D ECISION -M AKING 4
4.3 Rational or Analytical Decision-Making
The normative approach has been based on the model of the rational man who receives all necessary information to make decisions, knows well the alternative options, and has adequate time to trade-off options and find the optimal one. Classical decision-making was based on the utility principle, combining Bayesian probability theory with multi-attribute utility theory. In general, classical decision-making follows five stages:
1. Definition of problem parameters 2. Collection of relevant information 3. Identification of available options
4. Assessment of all options based on predetermined criteria 5. Selection and implementation of the optimum option
Classical decision-making has a solid mathematical foundation and it has been published in two variations. The first variation considers
Ambiguous Well structured
Problem structure
Intuitive Cognitive modes Analytical Naturalistic
decisions
Quasi rational decisions (heuristics,
“elimination by aspects”)
Normative rational decisions
Low High
Time availability, auditability of
decisions
Figure 4.1 Matching decision models to different situations in a cognitive continuum.
DECISION-MAKING 111
only discrete options where there is very little uncertainty in the selec-tion process. In this case, the overall value of each available opselec-tion depends on the magnitude of each attribute and the utility of each attribute.
This model uses the following mathematical formula:
Utility Equation
(4.1) Where:
• U(P) : overall utility of option P
• m(i) : magnitude of ith attribute
• u(i) : utility of the ith attribute
• n : total number of attributes
It is obvious that the optimum option is the one that maximizes the value of function U(P). This model of decision-making is appropriate for static problems where uncertainty can be reduced to zero and the selection time frame is very long.
The second variation is similar to the first one but incorporates uncertainty in terms of probabilities of outcomes. As a result, the concept of overall utility is replaced by the overall expected value. In mathematical terms, the expected value of each outcome depends on the probability of the ith outcome occurring and the value of the ith outcome:
Expected Utility Equation
(4.2)
Where:
• E(P) : overall expected value of option P
• v(i) : value of the ith outcome
• p(i) : probability of the ith outcome
• n : number of possible outcomes for option P
It is apparent that the optimum option is the one that maximizes the function E(P).
∑
=
=
( ) ( ) ( )
1
U P m i u i
i n
( ) ( ) ( )
1
E P v i p i
i
∑
n=
=
112 COGNITIVE ENGINEERING AND SAFETY ORGANIZATION
The following simplified example illustrates the application of classical decision-making (Malakis 2009). Consider, for example, an approach control unit that faces a staffing problem for a certain shift. For a time period it is uncertain whether the expected traffic will be kept at high or medium levels (e.g., public events may attract more unscheduled traffic to the airport than events expected by the approach unit). Hence, the unit manager faces the dilemma whether the shift will be staffed with four or three ATCOs. By calculating the probabilities of high and medium traffic conditions, a decision tree can be constructed as in Figure 4.2.
According to the decision tree, the two options are compared according to their expected value scores as shown in Table 4.1.
From the above example, it appears that the optimal choice is the work shift with three controllers that achieves the higher expected value score.
Classical decision-making models can be used in cases where uncertainty can easily be assessed in an approximate manner and
Shift staffing
Heavy traffic Medium traffic
Four controllers Three controllers Four controllers Three controllers
p = 0.3 p = 0.7
v = 9 v = 4 v = 5 v = 8
Figure 4.2 A decision tree for comparing teams of three or four controllers.
Table 4.1 Expected Values for Two Shift Options A SHIFT WITH FOUR
CONTROLLERS
A SHIFT WITH THREE CONTROLLERS Options Probability
of outcome
Value Expected value Value Expected value
p v p(i)v(i) v p(i)v(i)
Heavy traffic 0.3 9 2.7 4 1.2
Medium traffic 0.7 5 3.5 8 5.6
Overall expected value
E(P1) = 6.2 E(P2) = 6.8
DECISION-MAKING 113
there is ample time for selecting and comparing options. In the avi-ation domain, a widely used classical decision model has been the DECIDE model (Robson 2008) which involves six stages for pilots to make good decisions in a logical manner, that is,
D : Detect that the action is necessary E : Estimate the significance of the action C : Choose a desirable outcome
I : Identify actions needed in order to achieve the chosen option
D : Do the necessary action to make a change E : Evaluate the effects of the action
The DECIDE model has been extensively used in analyzing crew decisions in the aeronautical domain. It is a good example of the nor-mative approach, which relies on the following assumptions:
• A clear definition of the problem exists that requires a good decision; there are no circumstances that may change the nature of the problem that requires setting new goals
• Decision-making is seen as trading off options while little attention is paid to situation awareness
• All relevant information is timely available, options are known, and time is sufficient to do the work; information that is unreliable or delayed may create problems in the estimation of the optimal solution and so does the lack of adequate time
• The outcome should be an optimal decision and not a good enough or viable decision
• A formal method (e.g., MAUT) should be applied to find the optimal solution; other informal methods are not acceptable (for example, heuristics)
For decades, classical decision-making has been fully accepted by the majority of researchers and practitioners. The rational approach to decision-making suggested that problems could be formally rep-resented in a mathematical form, allowing optimization of resources.
As a result, failures in problem-solving were attributed to the prac-titioners rather than the adopted method of work. In the period 1965–1985, decision researchers produced evidence that humans do not usually follow rational models of decision-making. New studies
114 COGNITIVE ENGINEERING AND SAFETY ORGANIZATION
have built on the concept of bounded rationality (Simon 1957), that is, rationality limited by the tractability of the decision problem, the cog-nitive limitations of the human mind, and the time available to make the decision. In this view, decision makers act as satisfiers, seeking a satisfactory solution rather than an optimal one.
Simon’s findings triggered new studies that looked into rational and other informal models of how experienced people make decisions.
Kahneman et al. (1982) presented research indicating that humans use a range of heuristics that involve less cognitive effort than rational decisions, although the final decision may not be optimal. Among the well-known heuristics have been the availability heuristic (i.e., a reli-ance on a good solution that is readily available and worked well in the past) and the confirmation bias (i.e., a selective attention to informa-tion that confirms a preferred opinforma-tion). Heuristics are powerful tools in making decisions under time pressure because they do not entail the same cognitive effort that rational decisions do. Another quasi-rational model of decision-making proposed by the same researchers was elimination by aspects. According to this model, people often do not have time to consider and weigh all attributes of different options.
In this situation, they would start by establishing a minimal crite-rion and eliminating the options that fail to satisfy it. Subsequently, another criterion can be selected to eliminate more options until a stage is reached where the last option satisfies all the criteria that remain. However, as options get eliminated in a serial fashion, people may miss an option that has a low score in the first few criteria but compensates with a highest aggregated score for all criteria.