In this section, the aims of the paper are set out and the interested parties identified. Air traffic management is introduced and the utility of ATM-like micro-worlds assessed. Different conceptions of design problems are reviewed. Criteria for a design problem framework for Cognitive Ergonomics research are identified.

Aeronautics is a broad domain and includes aviation studies; air traffic management; aeronautical medicine; cockpit management and so forth. This paper is concerned with air traffic management (ATM), as planning and control. ATM may be the object of science, for example Cognitive Psychology (particularly Cognitive Modelling), and so lead us to an understanding of ATM cognitive phenomena. Alternatively, ATM may also be the object of Cognitive Ergonomics and so lead to the design of ATM systems. The approach taken here can be generally considered to follow that of Cognitive Ergonomics, as exemplified by: Hollnagel and Woods (1983); Wickens (1984); Woods and Roth (1988); Rasmussen and Vicente (1990); Vicente (1990); Dowell and Long (1998) and Hoc and Amalberti (1999). However, for the acquisition and validation of Cognitive Ergonomics design knowledge, there is a requirement from researchers (and indirectly from ATM system developers) for its design problems to be conceptualised and operationalised as in other areas of engineering (Helander, 1994). Without such conceptualisation and such operationalisation, design knowledge supporting the solution of design problems cannot be acquired or further validated (that is, tested and generalised) by researchers. This paper describes a set of models, constituting a framework for conceptualising ATM design problems and an operationalisation of those problems as data. A design problem is informally expressed as: to design users interacting with computers to perform work effectively. The framework models, that is, design problem; worksystem; domain; performance; and operator domain representation are intended to support researchers in the acquisition and validation of design knowledge (Boudes and Cellier, 1998; Grau, 2000; Hoc and Lemoine, 1998; Hörmann, 2000; Loiselet and Amalberti, 2000; Vanderhaegen, 1999). In turn, this knowledge is intended to be of use to developers, concerned with the design of operational ATM (Reuzeau, 2001; Shorrock, Kirwan, Kennedy, and MacKendrick, 2001). First, we consider the different “worlds” of ATM.

Operational ATM manages air traffic, for example, Manchester Ringway Control Centre in the UK. The Centre manages a terminal manœuvring area with 9 beacons; more than 2 airways; 1 stack; and 2 exits. Its traffic can be characterised as: departing; arriving; overflying; “low and slow”; high-level bunching and so forth. The management involves track, and vertical separation rules (Dowell, 1993).

In this section, operational problems are considered as a possible candidate for specifying design problems, as required by researchers to acquire and validate Cognitive Ergonomics design knowledge. However, such problems are rejected in favour of “design problems for research”, as described here, problems which can be initially conceptualised and operationalised in ATM-like micro-worlds.

Operational ATM problems come in at least three different, but not unrelated, forms—instances, statistical aggregations and user requi- rements. All are potential candidates for expressing ATM design problems.

Having rejected operational problems, as expressions of design problems for researchers, we now turn to research itself. The research literature on ATM suggests a range of possible candidates to express design problems. Some candidates appear to be ATM phenomena, for example, desynchronisation; abstraction; anticipation; time-management; pre-planning; planning; re-planning and so forth (Hoc, in press), and anticipation; temporal structure, etc. (Carreras, Valax, and Cellier, 1999). Other candidates reference ATM relations, for example, operator activity-traffic event relations, as they concern “near misses” (Hoc and Lemoine, 1998). Other candidates seem to reference different aspects of ATM performance, for example, “problems” (Vanderhaegen, 1999); “difficulties” (Boudes and Cellier, 1998); “practical benefits” (Hoc and Amalberti, 1999); and “effectiveness” as: acceptability; utility; and usability (Reuzeau, 2001). Other candidates reference a range of other issues, for example, “challenges” to include: errors and incidents; workload; delays; complexity; and traffic levels (Shorrock et al., 2001); “interference”, as it relates to: complexity; adaptability; workload variation; and temporal constraints (Loiselet and Amalberti, 2000); failures of situational awareness, involving: plan particularisation; external focusing; and unplanned situational awareness objects (Grau, 2000); and “system flaws” and “latent active failures” (Hörmann, 2001), etc.

The criteria are the same as the three criteria for design problems, applied earlier, and the general constituents of research problems, identified also earlier. First, the expression of the design problem, to be complete, must include: ATM worksystem (operator and devices); work; and performance. Second, to be coherent, the design problem must specify compatible relations between the: ATM worksystem (operator and devices); work; and performance. Last, to be fit-for-purpose, the expression of the design problem must be addressable by research to acquire design (and evaluation) knowledge to accommodate design practices of diagnosis of the design problem and prescription of the design solution. Before presenting a framework for design problems for research, the ATM-like micro-world, used to conceptualise and operationalise the design problems, is described.

This section describes an ATM-like micro-world, termed “reconstructed” ATM, rATM. Relations between ATM and rATM are described, along with rATM’s interface, traffic model, and goals. rATM is not an arbitrary micro-world, but one which has systematic relations with ATM and so the potential of being scaled up to operational ATM.

Manchester Ringway Control Centre in the UK was referenced earlier to illustrate the differences between operational and simulated ATM. rATM, an ATM-like microworld, simulating Manchester Ringway, is used here to conceptualise and operationalise design problems for research. Manchester Ringway manages air traffic in a terminal manœuvring area (TMA). The relations with other types of traffic management are shown in Figure 1. It can be seen that Manchester Ringway indeed constitutes a type of operational ATM with similarities and differences with other types of ATM. A comparison of the main features of Manchester Ringway ATM and rATM are shown in Figure 2 (following Dowell, 1993). It can be seen that the two share a number of features, for example, concerning sector, traffic, devices, etc.—hence the claim that rATM is ATM-like. rATM, however, also lacks important features of Manchester Ringway, for example, concerning operators and their training—hence the claim that rATM is a micro—, rather than a macro- (or even a micro-) world. Both claims are important. Taken together, the claims indicate rATM not to be an arbitrary micro-world, but one which has systematic relations with ATM and so the potential of being scaled up.

rATM features, shown in Figure 2, have been implemented in an interface by Dowell (1993 and 1998). The interface supports point-and-click interactions for traffic interrogations and interventions, as associated with simulated radar and flight progress strips. The flight strips, encapsulate the scenario for interaction with the rATM interface. The flight plan also reflects features of the rATM simulation (see Figure 2).

This section describes the design problem framework, intended to conceptualise and operationalise design problems for researchers. The framework is constituted of rATM: design problem; worksystem; domain; performance; and operator domain representation. It is concluded that the framework meets the general ATM requirements identified earlier and in a manner consistent with the design problem framework of Dowell and Long (1989 and 1998).

According to Dowell and Long (1989 and 1998), a design problem is expressed informally as: to design users interacting with computers to perform effective work. In the rATM expression, users become operators and computers become devices. More precisely, then, the rATM design problem becomes:

The rATM worksystem comprises the operator and the devices. Following Dowell and Long (1989), the two are modelled as structures, supporting behaviours—in both cases physical and mental (abstract). A model of the rATM worksystem is shown in the upper panel of Figure 3, following Timmer and Long (1997) and Timmer (1999). rATM operator physical behaviours (“pull down” (current speed menu—see earlier)) are supported by physical structures (“hand”). rATM devices’ physical behaviours ((display) “call sign”)) are supported by physical structures (“radar”). Operator physical behaviours interact with device physical behaviours (“write new speed”). Such operator physical behaviours are supported by mental behaviours, in turn supported by process and storage structures (“encode” (“transducer”); “categorise” (“working memory”); “form goal” (“goal store”); “execution” (“transducer”)).

The rATM domain model represents the domain of application of the rATM worksystem, that is, the work performed by the latter. Following Dowell and Long (1989), the domain is modelled as objects consisting of attributes having values. Physical objects embody abstract objects. The worksystem transforms the object-attribute-values from some actual state to some desired state. A model of the rATM domain, following Dowell (1993), is shown in the centre panel of Figure 3. Physical objects (Beacon-Alpha; Aircraft-WAL) embody respectively abstract objects (air- space; aircraft). Together, they support expression of the PASHT attributes (of Aircraft Position; Altitude; Speed; Heading; and Time of inter- vention). In turn, the latter support the expression of safety and expedition as attributes of air traffic vector objects.

According to Dowell and Long (1989), performance is expressed as two factors— task quality and worksystem costs. Task quality expresses how well a worksystem performs its tasks. Worksystem costs express the costs to the worksystem, that is, the workload of performing tasks that well. Costs are both structural and behavioural and are incurred by both the user and computer. The actual and desired performance, referenced in the expression of the design problem, thus needs to include both task quality and user and computer costs. Space limitations preclude the inclusion of both here. This paper only addresses the task quality factor of performance. The general argumentation, relating rATM to the expression of design problems for research, however, remains unaffected. Readers interested in the treatment of worksystem costs, in the expression of design problems, are referred to Dowell (1998).

rATM operator domain representations are not referenced explicitly by the design problem as such. However, they need to be included in the design problem framework in order to link the domain to the worksystem and so to performance. The inclusion is also motivated by the practice of diagnosis, which attempts to associate undesired performance with aspects of the worksystem structures and behaviours. Such associations provide support for the practice of prescription, that is, the specification of a worksystem, whose actual performance equals some desired performance. rATM operator representations of the domain, which are not consistent with the actual state of the domain, are prime candidates for diagnostic associations and so support for prescription.

The description of the design problem framework is now complete. The framework supports the conceptualisation of the rATM design problem for research and its operationalisation as data. It is concluded that the framework meets the general ATM requirements, identified earlier.

In this section, the design problem framework is applied to rATM simulation data to diagnose a candidate rATM design problem for research. The application is intended to illustrate the framework’s fitness-for-purpose for design in the form of diagnosis. First, two instance design problems are diagnosed, then a class design problem. It is concluded that the framework shows promise for being fit-for-purpose of diagnosing rATM design problems for research.

The framework can be applied to diagnose design problems, whenever actual performance deviates from desired performance by an unacceptable amount (see design problem expression Section III . 1 earlier). In rATM, deviations are judged with respect to the worksystem goals (see Section II . 2). Criteria for the unacceptability of deviations also relate to worksystem goals, but, in addition, require criteria, associated with the research, intending to acquire design knowledge to solve the design problem. Such criteria might include: representativeness, criticality, tractability, etc.—all as applied to the design problem. Such criteria, however, are not part of our concern here, other than that they can be applied to the diagnosis, supported by the framework. The latter is self-evidently the case.

In addition to diagnosing the behavioural associations of instances of design problems, classes of design problem may also be diagnosed. To this end, all recorded instances of undesired performance, associated with a particular performance criterion, are diagnosed, and a general class (that addresses the behavioural associations of each design problem instance) is constructed. Following data collection with rATM, for example, four instances of breaches of safety were recorded. From analysis of each instance, a general class can be constructed. In the first instance, safety is breached following: identification of two aircraft approaching one another at a common altitude; failure to plan an intervention in detail; and a failure to monitor the reducing separation between aircraft. In the second instance, safety is breached following: failure to identify an unoccupied flight level; allocating an aircraft an occupied flight level; and failing to establish the approaching breach of safety (situational awareness). In the third instance, safety is breached following: a failure to anticipate the future positions (crossing) of two aircraft; recognition of safety conflict and plan formation, but no follow-up monitoring; decay of plan and memory for approaching conflict. In the fourth instance, safety is breached following: failure to establish occupied flight levels in the vicinity of an aircraft near its exit beacon; execution of an unplanned intervention to allocate an aircraft its exit beacon; presence of another aircraft (insufficient horizontal separation) at an altitude passed-through by the aircraft reaching its exit altitude. This class design problem can be more formally expressed. [4]

The application of the design problem framework to diagnosis has illustrated its potential fitness-for-purpose for design in the form of diagnosis. Diagnosed design problems, and in particular class design problems, are constructed by means of the framework, and so become available to research, intending to acquire design knowledge for their solution. rATM and ATM design problems are related by means of rATM/ATM relations (see Figure 2).

In the introduction, it is argued that Cognitive Ergonomics research practices of acquiring and validating design knowledge require the specification of design problems, just as Cognitive Psychology research practices require the specification of phenomena to be understood. Current candidate specifications of design problems for researchers in ATM, derived from both operational ATM and research (simulation) ATM, are rejected on the grounds of lack of completeness, coherence and fitness-for-purpose. A framework for design problem diagnosis is proposed that is claimed to meet these criteria. The framework is conceptualised and operationalised in the ATM-like microworld, termed rATM. The framework comprises the rATM: design problem expression; worksystem, constituted of operator and devices; domain; performance; and operator domain representations. The framework is applied to rATM simulation data, describing the management of air traffic, to diagnose instance and class design problems. The framework, and so the design problem diagnosis, it is argued, are complete, coherent and fit-for-design purpose, here diagnosis. The resulting design problems are available for research to acquire and validate generic design knowledge for their solution. ATM research currently reports little by way of the validation of such knowledge. It is difficult to see how, without a better specification of design problems, such validation could occur and indeed be successful. As a result, competing theoretical knowledge, concerning ATM, is not testable or generalisable. Systematic progress in the incrementation of the knowledge to solve design problems is, thus, absent (Newman, 1994). The present framework offers a way forward for the better specification of design problems and so better progress in ATM research. The design problem framework has, so far, only been applied to the rATM micro-world. Its application to midcro-, macro-, and operational ATM is obviously required. However, its successful application here to rATM and the well-specified relations between rATM and ATM, suggests the framework has the potential to be scaled up in this way. It is concluded that, because they are less complex, micro-worlds may be a better initial development environment for specifying design problems than either operational or macro-worlds.