The future influence of new technologies such as Machine Learning and Artificial Intelligence will pose new challenges to the working environment in Air Traffic Control.

Under the umbrella of IFATCA, a group of motivated people has started to discuss and do research on these challenges. This group has produced several conferences papers and has finalised a draft guidance material. The aim being that this guidance material will serve the Federation in the future to assist in educating its membership and contribute to the regulation, certification, and ongoing research initiatives. This paper should be a major contribution to the discussion on the future of technology.

1.1. A group of dedicated professionals has started to tackle this important subject since over a year.

1.2. Composed of Ms Nora Berzina (MUAC ATCO, EGATS), Dr. Anthony Smoker (EASA representative IFATCA, GATCO UK), Dr. Stathis Malakis (SESAR representative ATCO, Greece), Mr. Andrea Potti (EASA representative IFATCA, Italy), Mr. Tom Laursen (former EVP Europe, Denmark), have meet 30 times and were joined by Dr. Marcello Scala (Italy) and Sergio Velotto (Italy) at the later stage of the work. Mr. Marc Baumgartner (SESAR/EASA Coordinator, Switzerland) has acted as coordinator of the group.

1.3. The aim of the group is to establish guidance material to be used in the future by IFATCA, the air traffic controllers and to engage with the outside world.

2.1. Setting the Scene

Today’s Air Traffic Management (ATM) system is heavily reliant on the interdependence between designers, technology, and the human operator. Although there have been several periods of time where the demise of the human operator has been predicted, it hasn’t materialized in significant changes to the designer, technology, and human operator equation. Especially in Europe, recent discussions, have created a push for a change of the role of the human operator. It seems that a system where the human operator is no longer necessary is attractive and is promoted by many stakeholders. The push and political agenda will probably continue to favor the notion of full automation as defined in the many different descriptions of automation.

There is an alternative to this approach. An approach where it’s important to try understanding the totality of the system before we try change the parts within the system. It assumes that the Aviation system is a complex system that is dynamic, adaptive, and evolving and where context, self-organization and emergence is always present. Taking this view leads us to a different path for designing today’s aviation system. An approach based on non-linearity, circular design, and the understanding of how the parts relate to each other. This reflects the centrality of a system’s approach to design and operation in the JCHMS narrative.

IFATCA is of the opinion that adopting the path of JCHMS and the thinking that the ATM system performs best by designing and integrating technology into the equation of designers, technology, and a human operator.

Changes in the ATM domain and at the ATM /CNS front are of permanent nature and challenges of research, development and transition to introduce these changes are daily life for ANSPs and their Staff. Be it Air Traffic Controllers, Technicians, Engineers, Managers and Decision Makers. The so-called New Technologies leading to digitisation or digitalisation, including Artificial Intelligence (AI) and Machine Learning (ML) are finding their ways into the ATM working environment.

Staying in control is important. The introduction of AI and ML will introduce new uncertainties that may reduce the span of control that we have. It’s challenges that needs to be dealt with and we think that using the JCHMS approach will ensure that.

IFATCA’s view on how to design and integrate new technology to the Aviation system is based on the following assumptions:

1. The more advanced a control system is, the more crucial may be the contribution of the human operator.
2. We need technology that makes us smarter – not smarter technology.
3. New technology needs to be designed by designers and operators with overall system goals in mind.

There is a need to move away from the idea of removing the human operator through automation AI and ML, and towards joint ‘thinking’ model, where the designer, the technology and the human operator is the model for future technological solutions. The aim of the joint view is to combine the strengths of humans and technology so that they amplify each other and that the ability to respond and anticipate surprises is increased or at least maintained.

IFATCA was invited to present the submitted working paper to conference. SESAR EASA coordinator was able to present the paper at the International Conference on Cognitive Aircraft Systems (ICCAS) . (Annex I)

Here is a summary of this paper:

Our approach to designs for human-system integration resulted in designs with reduced margins to manage and work with uncertainty and surprise within the work systems. This paper argues that technological designs often underperform compared to the promised benefits delivered. The reason for this is principally because designs have been based on a strategy where practitioners e.g., ATCOs, pilots etc, are expected to take over in abnormal conditions – the so called ‘left-over’ design strategy’ or the. There is a need to retain the human in the system to ‘complete the design, so as to adapt to the situations that designers never anticipated’. We argue that the need to change this philosophy of design is necessary, because we cannot think of engineering as a design without considering the people and the organisations that go with it. The operating environments of interest here, complex macro- cognitive work designs, are what is mentioned as socio-cognitive systems and are confronted with the challenge of digitisation and integration of artificial intelligence.

2.3. Challenges from the Introduction of Artificial Intelligence in the European Air Traffic Management System

IFATCA’s conference paper for the International Federation of Automatic Control was accepted as a conference paper which was published under the following link: https://www.sciencedirect.com/science/article/pii/S240589632201730X

Here is a summary of this paper:

The Air Traffic Management (ATM) system can be defined as a “Joint Cognitive System” of people, teams, and artifacts that adapts to the challenges and demands posed by familiar and unfamiliar situations in a dynamically evolving operational context. In the era of digitalization and Big Data we live, an incremental modernization of the ATM system is expected in the coming years with the pervasive implementation of Artificial Intelligence (AI) and Machine Learning (ML). In this paper, we present the findings from an initial attempt to detect and document the fundamental challenges of the introduction of AI, in the European ATM system through the lens of Cognitive Systems Engineering paradigm. We also discuss how these challenges give rise to difficult to resolve safety and performance related patterns in the ATM system.

2.4. Single European Sky and Resilience in ATM – Can This be a “Win-Win” for the Aviation Industry – the IFATCA Input

IFATCA submitted a paper to the FABEC Resilience Conference. EVP European and Stathis Malakis were able to present the ideas around Resilience in ATM.

Here is a summary of this paper:

In this paper we argued that SES has a large share of responsibility in institutionalizing financial and performance fragility to the European ANSPs, we provided a framework of possible solutions to counteract both financial and performance fragility and we stressed the importance of reframing the SES Environment performance indicator to embrace the interdependencies between several performance areas. Thus, by creating a systemic approach to manage the ATM contribution to decarbonization IFATCA claims that this could be a win- win situation. The research and development of the SESAR projects will have to focus on considering the aspects that we have highlighted so far. Although we are aware of the need to aim for an ATM system that is increasingly attentive to emissions and environmental issues, it is necessary to ensure that these objectives do not conflict with the highest safety standards. Furthermore, it is necessary that any new procedures studied are not going to provide additional burdens to the personnel involved with new tasks that could, in an already particularly congested situation, further aggravate the work of the operating personnel.

2.5. A Framework for Supporting Adaptive Human-AI Teaming in Air Traffic Control

IFATCA submitted a paper to the 25th International Conference On Human-Computer Interaction.

Here is a summary of this paper:

In recent years, the growth of cognitively complex systems has motivated researchers to study how to improve these systems’ support of human work. At the same time, there is a momentum for introducing Artificial Intelligence (AI) in safety critical domains. The Air Traffic Control (ATC) system is a prime example of a cognitively complex safety critical system where AI applications are expected to support air traffic controllers in performing their tasks. Neverthe- less, the design of AI systems that support effectively humans poses significant challenges. Central to these challenges is the choice of the model of how air traf-fic controllers perform their tasks. AI algorithms are notoriously sensitive to the choice of the models of how the human operators perform their tasks. The design of AI systems should be informed by knowledge of how people think and act in the context of their work environment. In this line of reasoning, the present study has set out to propose a framework of cognitive functions of air traffic controllers that can be used to support effectively adaptive Human – AI teaming. Our aim was to emphasize the “staying in control” element of the ATC. The proposed framework is expected to have meaningful implications in the design and effective operationalization of Human – AI teaming projects at the ATC Operations rooms.

(Laursen, T., Smoker, A.J., Baumgartner, M., Malakis, S. & Berzina, N.,  supported by the IFATCA JHMS group, International Federation of Air Traffic Controllers Associations, Montreal, Canada, tom.laursen@ifatca.org)

Introduction

Our approach to designs for human-system integration resulted in designs with reduced margins to manage and work with uncertainty and surprise within the work systems. This paper argues that technological designs often underperform compared to the promised benefits delivered. The reason for this is principally because designs have been based on a strategy where practitioners e.g., ATCOs, pilots etc, are expected to take over in abnormal conditions – the so called ‘left-over’ design strategy’ or the (Inagaki, T, 2014, p235)). Inagaki also argues, citing Rasmussen & Goodstein, that there is a need to retain the human in the system to ‘complete the design, so as to adapt to the situations that designers never anticipated’ (Inagaki, 2014, p235) We argue that the need to change this philosophy of design is necessary, as Boy argues: “We cannot think of engineering a design without considering the people and the organisations that go with it” Boy argues (Boy, 2020). The operating environments of interest here, complex macro-cognitive work designs, are what Boy refers to as socio-cognitive systems (Boy, 2020) and are confronted with the challenge of digitisation and integration of artificial intelligence.

Uncertainty and surprises will always be an element of complex systems,

Complexity research (Flach, 2014; Heylighen, Cilliers & Gershenson 2007, Cilliers, 2000) and the study of chaotic dynamics have demonstrated that uncertainty and surprise are fundamental aspects of the world around us (Eisenberg, Seager, Alderson, 2019; Lanir, 1983). Instead of an ordered system, such as machines, the aviation system is a complex system whose properties emerge from nonlinear interactions of numerous different agents. These interactions, and the interplay between them, create uncertainty and fundamental surprises (Eisenberg, Seager, Alderson, 2019. Woods et al, 2010, Lanir, 1983) which need to be managed in ways where as far as possible lead to being able to stay in control (McDaniel, Jr., Driebe, 2006).

The human or the machine?

Today’s aviation system consists of many different actors and agents that affect the ability to respond to uncertainty and surprises. There is a political level, an organisational level, a social dimension, training of practitioners, and numerous others. Knowing this, the only model of a human system is the system itself. We assume that there is a basic shared model of operation such as common ground in joint activity (Klein et al, 2005) between different actors. The basic model of operation consists of two interdependent processes. One is the process of preparing the other is the constant real-time adaptive capacity – that is the capacity to adapt to situational and fundamental surprises (Eisenberg et al 2019) and performance variability whilst sustaining production and system goals, which practitioners deliver principally.

Table 1: Differing design requirements of system designers and system end users (Klein, 2022)

Taking the designer’s view there are some caveats that we have to be aware of. Again, Bainbridge describes it in this way:

• One, that designer assumptions can be a major source of operating problems and,
• The second problem is that the designer who tries to eliminate the practitioner, the left-over functions, still leaves the practitioner to do the tasks which the designer cannot think how to automate.
• An additional problem is that the most successful automated systems, with rare need for manual intervention, may need the greatest investment in human practitioner training.

Taking the joint cognitive and the human system integration approaches (Hollnagel, Woods, 2005; Boy, 2020) are extant philosophies for collaboration between technology and the human which retain control in real-time operation.

Malakis Stathis*. Baumgartner Marc*. Berzina Nora*. Laursen Tom *. Smoker Anthony*. Poti Andrea*. Fabris Gabriele*.

* Supported by the IFATCA Joint Cognitive Human Machine System (JCHMS) group. International Federation of Air Traffic Controllers Associations, Montreal, Canada. (email: stathis.malakis@gmail.com)

Abstract

The Air Traffic Management (ATM) system can be defined as a “Joint Cognitive System” of people, teams, and artifacts that adapts to the challenges and demands posed by familiar and unfamiliar situations in a dynamically evolving operational context. In the era of digitalization and Big Data we live, an incremental modernization of the ATM system is expected in the coming years with the pervasive implementation of Artificial Intelligence (AI) and Machine Learning (ML). In this paper, we present the findings from an initial attempt to detect and document the fundamental challenges of the introduction of AI, in the European ATM system through the lens of Cognitive Systems Engineering paradigm. We also discuss how these challenges give rise to difficult to resolve safety and performance related patterns in the ATM system.

Keywords: Cognitive Systems Engineering, Joint Cognitive Systems, Air Traffic Control, Air Traffic Management, Artificial Intelligence, Machine Learning, Digitalization.

1. Introduction

It is widely accepted that a pilot in a cockpit is one of the earliest and most enduring symbols of what we can achieve with a positive and collaborative working relationship between humans and machines. If we make a step further and consider as the unit of analysis the airplane, i.e., the pilots plus flight control and automation, as the flight Joint Cognitive System (JCS) then the Air Traffic Management (ATM) is the environment. It is also possible to go one more step further and consider the pilots and the ATM as one system—the traffic flow JCS (Figure 1)— in which case the environment is the airlines, the airports, and the other aviation stakeholders (Hollnagel, 2007). In the era of digitalization and Big Data, this traffic flow JCS faces important and potentially disruptive challenges with the introduction of Artificial Intelligence (AI) both in the air and the in the ground components. It is ubiquitous that the use of AI is spreading rapidly in every industry with aviation and ATM making no exception.

However, innovative technologies not only provide capacity enhancement opportunities and other performance improvements but also raise new regulatory, safety, cognitive and operational challenges, and tradeoffs. Therefore, there is an urgent need to examine the introduction of AI cautiously. In this paper, we present an initial attempt to detect and document the fundamental challenges of implementing AI, in the European ATM system through the lens of Cognitive Systems Engineering (CSE) paradigm (Hollnagel and Woods, 2005; Woods and Hollnagel, 2006).

1.1 Motivation

The significant and continued growth in air traffic in the years prior to the COVID-19 pandemic has prompted considerable exploration of the use of AI in the ATM. It is expected that AI will provide the additional capacity to meet the challenges of increasing air traffic complexity due to sustained growth and new airspace users and support more efficient and environmentally friendly operations while maintaining and increasing current safety levels. Modern ATM systems comprise many airspace sectors with varying air traffic flows that interact in complex ways and evolve dynamically. ATM is a work domain that relies on the cognitive functions of Air Traffic Controllers (ATCOs) and their collaboration with flight crews, airport operators, network managers and the other aviation stakeholders to control the airspace, manage safety and adapt to the changing demands of new technological initiatives (Kontogiannis and Malakis, 2017). According to EASA (2021) while the concept of AI has been in existence since the 1950s, its development has significantly accelerated in the last decade due to three factors: a) significant improvements in the capacity to collect and store massive amounts of data (i.e., Big Data), b) significant increases in computing power and c) development of powerful algorithms and methods. From a purely AI view the ATM system is a real-time safety- critical decision-making process in highly dynamic and stochastic environments where human air traffic controllers monitor and direct many aircraft flying through its designated airspace sectors (Brittain et al, 2020). The extensive introduction of AI is expected to create a new ATM environment which will be tightly coupled, more complex to cope, with increased air traffic, and pressing needs for a) minimization of delays, b) accommodating a diverse array of autonomous aircraft, c) operating in adverse weather, d) smoothing out 4D aircraft trajectories and e) minimizing environmental impact. AI is expected to increase the resilience and the flexibility of the system (i.e., increase support during emergency in flight or on the ground or unusual situations, as severe weather, failures etc.) and to increase the situational awareness of the operators (e.g., ATCOs and Pilots). In parallel ongoing EASA’s projects such as Extended Minimum Crew Operations (eMCOs) and Single Pilot Operations (SiPOs) rely heavily on AI and the application of powerful Machine Learning (ML) methods. eMCOs are defined as operations where the flight time is extended by means of rest in flight with the minimum flight crew. It is achieved by allowing operations with one pilot at the controls, during the cruise flight phase; however, offering an equivalent overall level of safety through compensation means (e.g., ground assistance, advanced cockpit design with workload alleviation means, pilot incapacitation detection). SiPOs are defined as end-to-end single-pilot operations. Air operations regulation already foresees conditions and limitations under which these types of operations are allowed. In the future, it is expected that these conditions and limitations will need to evolve to extend single-pilot operations to large airplanes, provided that compensation means (e.g., ground assistance, advanced cockpit design with workload alleviation means, capability to cope with pilot incapacitation) are in place in order to provide for an overall level of safety equivalent to today’s two-pilot operations. EASA is working with interested industry stakeholders to explore the feasibility of such operational concepts, while maintaining current safety levels. It is evident that both projects that rely heavily on AI and ML will need ATM support and therefore introduce new operational requirements for the ATC. All these will impose a brand-new array of challenges to the ATM systems in the next 10 to 15 years. In this context we attempted to elicit and document fundamental challenges to the ATM system from the introduction of AI with of the view of drawing the attention to the potential side effects that we must act upon promptly.

2. Methodology

We used a range of methods over several phases of fieldwork, documentation analysis and finally divergent thinking, comparative reasoning, and integrative thinking to compile the final list.

In the fieldwork phase, we used many methods. These ranged from participation in structured group discussions (e.g., concerning AI certification) and in-depth discussions (e.g., AI related projects). From a Cognitive Systems Engineering (CSE) perspective, these techniques belong to the ‘natural history’ family of methods that are based on a diverse collection of observations in situ (Hoffman and Militello, 2009). This was a recurrent step that was used throughout the whole process. The result of each round was an improved version of the list of fundamental challenges. The next step was to consolidate the list by performing a documentation analysis. During this step we applied a documentation analysis of the most recent reports, white papers, position papers and technical documents from ATM and aviation organizations (EASA, 2020, 2021; EUROCONTROL, 2021a, 2021b; CANSO, 2021, 2021) regarding digitalization and AI in the European continent. The next step was to perform a literature review. There is an extensive body of high-quality research in the human factors and CSE literature that can inform the development and application of automated systems which is beyond the scope of this paper. Therefore, we decided to concentrate on some influential research publications and reports in the areas of automation AI, CSE and ATM (Bainbridge, 1983; Parasuraman and Riley,1997; Dekker and Woods, 1999; Moray and Inagaki, 1999; Parasuraman et al. 2000; Russell et al 2010, Woods and Sarter, 2000; Woods and Branlat, 2010; Woods et al, 2010; Norman, 2013). We also reviewed similar research efforts (Prevot et al. 2012), acceptance of automation studies (Westin, et al. 2016) and introduction of ML techniques can be used in developing classification rules and eliciting knowledge in the area ATM system (Malakis et al. 2020). Finally, we applied divergent thinking, comparative reasoning, operational expertise, and integrative thinking by capitalizing on the knowledge and operational expertise of the team members through successive rounds of drafting, commenting, and finalizing the list.

3.2 ANSP / Business

The challenges in the ANSP / Business area are the following:

• Organizational transformation issues.
• Insurance and liabilities.
• Constraints from the early adoption of new technology.
• Unclear cost-effectiveness benefits.
• Complexity of service borders and scope definition.
• Costly reinforced cybersecurity infrastructure.
• Complexity of service continuity requirements.
• Disruption of established knowledge sharing, learning procedures, and practices within ANSPs.
• Encourage simulation, training, and feedback paths from operators, to be shared with systems developers.
• Build a new change management policy to gradually drive operators through new technologies and working methods.

3.3 Technical

The challenges in the technical area are the following:

• Complexity of sharing of AI infrastructure between countries.
• Extreme dependency of the ML models on the datasets that are used for training, validation, and testing.
• Curse of dimensionality, which refers to the extraordinarily rapid growth of the complexity, as the number of variables (or dimensions) increases (Bellman, 1957).
• Significant differentiation of AI solutions for Tower, Approach and Area Control operations.
• Tailor made AI solutions for each use case that cannot be easily generalize across same type of units in different geographical areas.
• Consolidation of critical data to guarantee consistency, integrity, and safety of displayed information on the Controllers’ Working Positions.
• Develop backup systems to face failures or unusual situations in complex AI systems.
• Define effective AI methods to detect, frame into context, and interpret into decisions weak signals.

4. Discussion

Although the pandemic of Covid-19 significantly decreased the number of flights by slowing down the air transportation system in 2020 and 2021, this is not going to be a permanent situation. Commercial aviation already (June 2022) shows signs of a significant rebounce. The ATM systems is operating on its usual high traffic levels and the need from the introduction of AI in the ATM will gain momentum. AI is promising costs reduction, flight efficiency, improved strategic planning, enhanced trajectory prediction, and fuel efficiency to name but a few. However, it is evident from the challenges above that many and difficult to tackle challenges emerge. The range of challenges that were identified previously may well be the enablers of following safety and performance related patterns.

4.1 Difficult organizational and operational trade-offs

Operations rooms are hectic workplaces, and in many cases, work demands exceed resources, so ATCOs have to do their best and manage their traffic by adjusting their practices to meet existing conditions. In this sense, they are trying to maintain a continuous balance between demands and resources. Moreover, this does not only concern high workload situations, but also low workload periods, during which ATCOs need to maintain their performance and awareness levels despite not having enough to do to keep them engaged. The effort to tailor human performance to work demands can be described as if it involved a trade-off between efficiency and thoroughness. This view has received particular emphasis from Hollnagel’s (2009) proposition of the Efficiency-Thoroughness Trade-Off (ETTO) principle. To cater for efficiency, ATCOs generally try to achieve their goals by keeping their efforts and resources (for example workload) as low as possible. Safety, on the other hand, requires that more resources are spent on ensuring that the necessary conditions are in place, so that performance goals are achieved without risks. Safety implies that ATCOs spent more time in thinking whether preconditions for an activity are met, execution conditions are right and preparations for contingencies are made in advance. ATCOs have to reach both their safety and performance goals, neither of which should be achieved at the expense of others. In this line of reasoning unresolved challenges from AI projects will most certainly disrupt this fragile balance and create room for new and more complex dilemmas. Additionally, very thorough investigation will be required to determine how many, and which ATCO tasks will AI take over and/or assist with. For example, completing the simple and routine tasks would be beneficial in a high workload situation, as this would allow ATCOs to focus on the complex tasks, but detrimental in a low workload situation, as it would even furthermore reduce ATCOs engagement and could lead to reduced situational awareness and performance levels.

4.2 Difficult organizational and operational trade-offs

Current safety methodologies cannot cope well with AI related projects especially when it comes to learning assurance (EASA, 2020, 2021). EASA developed a multiyear project to address exactly this type of weakness in certification. ANSPs’ safety methodologies focus on the needs of single agents, but do not allow risks of different types and sources to be assessed with reference to each other, singly or in combination in dynamic environments. Hence, dysfunctional interactions arise that are beyond the control of single agents and current methodologies cannot adequately deal with them.

4.3 Transform or transfer hazards to other stakeholders

In many cases, risks may be transformed or transferred among ATM stakeholders since the solution of one’s own concerns may create problems elsewhere. For instance, adverse weather is a safety hazard for all flight operations. When weather cells are encountered, flight crews may request to circumnavigate cells which could increase traffic complexity for ATCOs particularly in congested airspaces. Hence, granting a cell circumvention to aircrews may increase the risk of separation minimum infringement and reduce the margin of maneuver for controllers. On the contrary, flight operation hazards are effectively reduced. The increase in air traffic complexity (e.g., more conflicts points) as several air crews are requiring changes to their routes to circumvent the weather cell cannot always be mitigated by ATFCM restrictions alone. Corver and Grote (2016) showed that area controllers, in addition to reducing and acknowledging uncertainty, may deliberately increase uncertainty to increase flexibility for other actors in the system to meet their operational goals. On severe weather days, ATCOs often forgo a few procedures, e.g., coordination requirements, to increase their overall efficiency in handling the complex traffic and to ensure the safety of the aircraft. These results are particularly important as uncertainty is likely to increase in future AI supported operations of area control ATC especially since explainability issues will remain an issue (Barredo Arrieta, 2020). Therefore, risk transferability between ATM stakeholders which is common now and, in a sense, understood (because work practices are built upon it) will most probably be disrupted with unwanted safety and performance consequences. For this reason, any AI that is introduced in ATC has to be modeled in a way that matches the ATCO working practice and culture.

Malakis, S., P. Psaros, T. Kontogiannis, and C. Malaki. 2020. “Classification of Air Traffic Control Scenarios Using Decision Trees: insights from a Field Study in Terminal Approach Radar Environment.” Cognition Technology and Work 22: 159–179.

Moray, N. and Inagaki, T. (1999) Laboratory studies of trust between humans and machines in automated systems. Trans. Inst. MC 21(4/5), pp203-211).

Norman, D. (2013) The Design of Everyday Things. MIT Press.

Parasuraman, R. & Riley, V. (1997). Humans and Automation: Use, Misuse, Disuse, Abuse. Human Factors, 39(2).

Parasuraman, R., Sheridan, T.B., Wickens, C.D. (2000). A Model for Types and Levels of Human Interaction with Automation. IEEE Transactions on Systems, Man and Cybernetics – Part A: Systems and Humans, Vol 30, No. 3, May.

Prevot, T., Homola, J. R., Martin, L. H., Mercer, J. S., and Cabrall, C. D. (2012). Toward Automated Air Traffic Control-Investigating a Fundamental Paradigm Shift in Human/Systems Interaction. International Journal of Human-Computer Interaction, 28(2), 77–98.

Russell, S. J., P. Norvig, and E. Davis. (2010). Artificial Intelligence: A Modern Approach. 3rd ed. Upper Saddle River, NJ: Prentice Hall.

Westin, C., Borst, C., and Hilburn, B. (2016). Strategic Conformance: Overcoming Acceptance Issues of Decision Aiding Automation? IEEE Transactions on Human-Machine Systems, 46(1), 41–52.

Woods, D. D. and Branlat, M. (2010). Holnagell’s test: Being in control of highly interdependent multi-layered networked systems. Cognition, Technology & Work, 12, 95–101.

Woods, D. D. and Hollnagel, E. (2006). Joint Cognitive Systems: Patterns in Cognitive Systems Engineering. Boca Raton, FL: CRC Press.

Woods, D. D., Dekker, D., Cook, R., Johannesen, L., and Sarter, N. (2010). Behind Human Error. Second Edition. Farnham: Ashgate Publishing.

Woods, D.D., Sarter, N.B., (2000). Learning from automation surprises and ‘going sour’ accidents. In: Sarter, N.B., Amalberti, R. (Eds.), Cognitive Engineering in the Aviation Domain. Lawrence Erlbaum Associates, Mahwah, NJ, pp. 327–354.

(IFATCA Joint Cognitive Human Machine Systems Team)

Introduction

Resilience is defined as the ability to succeed under varying conditions, so that the number of intended and acceptable outcomes is as high as possible (Hollnagel, 2018). In the context of resilience, Sustained adaptability refers to the ability to continue to adapt to changing environments, stakeholders, demands, contexts, and constraints (in effect, to adapt how the system in question adapts) (Woods, 2018). The EU’s Single European Sky (SES) legislation has introduced a Performance Scheme which defines mandatory performance targets for the Air Navigation Service Providers (ANSPs) of EU Member States. The Performance Review Body (PRB) advises and supports the Commission in setting up binding performance targets and thereby, in a way, acts as the regulator for ATM at the European level. In this sense, in the European ATM system, resilience and sustained adaptability are considered to be defined by the scope of the SES Performance Scheme.

The Covid-19 crisis demonstrated that forecasting and charging scheme deficiencies can produce results that provide a false reassurance about the degree of financial stability of the ANSPs. Undoubtedly, the SES has a large share of responsibility in institutionalizing the financial and performance fragility of the European ANSPs. Furthermore, by considering the unanticipated synergy between the effects of the pandemic of Covid-19 and the continuing war in Ukraine we argue that volatility in air traffic and consequently in financial results may be the norm and not the exception anymore.

For the ATM system to sustain resilient performance it needs the capabilities to respond to varying conditions typically through adaptation and coordination and to be able to sustain this adaptability. Traditionally, ANSPs, and the wider context of ICAO, EASA, EUROCONTROL, local CAAs and many more have built systems with resilient performance that have rarely led to fatal accidents. PRC (2022) and EASA (2022) argue that historical evidence shows that overall safety levels in the EUROCONTROL and EASA member states area are high, however there is no room for complacency. Whereas in terms of safety we are performing well, the ATM system has had bigger problems when it comes to efficiency. The SES is the ‘dark’ side of resilience and has reduced the ANSPs ability to adapt to disturbances and crisis. Focusing on the SES’s influence on the ATM systems ability to adjust it functions, configurations and sources of adaptive capacity mainly operational capacity, to the changes that have influenced the system for last two decades IFATCA contends that:

It is evident, in the authors view, that Covid-19 crisis uncovered all three patterns of adaptive breakdown in the European ATM system. In this line of reasoning, we propose a framework of possible solutions to counteract both financial and performance fragility:

1. Increased Personnel Redundancy: Hiring and training of new Air Traffic Controllers. Normalizing lean hiring and rostering personnel strategies is a certain path to performance deficiencies. A large share of delays this summer and the last 15 years is directly attributed to ATC staffing (PRC, 2022). This is a textbook example of exhausting capacity to adapt as disturbances/challenges cascade (Decompensation).
2. Introduction of Stress Tests. A stress test, in financial terminology, is an analysis or simulation designed to determine the ability of a given institution to deal with an economic crisis. Economic and financial models have limitations and constraints, including misspecification, estimation using assumed and sometimes inaccurate probability distributions, etc. Therefore, using such models to estimate the potential impact of shocks may lead to increasingly inaccurate estimates. Instead of doing financial projection on a “best estimate” basis, an ANSP and its regulators may do stress testing where they look at how robust an organization is in certain crashes. Useful performance indicators can emerge if these stress tests are properly designed and executed. Exploring the utilization of other criteria for the charging scheme than the ones based solely on traffic numbers may emerge as a consequence of these stress test. This is a textbook example of the world changes but the system remains stuck in what were previously adaptive strategies (Getting stuck in outdated behaviors).
3. Moving Away from Uninformative Safety, Financial and Performance Indicators. Current performance indicators belong to the ‘lagging category’ with minimal predictive value that may mislead from the complexities of the actual operational context. For instance, measuring an ANSP performance in terms of delay minutes accumulated is incorrect and highly misleading. It could have been a very busy day with adverse weather, lots of CBs (Cumulonimbus cells) and circumnavigation and controllers performing above standard to ensure separation and yet – delay minutes indicate low performance levels. A statistical analysis of performance-related information may be indicative of emerging threats from many sources. However, a resilient performance intelligence function requires a combination of reactive, proactive, and predictive indicators that will detect emerging threat patterns. The information published by PRC can only be used to assess how frequently organizations come close to the performance, financial and safety margins. Performance indicators are complemented with trends in the last decades and some elementary descriptive statistics. Therefore, stakeholders cannot discern anything more that some well- established but highly uninformative trends (Getting stuck in outdated behaviors).
4. Restructuring of the Air Traffic Flow and Capacity Management (ATFCM) system. ATFCM systems needs to move towards a more command and control system and consider rectifying the inefficiencies of the Computer Assisted Slot Allocation (CASA) algorithms. Slot allocation algorithms rely mostly on physical factors (e.g., prevailing winds estimated many hours earlier than the actual flight) but fail to account of actual aircraft performance, flight profile preferences by airlines and direct routings given by controllers in the en-route phase of flight. Hence, some ATC units may be stretched above their capacity or others may be underutilized because of under- specifications in the ATFCM algorithms. PRC (2022) acknowledged these deficiencies and provided some remedies. This is a textbook example of behavior that is locally adaptive, but globally maladaptive as it to coordinate different groups at different echelons as goals conflict (Working at cross-purposes).
5. Cultivating Adaptive Diversity. In the European ATM there are several micro- cultures of controllers. These are controllers’ communities exercising and promoting operations based on their local practices and local affordances. For instance, in a major ACC there are 3 different sector groups, and each group has a different working style/culture. These differences happen because the sector size, crossing points, traffic flows and hence needs vary. An adaptive way of working would increase the overall resilience. Institutionalizing a know-your-neighboring-units-work-practices awareness program would be beneficial to ATCOs during their basic or rating training with (familiarization visits to various ANSPs). Getting to know the working styles of your neighbors could help the ATCO understand them better, thus easier to adapt (Working at cross-purposes).
6. Tap local knowledge. Controllers at the sharp end of each unit know best the causes of safety and performance deficiencies. They rely on “localized performance and safety intelligence” at the individual and team level and respond accordingly. Likewise, parent ANSPs organizations may need to drive more of their sense making to local levels and exploit these more these localized performance and safety intelligence to detect and respond to weak signals of performance and safety deficiencies (Decompensation).

Baumgartner, M., Finger, M. (2014). The Single European Sky gridlock: A difficult 10-year reform process, Utilities Policy. http://dx.doi.org/10.1016/j.jup.2014.03.004.

EASA, (2022). Annual Safety Review 2021. European Union Aviation Safety Agency.

Hollnagel, E. (2018). Safety-II in Practice: Developing the Resilience Potentials. Routledge. ICAO, (2014).

IFATCA JCHMS Group, (2022a). Reducing the gap between designers and users, Why are aviation practitioners here again? Presented at The International conference on Cognitive Aircraft Systems – ICCAS, Toulouse France, June 2022.

IFATCA JCHMS Group, (2022b). Challenges from the Introduction of Artificial Intelligence in the European Air Traffic Management System. Presented at IFAC/IFIP/IFORS/IEA Symposium on Analysis Design and Evaluation of Human-Machine Systems, San Jose USA, September 2022.

IFATCA, (2007), A statement on the future of Global Air Traffic Management. Montréal, IFATCA.

Operational Opportunities to Reduce Fuel Burn and Emissions. Document 10013. First Edition. Montreal: International Civil Aviation Organization.

PRC, (2020). Performance Review Report 2019. Performance Review Commission. Eurocontrol, Brussels.

PRC, (2021). The role of ANS in a decarbonising world –Current knowledge and questions for debate. Performance Review Commission. Brussels, Eurocontrol.

PRC, (2022). Performance Review Report 2021. Performance Review Commission. Eurocontrol, Brussels.

Taleb, N.N. (2007). The Black Swan: The Impact of the Highly Improbable. Penguin, London.

Woods, D. D., Branlat, M. (2010). Hollnagel’s test: being ‘in control’ of highly interdependent multilayered networked systems. Cognition, Technology and Work 12(2), 95-101.

Woods, D. D., Branlat, M. (2011). Basic patterns in how adaptive systems fail. In E. Hollnagel, J. Pariès, D. D. Woods, & J. Wreathall (Eds.), Resilience Engineering in Practice. Farnham, UK: Ashgate.

Woods, D.D. (2018) The theory of graceful extensibility: basic rules that govern adaptive systems. Environment Systems and Decisions. 38:433–457.

Malakis Stathis* [0000-0003-2182-4597], Baumgartner Marc*, Berzina Nora*, Laursen Tom*, Smoker Anthony*, Poti Andrea*, Fabris Gabriele*, Velotto Sergio*, Scala Marcello*, Kontogiannis Tom**

* IFATCA Joint Cognitive Human Machine System (JCHMS) group. International Federa- tion of Air Traffic Controllers Associations, Montreal, Canada.

** Cognitive Ergonomics & Industrial Safety Laboratory, Department of Production Engi- neering & Management, Technical University of Crete, Chania Hellas stathis.malakis@gmail.com

In recent years, the growth of cognitively complex systems has motivated researchers to study how to improve these systems’ support of human work. At the same time, there is a momentum for introducing Artificial Intelligence (AI) in safety critical domains. The Air Traffic Control (ATC) system is a prime example of a cognitively complex safety critical system where AI applications are expected to support air traffic controllers in performing their tasks. Nevertheless, the design of AI systems that support effectively humans poses significant challenges. Central to these challenges is the choice of the model of how air traffic controllers perform their tasks. AI algorithms are notoriously sensitive to the choice of the models of how the human operators perform their tasks. The design of AI systems should be informed by knowledge of how people think and act in the context of their work environment. In this line of reasoning, the present study has set out to propose a framework of cognitive functions of air traffic controllers that can be used to support effectively adaptive Human – AI teaming. Our aim was to emphasize the “staying in control” element of the ATC. The proposed framework is expected to have meaningful implications in the design and effective operationalization of Human – AI teaming projects at the ATC Operations rooms.

In recent years, there has been a realization that the growth of cognitively complex systems has motivated researchers and practitioners from diverse backgrounds to study how to improve these systems’ support of human work. At the same time, there is a momentum for introducing Artificial Intelligence (AI) in the safety critical domains building upon:

a) Advances in capacity to collect and store massive amounts of data.

b) Significant increases in computing power, and

c) Development of increasingly powerful algorithms and architectures.

The Air Traffic Control (ATC) system is a prime example of a cognitively com-plex safety critical system where AI applications are introduced. Soon AI systems are expected to support even more than now air traffic controllers, flow controllers, operational supervisors, and flight information officers in performing their safety critical tasks within the ATC ecosystem. Furthermore, the widespread introduction of AI is expected to create a new ATC environment, which will be tightly coupled, more complex to manage, and with pressing needs for:

a) Minimization of delays.

b) Accommodating a diverse array of autonomous aircraft.

c) Operating smoothly in adverse weather conditions.

d) Smoothing out 4D aircraft trajectories, and

e) Minimizing environmental impact.

European Union [1, 2] is envisioning a Digital European Sky and an irreversible shift to low and ultimately no-emission mobility. With this goal in mind, ATC and aviation will evolve into an integrated digital ecosystem characterized by distributed data services. This is planned to be accomplished mostly by leveraging digital tech-nologies to transform the aviation sector. The aim is to deliver a fully scalable Air Traffic Management (ATM) system for manned and unmanned aviation that is even safer than today’s, and, based on higher air ground integration. The ATM provides an essential service for aviation. While the essence of ATM is and will always re-main, to ensure the safe and orderly execution of all flights it needs to do so in the most environmentally friendly and cost-efficient way [1].

In this context, International Federation of Air Traffic Controllers (IFATCA) has created the concept of Joint Cognitive Human Machine System (JCHMS) or a Hu-man Centric approach and wishes to influence ICAO and standardization bodies such as EUROCAE and EASA in Europe as well as other global organization such as RTCA, SAE in the way these new technologies are designed. Nevertheless, the de-sign of AI systems that effectively support humans poses significant challenges [3].

In the ATC domain, as early as 1951, Fitts [4] proposed the use of automation to replace human functions with technology. In this approach the capabilities (and limitations) of people and machines are compared on a number of salient dimensions, and function allocation is made to ensure that the respective capabilities are used in an optimal way (Compensatory principle). However, it failed to acknowledge the essential condition that it is necessary at all times to remain in control of what hap-pens. Humans are aware of what they are doing and can imagine what the outcomes may be. Machines and technology can do neither. Digitalization relies on highly effective but poorly understood algorithms. By replacing human functions with technology that is not fully comprehensible, control is gradually and irretrievably lost.

It is acknowledged that although AI has many potential benefits, suffers from a number of challenges for successful performance in complex safety critical real-world environments, including brittleness, perceptual limitations, hidden biases, and lack of a model of causation important for understanding and predicting future events. These limitations mean that AI will remain inadequate for operating on its own in many complex and novel situations for the foreseeable future, and that AI will need to be carefully managed by humans to achieve their desired utility [5].

It is evident that in the era of digitalization and Big Data, the ATM ecosystem faces important and potentially disruptive challenges with the introduction of AI both in the air and the in the ground. It is ubiquitous that the use of AI is spreading rapidly in every industry with aviation and ATM making no exception.

However, innovative technologies not only provide capacity enhancement opportunities and other performance improvements but also raise new regulatory, safety, cognitive and operational challenges, and trade-offs [3]. Therefore, there is an urgent need to examine the introduction of AI and more importantly the human-AI teaming cautiously.

The significant and continued growth in air traffic in the years prior to the COVID-19 pandemic has prompted considerable exploration of the use of AI in the ATC. It is expected that AI will provide the additional capacity to meet the challenges of increasing air traffic complexity due to sustained growth and new airspace users and support more efficient and environmentally friendly operations while maintaining and increasing current safety levels. Modern ATM systems comprise many airspace sectors with varying air traffic flows that interact in complex ways and evolve dynamically. ATC is a work domain that relies on the cognitive functions of Air Traffic Controllers and their collaboration with flight crews, airport operators, network managers and the other aviation stakeholders to control the airspace, man-age safety and adapt to the changing demands of new technological initiatives [10].

From a purely AI view the ATM system is a real-time safety-critical decision-making process in highly dynamic and stochastic environments where human air traffic controllers monitor and direct many aircraft flying through its designated airspace sectors [11].

We followed the CSE approach research methods [6,7,8] We used a range of meth-ods over several phases of fieldwork, documentation analysis and finally divergent thinking, comparative reasoning, and integrative thinking to compile a framework of cognitive functions of Air Traffic Controllers that can support adaptive Human-AI teaming.

We applied a documentation analysis of the most recent reports, white papers, po-sition papers and technical documents from ATM and aviation organizations [12, 13, 14, 15, 16, 17] about digitalization and AI/ML in the European continent. The next step was to perform a literature review. There is an extensive body of research CSE literature that can inform the development and application of automated systems which is beyond the scope of this paper. Hence, we decided to concentrate on some influential research publications and reports in the areas of automation AI/ML, CSE and ATM [18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29]. In the final phase, we performed an extensive literature review focused on models of human performance, team performance, decision-making, anomaly response and cognitive functions [30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54 , 55, 56, 57, 58, 59]

4. A Framework for Supporting Adaptive Human-AI Teaming

The framework we propose examines six cognitive functions as follows:

• Steering or Goal Setting.
• Sensemaking and Mental Models.
• Common Operating Picture or Shared Mental Models.
• Coordination and Transfer of Control.
• Managing Changes.
• Operating or Planning-Doing-Checking cycle.

4.1 Steering or goal setting

An orientation of performance towards specific goals that can be modified later during events. It involves an assessment of the context of work in terms of threats, constraints, and resources. Sometimes explorative goals may be taken by air traffic controllers to see how the system reacts. Steering broadly refers to the choices and goals that people set in order to achieve their mission. In a complex environment, people may have to tackle multiple goals that compete for limited resources. To be successful in steering, people need to demonstrate willingness to make the necessary trade-offs and show an ability to spot and exploit leverage points [35]. Fundamental to steering is the presence of a large ‘set of action repertoires’ in the form of work practices honed through ‘enlightened’ experimentation [60].

AI could very well assist air traffic controllers in making more environment friendly decisions in situations where either solution method has no consequence on the overall traffic picture. If a conflict needs to be solved by either a level change or headings and it does not make a difference which one it is, AI would have the capability to advise which solution would be better for the environment based on the wind and atmospheric conditions. Making aviation more climate friendly is a goal made by the EU to which individual air traffic controllers have no chance to contribute to unless more assistance is provided in their decision making. Obviously, safety would come first, but with the way, we see AI potential in ATC as an assistant, air traffic controllers do not normally care for environmentally wise headings or a level change if there is no other traffic to consider.

1. SESAR JU.: Strategic Research and Innovation Agenda – Digital European Sky. (2020).
2. SESAR JU.: EUROPEAN ATM MASTER PLAN- Digitalising Europe’s Aviation Infrastructure. (2020).
3. Malakis, S. Baumgartner, M. Berzina, N. Laursen, T. Smoker, A. Poti, Andrea. Fabris, Gabriele. : Challenges from the Introduction of Artificial Intelligence in the European Air Traffic Management System, IFAC-PapersOnLine, Volume 55(29), 1- 6 (2022).
4. Fitts, P. M. (Ed). : Human engineering for an effective air navigation and traffic control system. Ohio State University Research Foundation. Columbus, OH (1951).
5. National Academies of Sciences, Engineering, and Medicine. : Human-AI Teaming: State-of-the-Art and Research Needs. The National Academies Press.  Washington, DC (2022).
6. Hollnagel, E. and Woods, D. D. Joint Cognitive Systems. Foundations of Cognitive Systems Engineering. Taylor and Francis, London (2005).

7. Woods, D. D. and Hollnagel, E. Joint Cognitive Systems: Patterns in Cognitive Systems Engineering. CRC Press, Boca Raton, FL (2006).
8. Militello L. G., Dominguez, C. O., Lintern, G., Klein G. The Role of Cognitive Systems Engineering in the Systems Engineering Design Process. Systems Engineering, 13: 261-273 (2010).
9. Hollnagel, E. Flight decks and free flight: Where are the system boundaries? Applied Ergonomics, 38, 409–416. (2007).
10. Kontogiannis, T., and S. Malakis. Cognitive Engineering and Safety Organization in Air Traffic Management. CRC Press; Taylor & Francis, Boca Raton, FL (2017).
11. Brittain, M. Yang, X. Wei, P. A Deep Multi-Agent Reinforcement Learning Approach to Autonomous Separation Assurance. arXiv 2020, arXiv:2003.08353. (2020).
12. EASA, Artificial Intelligence Roadmap 1.0. European Aviation Safety Agency. (2020).
13. EASA. First usable guidance for Level 1 machine learning applications. European Aviation Safety Agency. (2021).
14. Eurocontrol. A Patterns in How People Think and Work Importance of Patterns Discovery for Understanding Complex Adaptive Systems. Brussels: Eurocontrol. (2021).
15. Eurocontrol. Digitalisation and Human Performance. Hindsight 33. Winter 2021- 2022. Eurocontrol, Brussels (2021).
16. CANSO. Artificial Intelligence. CANSO Whitepapers Emerging Technologies for Future Skies. Civil Air Navigation Services Organization. (2021).
17. CANSO. Virtualisation. CANSO Whitepapers Emerging Technologies for Future Skies. Civil Air Navigation Services Organization. (2022).
18. Bainbridge, L. Ironies of automation. Automatica, 19, 775–780. (1983).
19. Parasuraman, R. Riley, V. Humans and Automation: Use, Misuse, Disuse, Abuse. Human Factors, 39. (1997).
20. Dekker, S. W. and Woods, D. D. To intervene or not to intervene: The dilemma of management by exception. Cognition Technology and Work, 1, 86–96. (1999).
21. Moray, N. and Inagaki, T. Laboratory studies of trust between humans and machines in automated systems. Trans. Inst. MC 21(4/5), 203-211. (1999).
22. Parasuraman, R., Sheridan, T.B., Wickens, C.D. A Model for Types and Levels of Human Interaction with Automation. IEEE Transactions on Systems, Man and Cybernetics – Part A: Systems and Humans, Vol 30, No. 3, May. (2000).
23. Russell, S. J., P. Norvig, and E. Davis. Artificial Intelligence: A Modern Approach. Third edition. Upper Saddle River, NJ: Prentice Hall. (2010).
24. Woods, D.D., Sarter, N.B., Learning from automation surprises and ‘going sour’ accidents. In: Sarter, N.B., Amalberti, R. (Eds.), Cognitive Engineering in the Aviation Domain. Lawrence Erlbaum Associates, Mahwah, NJ, pp. 327–354. (2000).
25. Woods, D. D. and Branlat, M. Holnagell’s test: Being in control of highly interdependent multi-layered networked systems. Cognition, Technology & Work, 12, 95–101. (2010).
26. Woods, D. D., Dekker, D., Cook, R., Johannesen, L., Sarter, N. Behind Human Error. Second Edition. Farnham: Ashgate Publishing. (2010).
27. Norman, D. The Design of Everyday Things. MIT Press. (2013).
28. Domingos, P. A few useful things to know about machine learning. Communications ACM 55(10):78–87. (2012).

29. Alpaydin, E. Introduction to machine learning, third edition. MIT Press, Massachusetts Institute of Technology, Cambridge. (2014).
30. Cohen, M.S., Freeman, J.T., Wolf, S.P., Meta-Recognition in time stressed decision making: recognizing critiquing and correcting. Human Factors 38, 206–219. (1996).
31. Colvin, K., Funk, K., Braune, R., Task prioritization factors: two part-task simulator studies. International Journal of Aviation Psychology 15, 321–338. (2005).
32. D’Arcy, J.F., Della Rocco, P., Air Traffic Control Specialist Decision Making and Strategic Planning – A Field Study (DOT/FAA/CT-TN01.05). DOT/FAA William J, Hughes Technical Center, Atlantic City International Airport, NJ. (2001).
33. Gronlund, S.D., Dougherty, M.R., Durso, F.T., Canning, J.M., Mills, S.H., Planning in air traffic control: impact of problem type. International Journal of Aviation Psychology 15, 269–293. (2005).
34. Klein, G.A. Recognition-primed decisions. In: Rouse, W.B. (Ed.), Advances in Man- Machine Systems Research, vol. 5. JAI Press, Greenwich, CT, pp. 47–92. (1989).
35. Klein, G.A., Sources of Power: How People Make Decisions. MIT Press, Cambridge, MA. (1998).
36. Kontogiannis, T. Stress and operator decision making in coping with emergencies. International Journal of Human-Computer Studies 45, 75–104. (1996).
37. Kontogiannis, T. Training effective human performance in the managing of stressful emergencies. Cognition Technology and Work 1, 7–24. (1999).
38. Rantanen, E.M., Nunes, A., Hierarchical conflict detection in air traffic control. International Journal of Aviation Psychology 15, 339–362. (2005).
39. Reynolds, T.G., Histon, J.M., Davison, H.J., Hansman, R.J. 2002. Structure, intent and conformance monitoring in ATC. In: Proceedings of the Air Traffic Management (ATM) Workshop on ATM System Architectures and CNS Technologies, Capri, Italy, 22–26 September (2002).
40. Seamster, T.L., Redding, R.E., Cannon, J.R., Purcell, J.A. Cognitive task analysis of expertise in air traffic control. International Journal of Aviation Psychology 3, 257–283. (1993).
41. Woods, D.D., Cognitive demands and activities in dynamic fault management: abduction and disturbance management. In: Standon, N. (Ed.), Human Factors of Alarm De-sign. Taylor & Francis, London. (1994).
42. Bowers, C.A., Jentsch, F., Salas, E., Brawn, C. Analyzing communication sequences for team training needs assessment. Human Factors 40, 672–679. (1998).
43. Cannon-Bowers, J.A., Tannebaum, S.I., Salas, E., Volpe, C.E. Defining competencies and establishing team training requirements. In: Guzzo, R.A., Salas, E. (Eds.), Team Effectiveness and Decision Making in Organizations. Jossey-Bass, San Francisco, pp. 333–381, (1995).
44. Entin, E.B., Entin, E.E. Assessing team situation awareness in simulated military mis-sions. Proceedings of the Human Factors and Ergonomics Society 44th Annual Meeting. Human Factors and Ergonomics Society Press, San Diego, CA. 73–77. (2000).
45. Salas, E., Sims, D.E., Burke, C.S. Is there a ‘‘big five’’ in teamwork? Small Group Re-search 36, 555–599. (2005).
46. Salas, E., Cooke, N.J., Rosen, M.A. On teams, teamwork, and team performance: discoveries and developments. Human Factors 50, 540–547. (2008).

47. Salas, E., Fiore, S.M. Team Cognition: Understanding the Factors that Drive Process and Performance. American Psychological Association, Washington, DC. (2004).
48. Orasanu, J.M., Decision Making in the Cockpit. In: Wiener, E.L., Kanki, R.G., Helmreich, R.L. (Eds.), Cockpit Resource Management. Academic Press, San Diego, pp. 137–172. (1993).
49. Patterson, E.S., Watts-Perotti, J., Woods, D.D. Voice loops as coordination aids in space shuttle mission control. Computer Supported Cooperative Work 8, 353–371. (1999).
50. Kontogiannis, T. Adapting plans in progress in distributed supervisory work: Aspects of complexity, coupling, and control. Cognition Technology and Work, 12, 103–118. (2010).
51. Malakis, S. and Kontogiannis, T. A sensemaking perspective on framing the mental picture of air traffic controllers. Applied Ergonomics, 44, 327–339. (2013).
52. Malakis, S. and Kontogiannis, T. Exploring team sensemaking in air traffic control (ATC): Insights from a field study in low visibility operations. Cognition Technology and Work, 16, 211–227. (2014).
53. Malakis, S., Kontogiannis, T., Kirwan, B. Managing emergencies and abnormal situations in air traffic control (Part I): Taskwork strategies. Applied Ergonomics, 41, 620–627. (2010).
54. Malakis, S., Kontogiannis, T., and Kirwan, B. Managing emergencies and abnormal situations in air traffic control (Part II): Teamwork strategies. Applied Ergonomics, 41, 628–635. (2010).
55. Klein, G., Snowden, D., Pin, C.L. Anticipatory thinking. In: Mosier, K., Fischer, U. (Eds.), Informed knowledge: Expert performance in complex situations. Psychology Press, London, UK. (2010).
56. Langan-Fox, J., Canty, J.M., Sankey, M.J. Human–automation teams and adaptable control for future air traffic management. International Journal of Industrial Ergonom-ics 39 (5), 894–903. (2009).
57. Luokkala, P., Nikander, J., Korpi, J., Virrantaus, K., Torkki, P. Developing a concept of a context-aware common operational picture. Safety Science 93, 277–295. (2017).
58. Steen-Tveit, K., Munkvold, B.E. From common operational picture to common situational understanding: An analysis based on practitioner perspectives. Safety Science 142. https://doi.org/10.1016/j.ssci.2021.105381. (2021).
59. Stein, E.S., Della Rocco, P.S., Sollenberger, R.L. Dynamic re-sectorization in air traffic control: a human factors perspective. FAA, U.S. Department of Transportation. DOT/FAA/TC-TN06/19 report. (2006).
60. Thomke, S. Enlightened experimentation. The new imperative for innovation. Harvard Business Review 79 (2), 66–75. (2001).
61. Grote, G. Management of Uncertainty: Theory and Application in the Design of Sys- tems and Organizations. Springer, Berlin. (2009).
62. Laursen, T., Smoker, A.J., Baumgartner, M., Malakis, S. Berzina, N.: Reducing the gap between designers and users, Why are aviation practitioners here again? International conference on Cognitive Aircraft Systems – ICCAS, Toulouse France, June 2022. (2022).
63. Boy, G. Human-systems integration. CRC Press. (2020).