisation makes it important to train communication and understanding of shared frameworks and goals. Team decision making can be viewed as distributed decision making where the co-operating actors have different roles, tasks, and items of information in the decision process (Klein & Thordsen 1989; Orasanu & Salas 1993).

Brehmer (1991a), describes decision making in a complex system as a process of regulating a system. The decision can be viewed as a task of finding one process in the system that can be used to control some other process in the system. The goal of the decision maker is to analyse the state of the system and try to see what processes in the system he or she can use and control to fulfil his or her goal.

Brehmer (1991c) has also characterised decision making in a complex system by the following properties:

It demands a series of decisions.
The decisions are dependent on each other.
The states of the system are changed both autonomously and as a function of the actions that the decision maker makes.
The decision makers have limited time to make their decisions.
Different decision makers often work on different time scales in the system.
They are trying to find a process to control another process.
The decision makers may have inconsistent information.

Co-operation Between Time Scales

Organisations that try to control complex dynamic system are often organised in some form of hierarchy. An important reason for this is to let the decision makers consider different abstraction levels and time scales. The time scales are relative. Often a higher time scale has the authority to command a lower. An organisation containing a commanding and controlling staff and its sub­or­dinates can in it simplest form contain two time scales. The staff works on a higher and slower time scale, while the subordinate units work on a lower and quicker time scale. For this type of organisation it is possible for those with more general and broader information on the higher time scale to define operations for the lower time scale. When people on a higher time scale set goals for the lower time scales, they can do it so that operators on the lower time scale can work co-operatively within their time scale to process something which is appropriate from the higher time scale point of view, but not for those on the lower time scale. An example would be to save a town from catching fire.

Tactical Reasoning

One way to view the decision making in real world tasks for a commander and staff in an emergency organisation is to view the decision making as a process that exists in a tactical reasoning process. Rogalski & Samurçay (1993 a, b) start by describing the work performed in an emergency organisation as a loop including information gathering, prognosis and planning, decision making and order giving, execution and control. See Figure 3. This figure is a good description of the activity and information flow in a three-level hierarchical emergency organisation. The com­mander and staff work at the top level where they are responsible for the tactical decisions. At the bottom level are the tactical units that perform the activities on the target system.

Figure 3. Information flow and task distribution in emergency management.

The commander and staff's work of assessing the situation and co-ordination of actions have been described in the method for tactical reasoning (MTR) by Rogalski & Samurçay (1993 a, b). This method for tactical reasoning was developed by expert emergency officers for a more rational and efficient decision making. The method covers common processes for elaborating and choosing strategies. The basic steps in the MTR are shown in figure 4. The staff collects information from the emergency organisation and the other information resources so that they gain a situation awareness and understanding of the risk status. On this basis they define the task, plan and transmit orders to their subordinates in order to direct and co-ordinate actions between the subordinate units. The method for tactical reasoning defines a framework for task analysis that satisfies three functions: one, it defines the field of necessary and sufficient data to be processed and hence provides a basis for orienting information-gathering and data-processing; two, it constrains operators to define the various courses of evolution of the situation exhaustively and to design possible tactics leading to the target state - it prescribes a search for possible means for reaching the goal; three, it prompts operators to define a set of choice criteria and to evaluate them in terms of the proposed tactics.

Figure 4. Some steps in Samurçay & Rogalski’s MTR.

The method for tactical reasoning should be seen as a rational view of the staff’s work process. In a real situation the decision makers do not follow the process step by step. The staff work on a continuous process where more than one step is processed at the same time. A typical example is the situation assessment step, which proceeds the whole time a staff is working. In the situation assessment process (see figure 5) the decision maker seeks more information to the point where he or she has enough information to create a proper situation awareness. The decision maker then goes to the next step, but if the decision maker gets information that conflicts with the current mental picture of the situation, he or she goes back to the situation assessment step and creates a new mental picture of the situation.

Figure 5. Situation assessment steps in the RPD model.

Concluding Remarks ...

Decision making in emergency management staffs is a difficult and complex task. An important step in the decision process is, as Nobel, Klein, Samurçay & Rogalski say, the situation assessment process, which has been the focus of my work. This process can be viewed as taking three steps: defining the current state, analysing possible evolution and performing a risk analysis. To be able to train situation assessment it is important that the training environment generates a decision environment for the trained staff that enhances the learning processes. This means that to meet training goals the training system should have the same information cues and dynamic behaviour as the real world. The training task for the trained staff should be to learn to collect appropriate information from their subordinate units so to gain good situation awareness.


The following references are used on this www page.

Bidgoli, H. (1989).
Decision Support Systems - Principles & Practice. West Publishing Company,
St. Paul, MN. 1989.

Brehmer, B. (1991a).
Distributed Decision Making: Some Notes on the Literature. In Distributed Decision Making: cognitive models for co-operative work, J. Rasmussen, B. Brehmer & J. Leplat (eds.) John Wiley & Sons, New York. ISBN 0-471-92828-3.

Brehmer, B. (1991c).
Dynamic Decision Making: The Effects of Task Complexity and Feedback Delay. In Distributed Decision Making: cognitive models for co-operative work, J. Rasmussen, B. Brehmer & J. Leplat (eds.) pp. 193–200. John Wiley & Sons, New York. ISBN 0-471-92828-3.

Brehmer, B., & Svenmarck, P. (1995)
Distributed Decision Making In Dynamic Environments: Time Scales and Architectures of Decision Making. In Contributions to Decision Making - I, J.-P. Caverni, M. Bar-Hillel, F.H. Barron, H. Jungermann (eds). Elsevier Science B.V. 1995.

Duncker, K. (1945).
Onproblem-solving. In Psychological Monographs, vol. 58. Dashiell, J.F., (ed).
The American Psychological Association, Inc., Washington, D.C.

Edwards, W. (1954).
The theory of decision making. In Psychological Bulletin, 51, pp. 380-417. 1954.

Hammond, K. R., McClelland, G. H., and Mumpower, J. (1980).
Human Judgement and Decision Making. Hemisphere Publishing Coperation,
New York, NY. 1980.

Hammond, K. R. (1993).
Naturalistic Decision Making From a Bruswikian Viewpoint: It Past, Present Future.
In Decision Making in Action: Models and Methods, G. A. Klein, J. Orasanu, R. Calderwood, and C. E. Zsambok (eds.), ISBN 0-89391-794-X, pp. 205 - 218, 1993.

Klein, G. A. (1993).
A Recognition-Primed Decision (RPD) Model of Rapid Decision making.
In Decision Making in Action: Models and Methods, G. A. Klein, J. Orasanu, R. Calderwood, and C. E. Zsambok (eds.), ISBN 0-89391-794-X, pp. 138 - 147, 1993.

Klein, G. A., Calderwood, R., & Clinton-Cirocco, A. (1986)
Rapid decision making on the fire ground. In proceedings of the Human Factors Society 30th Annual Meeting, 1, pp. 576-580. 1986

Klein, G. A., Thordsen, M. (1989).
Cognitive processes of the team mind. Ch2809-2/89/0000-0046. IEEE. Yellow Springs: Klein Associates. 1989.

Lipshitz, R. (1993).
Converging Themes in the study of Decision Making in Realistic Settings.
In Decision Making in Action: Models and Methods, G. A. Klein, J. Orasanu, R. Calderwood, and C. E. Zsambok (eds.), ISBN 0-89391-794-X, pp. 105 – 109, 1993.

Montgomery, H. (1993).
The search for a dominance structure in decision making: examining the evidence.
In Decision Making in Action: Models and Methods, G. A. Klein, J. Orasanu, R. Calderwood, and C. E. Zsambok (eds.), ISBN 0-89391-794-X, pp. 182 - 201, 1993.

Noble, D. (1986)
Application of a theory of cognition to situation assessment.
Vienna, VA:Engineering Research Associates. 1986.

North, D. W. (1968).
A tutorial introduction to decision theory. In Shafer,G. and Pearl, J., (eds.),
Uncertain reasoning, pp. 68-78. Morgan Kaufmann Publishers.1968.

Orasanu, J., & Connolly, T. (1993).
The Reinvention of Decision Making. In the book, Decision Making in Action:
Models and Methods. G. A. Klein, J. Orasanu, R. Calderwood, and C. E. Zsambok (eds.),
1993, ISBN 0-89391-794-X,

Orasanu, J., & Salas, E. (1993).
Team Decision Making in Complex Environments. In Decision Making in Action: Models and Methods, G. Klein, J. Orasanu, R. Caldewood, C. E. Zambok (Eds.). New Jersey: Ablex. 1993.

Rogalski, J.,& Samurçay, R. (1993a).
A Method for Tactical Reasoning (MTR) in Emergency Managment: Analysis of Individual Acquisition and Collective Implementation. In the book Distributed Decision Making: cognitive models for co-operative work. J. Rasmussen, B. Brehmer & J. Leplat. (eds.), ISBN 0-471-92828-3. John Wiley & Sons, New York. pp. 287-298. 1993.

Rogalski, J.,& Samurçay, R. (1993b).
Analysing communication in complex distributed decision-making. Ergnomics, 36, pp. 1329-1343. 1993.