Decision-Making and Goal State Modification
Because of the variability in the operating environment, task modification may also require the selection of a new goal state. Orasanu and Martin (1998) propose a framework of decision types characterised by the nature of the information available and the responses available. These are summarised in Table 5.2. The simplest course of action is a ‘codified’ strategy where an event triggers a rule-based response. As a decision type, a rule-driven situation has a single response, and the go/no go decision in aviation is a classic example of this. Successful performance is related to the experience of the decision-maker, and thus the richness of the response sets available, and to the quality of the information available. Typical failures involve not being able to retrieve the correct response from memory.
The second strategy, associative, is a situation in which multiple options are available and involve trade-offs. Various suggestions have been offered for how humans choose between the alternatives. The pairwise comparison suggests that we set up alternatives and delete one that fails to meet the required or certain criteria, working through the list until one option remains. The ‘first past the post’ strategy suggests that we implement a plan and stick with it until it no longer works. The ‘dominant search’ model proposes that unsuitable options are discarded early, and the most obvious solution applied. The solution will remain in use until either the task is achieved or additional information renders the solution no longer suitable. Again, experience and the availability of information are significant variables in choice decisions as well as the effectiveness of simulation based on stored mental models. The final strategy is the analytic or creative approach and is characteristic of ill-defined situations. This is the domain of problem-solving and will be discussed later in this chapter.
In dynamic contexts, a decision generates a future action. That action, in turn, may require a further decision. Rule-based problems tend to function in a feed-forward manner: the decision is made and action is taken. The execution of intervention then brings about a state change. However, in situations requiring more creative solutions, action will feedback to the decision-maker and may well change the understanding of the situation as well as the freedom of action for any responses. Reason (1990) identifies two intervention conditions: reactive and multiple dynamics. In the former, the action of the decision-maker influences the state of the task in a lock-step fashion. In the latter, the problem space is influenced by the actions of the decision-maker but also, independently, by external influences.
The Course of Action Strategies
We have already seen that sense-making is fundamental to successful intervention in the world, and studies of decision-making have added to our understanding of the process. Quite often, when presented with a problem the tendency is to consider ‘have I seen this before and what did I do last time’. If our analysis of the situation is wrong then the solution we apply will not work. One study (Fischer, Orasanu & Montalvo, 1993) found that more effective pilot decision-makers spent longer clarifying the situation, worked through checklists to completion and took time to prevent misdiagnosis. Fischer, Orasanu and Wich (1994) found that the situation structure was not constant across team roles. They found that captains tended to view a situation from a severity and flight safety perspective whereas FOs considered situations in terms of the response required - what will they be required to do. Both groups did consider the time available as a factor. The authors also found that better decision-makers balanced ongoing tasks, avoided cognitive lock-up (fixation) and also retained options for as long as possible.
Goals can also differ in terms of their structure, which can affect situation comprehension. Goals can be either open or closed. A closed goal would be easily defined: land on the correct runway, for example. An open goal might be something like ‘how to efficiently and safely dissipate excess energy in the descent in order to achieve a stable approach while negotiating storm cells’. The structure of the goal might be visible or hidden. So, the relationship between the inputs and outputs is clear in a situation where the structure is visible but is less clear where systems are opaque. Time pressure can be high or low, and the information available may be dense - and therefore difficult to process - or sparse - leading to ambiguity. These factors, especially time pressure and structure, will affect the ability to prioritise, which is simply the act of selecting the next best action sequence. Breaking goals down into meaningful sub-goals and establishing a purpose for each goal has been found to be an effective behaviour (Dorner and Schaub), but the oversimplification of situations is, actually, a recognised bias. Simplification is a counter to complexity but can result in a flawed analysis. Pilots often use checklists to frame situations but this runs the risk of falling victim to the ‘availability’ heuristic. Just because a checklist appears to match the situation does not mean that it is the correct checklist.
Two further issues that relate to situation assessment are bounded rationality and the structure of the actor’s mental model. Bounded rationality relates to the fact that, unlike supposed rational decision-makers, ordinary people tend to work within a subset of available information. They do not interact with an absolute model of the world but, instead, work with a skewed, biased, approximate version. Furthermore, because of flaws in training or through aberrant schemata modification as a result of misinterpreted experience, individuals’ mental models may be incomplete or inaccurate. Therefore, their ability to run simulations of events will be jeopardised.
Work unfolds by individuals enacting sequences of behaviour directed at achieving a goal state, each sequence triggered by prior action or by some change in the environment. In aviation, work is often conducted in conditions of ambiguity and with high risks involved. Often, crews have to reconcile conflicting goals, and their actions can have unintended consequences. Selecting between the courses of action often involves incurring known costs in order to mitigate unknown risks. Unlike normative models, where there is usually a single ‘best’ solution, this is not the case
Metacognition and Task Management
Source: After Rasmussen and Svedung.
in aviation: a choice is ‘good enough’ in that it works under the circumstances. Our behaviour can be classed as satisficing in that it is acceptable although probably not always optimal. A course of action, with hindsight, can often be criticised, but at the time, decisions are usually judged acceptable if ‘nothing went wrong’. These ‘decision-action’ sequences occur in real time as the flight progresses.
We can now see that effective performance comprises both action and control of that action. Metacognition is the term we use to describe this controlling function. Table 5.3 elaborates the skill/rule/knowledge framework, discussed in Chapter 1, in terms of metacognitive functions.
To be successful, the work needs to be monitored. Monitoring requires us to be able to visualise the task being controlled and, thus, is a function of the completeness of the mental model held by the operator. Monitoring requires the current status of the aircraft to be established with reference to information distributed around the aircraft and shared between the aircraft and outside agencies. We will look at monitoring in more detail in Chapter 7.