The Special Case of Problem-Solving

It is customary to differentiate between decision-making and problem-solving as somehow separate processes. A decision can be seen as a choice between two or more known alternatives. With a problem, it is not necessarily clear at the outset what the current situation entails and, therefore, what alternative solutions are available and which one will have the greatest efficacy. Decision-making is often fast and intuitive whereas problem-solving is often slow and laborious (Kahneman, 2011). Problems are what Orasanu and Martin called ‘creative’ or ‘analytic’ decision types, often require the creation of novel solutions. On closer examination, however, what we call problem-solving is actually a process of making decisions about decisions.

Problem-solving is usually triggered by either an impediment to progress or a realisation that there is no obvious course of action available. In the event of a non-normal or emergency situation like the B-777 on approach to Heathrow, the fact that the crew were presented with a ‘problem’ was apparent. In Chapter 3,1 suggested that the margin was often the zone in which discrepant signals emerged, and in the discussion of SA, we saw that the final stage in Endsley’s model was ‘projecting ahead’. Given Chater’s caution that the future is unknowable, a key element of task management is tracking the current status of the task and validating the future goal state constraint set. This prompts two key questions: ‘will the conditions that prevail now remain applicable until the next goal state?’ and ‘are there any signals in the current situation that suggest either a significant change to the prevailing conditions before the next goal state or a change to the goal state constraints?’. If the answer to either question is positive then there is a need to consider a change to the current, active plan or to select a new course of action. From a systems perspective, we once again come across the issue of scale effect. Quite often, as conditions deteriorate (at least, from the perspective of the outside observer), pilots continue to apply skill-based and rule-based behaviour to sustain performance within bounds: they are engaged in what I call ‘trajectory trimming’ in order to remain safe and legal. Their actions are directed at the immediate situation. Unfortunately, the system will be operating at a larger scale and maybe changing in more fundamental ways. By the time they recognise the severity of their situation, their range of options has narrowed significantly. At some point along a notional trajectory, the boundary has shifted, and the change in conditions has reduced both the margin and the buffering capacity. Because of the inadequate tracking of the current status of the operation, actors often fail to consider the time and space required for a manoeuvre or goal state transition. Furthermore, where actions are deferred, actors rarely anticipate that any further change in circumstance might impact the ability to perform the delayed action. In part, this is reflected in task forgetting.

The traditional view of problem-solving is that it is a normative process in which the decision-maker has identifiable, stable expectations. The choice is then between a set of known alternatives, each of which has a specific value, efficacy or perceived benefit. Prescriptive models, derived from such an approach, are common in aviation, and Table 5.4 presents some popular examples. Although referred to as ‘decision-making models’, the first five rows in the table actually describe what we might call the steps in a problem-solving process. The last three rows describe the decision point and subsequent control activity. The ‘problem-solving’ phase comprises a set of decisions about recognising a need for additional activity, about how and where to search for information and about what future actions have a tolerable probability of achieving the desired goal. Since the early 1990s, attempts have been made to offer descriptive models of naturalistic decision-making (NDM) that better capture real-world performance. It is best to think of NDM as a cluster of views about decision-making rather than a fully developed theory.

Table 5.5 presents a model of problem-solving drawn from work using business simulations (Dorner & Schaub, 1994). The table lists the steps and the potential error types. The model offers an elaboration of the first five rows in Table 5.4 and, importantly, some of the key reasons why problem-solving, especially under stress, can fail. The authors make the point that we have limited capacity for thinking and, as we have seen, are prone to forgetting. They also stress that humans tend to guard feelings of competence and efficacy: we want to feel that what we are doing is serving a purpose. Kahneman has demonstrated that we tend to be overconfident: we think we know more

TABLE 5.4

'Decision-Making' Models

TABLE 5.5

Stages in Problem-Solving

than we do. Furthermore, our natural loss aversion makes us more likely to take risks in order to avoid a potential loss, and our inability to recognise a ‘sunk cost’ means that we are likely to continue to throw good money after bad in order to avoid feelings of regret.

One solution proposed for creative situations is critical thinking or narrative creation. In situations like this, actors attempt to create dynamic descriptions of events - a narrative or story - that they can then use to generate future actions. As long as the outcomes observed can be explained by the narrative, then the model is deemed adequate to support future action. If subsequent events cannot be explained by the narrative then the story must be reviewed and a new narrative is created.