ATC Work: Behavioural, Cognitive & Sociological Perspectives.

Opening up ATC work: Behavioural, Cognitive and Sociological Perspectives

Liam Bannon & Dan Shapiro

University of Limerick Lancaster University

In Proceedings 21st Western European Association for Aviation Psychology (WEAAP) Conference, Trinity College, Dublin. March 1994.

Introduction

This paper is intended to set the scene for a discussion of how different disciplinary perspectives can contribute towards our understanding of complex human-computer interaction situations, specifically focused on air traffic control (ATC). While the title of the Symposium "Psychological versus Sociological Approaches to ATC Work" might appear unnecessarily combative and divisive, the intent is to open up a direct debate about strengths and weaknesses of different approaches. Here we concentrate on developing the third, and probably least discussed to date - namely the sociological approach. It is important to note that a variety of different sociological approaches could illuminate ATC work, and what we represent here is mainly just one form of contribution, namely that of ethnomethodologically-inspired ethnography. All 3 papers in this session involve members of a group of sociologists originally at Lancaster University who have carried out a study on ATC in the UK (Hughes et al, 1992). Our purpose in this introductory paper is more to open up the discussion, and put it in a larger context of current debates within the general HCI and CSCW community.

The Context: Evolving HCI Research

While the field of HCI is itself perhaps best viewed as an arena in which different disciplinary viewpoints vie for attention, the cognitive science perspective has been the most dominant in recent years. This approach has come up against a number of problems, both in terms of its research agenda and the usability and utility of its empirical results (Bannon & Bødker, 1991). For example, many experimental studies tend to analyse the individual without reference to their community, or their history, performing on a task designed by the experimenter in an unfamiliar environment. Performance is measured relative to a certain "ideal", rational model of problem-solving, and the deviations of subjects from rational action is noted. In an influential book, Lave (1988) provides a strong critique of both the theoretical underpinnings of the mainstream approach and the experimental manipulations often associated with it, arguing that "Cognition observed in everyday practice is distributed -stretched over, not divided among- mind, body, activity and culturally organised settings (which include other actors)". There is a call for changes, of a more or less radical kind, in the conceptual frameworks employed, the kinds of research undertaken, and methods used, in HCI.

The critique of the general HCI field, as exemplified by work appearing in such flagship conferences as the ACM CHI series, needs to be moderated by a number of important factors as we turn towards HCI work in areas such as process-control or aviation, where there has been a somewhat different research tradition. For example, the exemplary work of people like Hollnagel and Woods (1983), Rasmussen (1986) and others in the "cognitive engineering" perspective is of particular importance to our present concerns. Such work takes the features of the task domain as vital to an understanding of human-system interaction. So the nature of the work is factored into their models. Secondly, they usually employ skilled operatives in their studies, in very realistic situations, if not conducting actual on-site studies, so this diffuses further some of the criticisms noted above. However, emphasis has still tended to focus on the individual user's model of the task, the actual behaviour of each user, their errors, etc. This is adequate for certain purposes, yet the uncritical acceptance of this situation as the norm in the field has meant that the everyday needs of people to communicate with others both around and through the computer system have received insufficient attention. Current concerns in HCI, and in the emerging CSCW ( Computer Supported Cooperative Work) field, are on extending the focus of concern from the human-computer dyad to larger groups of people and machines engaged in collaborative tasks.

It is specifically on this issue that there has been interest in alternative concepts and methods. This focus has lead to the need for extensions to existing conceptual and methodological frameworks as well as the search for more appropriate frameworks from other disciplinary areas. While more traditional sociological and anthropological concepts - division of labour, issues of power and control, symbolism, etc. are of importance, there has been particular interest in studies of an ethnomethodological nature. The emphasis in these studies is on the work that members do in order to make their work accountable to themselves and each other, focusing on the "working division of labour" as distinct from viewing the division of labour as an analytical category. Even within the cognitive science tradition, we see an extension of the basic conceptual apparatus in a number of directions, e.g. examining the role of artifacts in human cognition (Norman, 1991), and the way cognition is distributed among people and artifacts -"distributed cognition" (Hutchins, ms). Another alternative conceptual framework that has been applied is the Russian tradition of socio-cultural or activity theory (Kuutti, 1991). While the conceptual apparatus of this framework is difficult to apply directly, its focus on the concept of activities, which are inherently collective, not individual, phenomena, and on mediation of activity via instruments, tools, procedures, methods, etc. make it quite apropos.

The Case of ATC Work

Air Traffic Control (ATC) is a very complex work activity, involving a variety of tasks that must be interleaved, with obvious time constraints, including communication and cooperation not just with cockpit crew, but with other controllers on the ground. Given the safety critical nature of the activity, there has been a strong interest in understanding and either automating or supporting the work of the controllers with computer aids for many years. Simplistic automation strategies will not work (Harper, Hughes & Shapiro, 1991). A more considered, evolving piecemeal strategy for supporting the work of ATC officers with technology is now developing among a number of researchers in the area. There remain however serious questions about human-machine task allocation in the design of the supporting technology systems, for ATC, just as for other areas involving intense monitoring of workflows and processes. Reducing the involvement of controllers in the actual controlling activity through automation can have the effect of making them less competent as controllers when in fact a system malfunction requires them to intervene and assume direct control for a period (Bainbridge, 1983).

The work of air traffic controllers (ATCs) has been the subject of study for many years by human factors (HF) researchers. As the work involves a high degree of expertise, it has been evident for some time that extrapolation of simplified laboratory studies to the world of work is deeply problematic in this area, so there has been a tradition of performing studies in as rich a setting as possible, and using experienced people in the studies. The conceptual frameworks employed in such studies have been varied. On the one hand, in an effort to eschew conceptual confusions, some researchers have adopted an extremely behaviouristic task analysis approach, focusing very much on identifiable tasks, and overt behaviours required of individual ATC's. This was the norm in most ATC studies until 10 years ago. Then, the influence of cognitive psychology (mentioned above) led to a different approach, going beyond simple behavioural studies. In particular there has been an interest in what has been termed "cognitive task analysis" (CTA). This differs from traditional task analysis, which tends to focus on behavioural performance and training objectives, by putting more emphasis on the cognitive processes involved in producing behaviour and on the nature of the learning process. As noted by Redding & Seamster (in press): " CTA focuses on decision making and mental models, learning and skill development, the interrelationships among job concepts and task elements, and group performance as well as individual differences." They claim that most CTA’s identify the following: 1) the key job components, 2) the knowledge and skills required for similar job components, 3) important knowledge and skill differences between novices, (perhaps intermediates) and experts, or between good and bad performers and 4) the conditions which best facilitate learning. While it is generally claimed that cognitive and behavioural methods ought to be complementary, it is not always easy to discuss them in such a manner, as the categorisations and concepts may not mesh neatly.

We can add a further development, or perhaps note a parallel development, where more sociological methods have been applied. There has been an increased interest recently in studies that focus more explicitly on the socially organised character of ATC work, and the effects that new technology might have in supporting or indeed hindering this collaboration. Much of this work has been reported within the area of CSCW - Computer Supported Cooperative Work, and is often identified with an ethnographic approach to studying ATC work, in situ. This is quite distinct from the individualistically-oriented task-analytic studies done in laboratories, as often found in traditional human factors work, and while as we shall see, there are overlaps between this approach and CTA, there are also quite large differences in approach. As expressed by Hughes et al., 1992: "There is no one method of ethnographic analysis. ... The field workers immersed themselves in the work by spending several months observing activities on and around the suites, talking to staff, and discussing with them the researchers’ developing understanding of what controllers do. While attempting to avoid prejudices and to allow the work situation to ‘speak for itself’ as much as possible, researchers cannot claim to address it innocent of any theoretical orientation; and their results would be much impoverished if they did. The purpose of an ethnographic approach is not so much to show that work is socially organised (which is rather easy) but to show how it is socially organised."

The ethnographic studies note, for example, how much individual work done by ATC officers is also done in a way so as to make available to other ATC officers information as to the state and conduct of the task, without requiring overt messaging between officers. In any attempt to "improve" the tools of the ATC officers, they point out how great care must be exercised so that the current, often non-obvious, fluid use of artifacts and signs to communicate the state of affairs in the airspace to all in the control room are not disturbed by new technology. The technology may, while making specific tasks easier, have the undesirable side-effect of occluding some vital information about the state of affairs to others. An example that has been studied in this regard is how properties of the physical flight control strips are used to make available information about the state of affairs to others in simple yet undemanding ways that would be difficult to replicate in simple electronic replications of such strips.

Comparing Approaches

While any attempt at direct comparison of approaches is bound to be difficult, perhaps in some sense impossible due to the incommensurability of different perspectives, from the point of view of systems design it would appear reckless if one did not at least try to utilise all the concepts and results available, from whatever disciplinary background, or research method, so as to develop appropriate technology to support ATC's. While such comparative studies are rare, there has been an initial attempt to investigate the complementarities of certain approaches in a recent paper by one of the authors (Shapiro, 1993) which we will briefly refer to here as a pointer to the kinds of studies that may be fruitful. The paper contrasts two studies undertaken from different disciplinary perspectives but with exactly the same subject matter: the use and possible electronic replacement of the paper flight progress strip in en-route air traffic control. One involves an ethnographic method and an ethnomethodological theoretical perspective (Hughes et al., 1992). The other involves an experimental behavioural study undertaken in the US by O.U. Vortac et al. (1992).

Shapiro notes the obvious problems with an exercise of this kind which takes just two specific studies as its basis. We should also note that several of the ‘standard’ criticisms of task analysis either do not apply or only partially apply to the Vortac et al. study. While the study did take place using a training simulator ‘experiment’ rather than a real-world setting, that is not intrinsic to the method and an equivalent real-world study is perfectly feasible. While the study used ATC instructors rather than real current controllers, that too could be corrected provided a sufficiently unobtrusive means of observation could be devised, which again should not be insuperable. It looks only at individual controllers, in communication through the simulator with ‘ghost’ pilots and a ‘ghost’ controller, but the researchers recognise this as a problem and describe their intention to study a pair of controllers co-operating at a suite. Commenting on the study, Shapiro suggests that the approach needs a much more purposive and directed means of initially choosing and then refining what are the pertinent units of activity to observe. He complains that while it was possible to conclude that strip use was important; it was not able to determine the implications for the possible automation of this function, or how it should be done. Finally, Shapiro notes: "the sociality of the work, which is so fundamental to the ethnographic account, is entirely missing from the task analysis, and it is not easy to see how it could be recovered". At the same time, he is aware of limitations of the ethnographic-ethnomethodological account concerning the degree of independence between data and interpretation.

What would be of interest now is if we could have a comparison exercise between an ethnographic account and a cognitive task analysis of similar activities. CTA methods include such methods as observation and interviewing, protocol analysis, psychological scaling, cognitive and performance modelling, and error analysis of different forms. As noted above an ethnographic account would also encompass some of these methods: observing and interviewing in particular. At least at this level, the kinds of material produced might be comparable. While the ethnographic account does not develop models of "internal cognitive processes", it does try to explain observed behaviour by recourse to concepts which describe the nature of the work from the point of view of those carrying it out. While one might expect vast differences here, we can detect, at least tentatively, some interesting commonalities in terms of what the different views call attention to in the work of ATC.

For example, in an overview of CTA, Redding & Seamster claim that 2 key findings emerging from such studies are that the 2 key tasks for ATC's are: to maintain Situation Awareness & to develop and revise the Sector Control Plan. Let us note one of the key features of ATC work as identified by Hughes et al(1992) from their ethnographic work - "getting the picture". To quote them directly on this: " Controllers ‘build a picture’ of the state of the sector by relating traffic within it to a ‘schematic’ of that sector’s characteristics. ....‘Getting the picture’ therefore means working out what proportion of flights fit the scheme and what proportion are exceptions; assessing these for the kinds of control actions that will be required, and planning for those actions; and issuing the instructions to create an appropriate current order, and make the traffic flow." At least as a first pass, there seems to be some agreement on what the key aspects of ATC work are. Exactly what is involved in "getting the picture" would seem to be an interesting exercise to explore and develop within different approaches. Even more so, in thinking about system design, how could one support ATC's in this demanding task? Whether one refers to something going on "in the head" of the ATC, in terms of mental models or such like, or whether one prefers a phenomenological account, both kinds of reports would seem to indicate the need for support tools to provide some form of Gestalt to help controllers follow the air traffic in a sector.

In conclusion, the purpose of this paper was to open up a discussion on the variety of ways that one might study ATC work, and note some features of each approach. While they are not directly comparable, certainly there should be scope for collaboration across the disciplinary divides.

References

Bainbridge, L. (1983) Ironies of Automation. In G. Johannsen & J. Rijnsdorp

(Eds.) Proceedings of IFAC/IFIP/IFORS/IEA Conference on Analysis,

Design & Evaluation of Man-Machine Systems, pp 151-8, Baden Baden

Pergamon.

Bannon, L. & Bødker, S. (1991) Beyond the Interface: Encountering Artifacts

in Use. Book Chapter in J.M. Carroll (Ed.) (1991) Designing Interaction:

Psychology at the Human-Computer Interface, pp.227-253. (New York:

Cambridge University Press)

Harper, R.R., Hughes, J.A., & Shapiro, D.Z. (1991) Harmonious Working and

CSCW: Computer technology and air traffic control. In J. Bowers & S.

Benford (Eds.) Studies in Computer Supported Cooperative Work.