The autonomous crewmate : A sociotechnical perspective to implementation of autonomous vehicles in sea rescue

(1)

Linköping University | Department of Computer and Information Science Master thesis | Cognitive Science Spring semester 2020| LIU-IDA/KOGVET-A--20/009--SE

The autonomous crewmate

A sociotechnical perspective to implementation of

autonomous vehicles in sea rescue

Author: Oscar Lundblad

Supervisors: Peter Berggren, Björn Johansson & Emma Jonsson Moderator: Arne Jönsson

(2)

Upphovsrätt

Detta dokument hålls tillgängligt på Internet – eller dess framtida ersättare – under 25 år från publiceringsdatum under förutsättning att inga extraordinära omständigheter uppstår.

Tillgång till dokumentet innebär tillstånd för var och en att läsa, ladda ner, skriva ut enstaka kopior för enskilt bruk och att använda det oförändrat för ickekommersiell forskning och för undervisning. Överföring av upphovsrätten vid en senare tidpunkt kan inte upphäva detta tillstånd. All annan användning av dokumentet kräver upphovsmannens medgivande. För att garantera äktheten, säkerheten och tillgängligheten finns lösningar av teknisk och administrativ art.

Upphovsmannens ideella rätt innefattar rätt att bli nämnd som upphovsman i den omfattning som god sed kräver vid användning av dokumentet på ovan beskrivna sätt samt skydd mot att dokumentet ändras eller presenteras i sådan form eller i sådant sammanhang som är kränkande för upphovsmannens litterära eller konstnärliga anseende eller egenart.

För ytterligare information om Linköping University Electronic Press se förlagets hemsida

http://www.ep.liu.se/.

Copyright

The publishers will keep this document online on the Internet – or its possible replacement – for a period of 25 years starting from the date of publication barring exceptional circumstances.

The online availability of the document implies permanent permission for anyone to read, to download, or to print out single copies for his/hers own use and to use it unchanged for non-commercial research and educational purpose. Subsequent transfers of copyright cannot revoke this permission. All other uses of the document are conditional upon the consent of the copyright owner. The publisher has taken technical and administrative measures to assure authenticity, security and accessibility.

According to intellectual property law the author has the right to be mentioned when his/her work is accessed as described above and to be protected against infringement.

For additional information about the Linköping University Electronic Press and its procedures for publication and for assurance of document integrity, please refer to its www home page: http://www.ep.liu.se/.

(3)

Abstract

The usage of autonomous vehicles is starting to appear in several different domains and the domain of public safety is no exception. Wallenberg Artificial Intelligence, Autonomous Systems and Software Program (WASP) has created a research arena for public safety (WARA-PS) to explore experimental features, usages, and implementation of autonomous vehicles within the domain of public safety. Collaborating in the arena are several companies, universities, and researchers. This thesis examines, in collaboration with Combitech, a

company partnered in WARA-PS, how the implementation of autonomous vehicles affects the sociotechnical system of a search and rescue operation during a drifting boat with potential castaways. This is done by creating a case together with domain experts, analyzing the sociotechnical system within the case using cognitive work analysis and then

complementing the analyses with the unmanned autonomous vehicles of WARA-PS. This thesis has shown how the WARA-PS vehicles can be implemented in the case of a drifting boat with potential castaways and how the implementation affects the sociotechnical system. Based on the analyses and opinions of domain experts’ future guidelines has been derived to further the work with sociotechnical aspects in WARA-PS.

(4)

(5)

Foreword

I would like to thank my supervisors for the many meetings and feedback. Additionally, I would like to thank Rickard for the collaboration with the case and domain knowledge. I would also like to thank Combitech for a fun and interesting opportunity to bring a

sociotechnical perspective to WARA-PS. Finally, I want to express my gratitude towards all the participants in this thesis for participating in interviews, focus groups and workshops.

(6)

Abbreviations

WASP - Wallenberg Artificial Intelligence, Autonomous Systems and Software Program WARA-PS – WASP Research Arena Public Safety

SMA - Swedish Maritime Administration JRCC - Joint Rescue Co-ordination Centre SAR – Search and Rescue

SSRS – Swedish Sea Rescue Society OSC – On Scene Coordinator

UxV – Unmanned Vehicle

UAV – Unmanned Aerial Vehicle USV – Unmanned Surface Vehicles CWA – Cognitive Work Analysis WDA – Word Domain Analysis ConTa - Control Task Analysis

(9)

(10)

1

1 Background

1.1 Introduction & Purpose

The usage of autonomous vehicles is starting to appear in several different domains and the domain of public safety is no exception. Wallenberg Artificial Intelligence, Autonomous Systems and Software Program (WASP) has created a research arena for public safety (WARA-PS) to explore experimental features, usages, and implementation of autonomous vehicles within the domain of public safety. Collaborating in the arena are several companies, universities, and researchers. The arena focuses on how autonomy can contribute to a safer and more efficient sea rescue. However, the recent advancements in the arena has been focusing on the technical developments, leaving the sociotechnical perspective behind. There is a limited understanding as to how search and rescue personnel can use and implement autonomous vehicles in their organizations and work processes. In a domain where time is scarce and safety a major factor, the development should consider the sociotechnical perspective. Otherwise the processes, allocations and interactions of implemented autonomous vehicles may result in a less safe system prone to failure, frustration, inefficiency.

As WARA-PS has mainly been technocentric, Combitech, a company and partner within WARA-PS called for a perspective including the organizations, users, and existing processes of potential future systems, which resulted in the initiation of this thesis. Hence this thesis is in collaboration with Combitech. There is a real risk that WARA-PS becomes too detached from the real working environment and situations of search and rescue personnel, resulting in features, systems and vehicles based upon the need for technological development rather than on the current problems and development opportunities within the search and rescue at sea domain. To provide a sociotechnical perspective and to bring the challenges in human autonomy to light, this thesis aims to examine the sociotechnical system which emerges during a specific case of an empty drifting boat with potential castaways. While also examining how an implementation of autonomous vehicles might look like based on the expertise of domain experts and how an implementation may influence the sociotechnical system within the described case. This is done by utilizing the tools provided by the method cognitive work analysis.

(11)

2

The purpose of this thesis is hence to contribute to WARA-PS by providing insights in the possible implementations of the WARA-PS vehicles and how the implementation may influence the preexisting sociotechnical system. As well as providing guidelines for future development to match the current issues and improvement opportunities based on the expertise of experts within the search and rescue at sea domain.

The results are presented in several different illustrations and it would be of benefit to the reader to have appendixes 8.5 and 8.9 easily accessible for a more convenient read.

1.2 WARA-PS Research Arena

Wallenberg Artificial Intelligence, Autonomous Systems and Software Program (WASP) is Sweden’s largest individual research program. WASP was started when five major Swedish universities came together to strengthen, expand, and renew the Swedish national competence in autonomous systems and artificial intelligence. WASP has created a research arena for public safety (WARA-PS) which aims to provide a realistic, large scale and industrially relevant environment for public safety at sea. The partners which partake in WARA-PS are Saab Kockums, Saab Aeronautics, Saab Surveillance, Combitech, Ericsson AB, Axis Communications, Linköping University, UMS Skeldar, Swedish Maritime Robotics Centre and the Swedish Sea Rescue Society. Together the partners integrate existing products and software, each contributing with different aspects to the research arena. The result is a research arena that consist of researchers and engineers from both industry and academia. WARA-PS research arena conducts demonstration days where technologies, vehicles and concepts are showcased between the different partners and external potential stakeholders. This thesis will examine two autonomous unmanned vehicles used in WARA-PS. These are the unmanned surface vehicle named Piraya, and a quadcopter drone. The Piraya is a small unmanned boat of 4 meters length, it is weatherproof, capable to record visuals of 360-degree radius and to navigate autonomously when instructed by waypoints. See Figure 1 which depicts the Piraya. The quadcopter drone is capable of hovering and flying. It is capable to record visuals and to autonomously avoid objects. See Figure 2 which depicts the quadcopter. Both vehicles are capable to mount several different sensors and hardware, while they are also used for heavy development of experimental features.

(12)

3 Figure 1: Piraya (Picture owned by WARA-PS, WASP.)

Figure 2: Quadcopter (Picture owned by WARA-PS, WASP)

1.3 Swedish maritime search and rescue

Swedish maritime search and rescue is coordinated by the Swedish Maritime Administration (SMA) which is a government agency responsible for maritime safety and availability. The SMA coordinates rescue operations from a rescue central located in Gothenburg. This rescue central severs as a Joint Rescue Co-ordination Centre (JRCC). JRCC has 28 employees with the necessary competence to act as a rescue leader during a rescue operation. The center is manned all hours of the day by at least four persons. SMA pursues collaboration in maritime and aeronautical Search And Rescue (SAR) (Sjöfartsverket, 2013). SMA itself has resources such as pilot boats and rescue helicopters but relies heavily on the resources found in the Swedish Sea Rescue Society (SSRS), the Swedish coast guard, the police, and municipalities emergency services across Sweden. These are primarily vessels, crew, and airborne units.

(13)

4

Whenever there is a rescue operation, the JRCC has the joint resources of these actors at their disposal (Sjöfartsverket, 2019).The rescue leader at JRCC can assign an On Scene

Coordinator (OSC) which can operate on location of the accident site together with the rescue leader at JRCC. SMA is responsible for conducting education and courses for personnel to be able to be called upon as an OSC (Sjöfartsverket, 2013).

The SSRS is involved in around 80 percent of all sea rescue in Sweden, it is financed by membership fees, donations and voluntary work. Founded in 1907, SSRS now has 72 stations and 230 rescue boats around the coast and lakes of Sweden with the goal to be on route within fifteen minutes after the alarm has been received (Sjöräddningssällskapet, n.d.).While the municipal emergency service is responsible for firefighting, rescue during traffic accidents and industrial accidents, they also have a responsibility to conduct and lead rescue operations in harbors, canals and in areas close to coast. However, in the case of an accident occurring at sea on government owned waters, SMA is responsible to conduct and lead a rescue operation (Myndigheten för samhällsskydd och beredskap, 2019).

The Swedish national Communication system for co-ordination and management,

RAdiokommunikation för Effektiv Ledning (RAKEL) is used by over 600 organizations and it is designed to be as secure as possible as well as able to cope with tough weather situations and power cuts (Myndigheten för samhällsskydd och beredskap, 2020).

1.4 Delimitations

The domain of search and rescue at sea is dynamic and ever changing. To provide practical context to a more or less intangible concept, a specific case will be created. This will allow the thesis to explore the challenges and issues in a specific context and what this could mean for an implementation of autonomous unmanned vehicles. Therefore, this thesis is limited to that case.

During WARA-PS demonstration days there may be several vehicles and different technology showcased. However, this thesis will only examine two of them, the quadcopter and the Piraya.

The two described vehicles are under constant development, examples of future features are automatic detection of castaways and dropping emergency supplies. In order to specify the case and to conduct the analysis, the features of these vehicles had to be fixed. Therefore, this thesis will be conducted on the basis that the vehicles are capable to navigate to a certain

(14)

5

location autonomously using navigation waypoints and able to relay back a visual feed of either video or pictures, as well as an audio feed.

(15)

(16)

7

2 Previous literature

This section presents previous literature to provide context and a connection to this thesis’s research questions and its results. Firstly, previous work with autonomous vehicles at sea followed by previous work in human autonomy collaboration will be presented. Previous work in integration of novel technology and sociotechnical systems are also presented here. The main points of presented previous literature is summarized at the end of this section.

2.1 Autonomous vehicles at sea

The use of unmanned vehicles (UxV) is becoming more widespread in a wide range of industries and fields. The usage of unmanned vehicles has made various appearances in crisis management and emergency services worldwide. Tanzi, Chandra, Isnard, Camara, Sebastien and Harivelo (2016) put forth one example where an Unmanned Aerial Vehicle (UAV) was used by a local company to detect feasible roads for the emergency service to traverse on after the rampaging typhoon Yolanda in Tacloban, Philippines. While the study above shows a clever usage of technology, there is a fear that remote-controlled vehicles could be more of a hassle to emergency service personnel as controlling and managing these vehicles require time, time that could be spent to save lives (Tanzi et al., 2016). Nevertheless, the level of automation in UxV are increasing to the point where they can navigate autonomously. This entails that personnel can spend less time controlling UxV and more time doing the work which they were trained to do while also reaping the potential benefits of UxVs. Thus, automation is key to apply unmanned vehicles in the emergency service domain. In order to accomplish an appropriate level of autonomy specific features must be developed which corresponds to the tasks at hand. An example of this could be to discriminate castaways from debris or being able to locate something that might not be optically visible. This requires development of algorithms and a range of appropriate sensors to be attached to the vehicle (Tanzi et al., 2016). One could assume that an UAV able to detect victims might be enough, but Tanzi et al. (2016) suggests that real world situations at sea might require an UAV to differentiate between different types of victims and SAR actors. This will enable the feature to count victims and to identify a certain type of victim in more urgent need of rescue than others, such as children. A feature like this would allow SAR teams to make more informed decision during a rescue operation at sea. Another feature is also mentioned by the authors which allows the pinpointing of mobile devices using sensors detecting electromagnetic emissions. Another example put forth by Tanzi et al. (2016) describes how an UAV can create

(17)

8

a high resolution map of a disaster struck area which can aid the decision making in emergency service teams.

As mentioned, most of the mentioned features need to have specific algorithms developed, but the vehicle in question also needs to have the appropriate hardware and sensors on board. Tanzi et al. (2016) are mentioning the need for low energy consumption and increased processing power hardware but also stresses the rapid development of hardware in the UxV domain which will significantly affect the limitations of unmanned vehicles.

Xiao, Dufek, Woodbry and Murphy (2017) present their study where UAV collaborates with an Unmanned Surface Vehicle (USV) during a Search and Rescue (SAR) operation in a mass casualty shipping incident at sea. The authors argue that a mass casualty shipping incident demands fast and comprehensive situational awareness. This situational awareness is

especially difficult to achieve at sea because of vast amount of area and the flat nature of the sea. Objects at sea are at the same evaluation and this makes depth perception more difficult than on land. In other words, victims and objects at sea will appear closer to the SAR teams than what they are in reality. Because of this it can be troublesome for an USV to navigate to castaways in distress. To solve this, the authors propose that an UAV can provide information which may guide the USV and give the SAR teams with an overview of the accident.

The idea of pairing different types in unmanned vehicles is not unique. Ramirez, Benitez, Portas and Orozco (2011) present a sea recuse system which is also based on the collaboration between an UAV and USV, where the USV utilizes the measurements and information

gathered by the UAV to calculate locations of castaways. The authors highlight the dynamic environment of the sea, and that wind and water currents may spread objects from the original incident location across a large area. With the aid of algorithms and artificial neural networks the future positions of objects and castaways can be predicted. An UAV can search in the predicted search areas and then share the information with an USV. The coordinated use of the different characteristics of these vehicles, meaning that the UAV is fast but fragile and the USV robust but slow, the time spent by the SAR team searching could be reduced.

Continuing the idea of cooperation between unmanned vehicles, Tanzi et al. (2016) presents the usage of different types of UAVs and argues that applying different types may greatly increase effectiveness in a SAR operation. The three different suggested UAVs are different in physical and algorithmic capabilities, these are the blimp, fixed wing drone and quadcopter drone. The blimp has a high level of autonomy and stability which can provide

(18)

9

communication infrastructure. The fixed wing drone can fly at a high altitude which can provide a large vision angle than drones at lower altitude. Lastly the quadcopter drone is able to hover at a fixed location and fly in all directions. This makes the quadcopter ideal for examining incident sites closely. With that said, UxV are not without drawbacks, in some weather conditions it may be impossible for UAVs to operate.

As unmanned vehicles become less remote controlled and more autonomous, they can free humans from the burden of constantly controlling and monitoring them (Tanzi et al., 2016). However, a high level of autonomy does not necessarily imply that instructions from outside the system will seize. Quite the opposite as Tanzi et al. (2016) suggests that unmanned vehicles should, while autonomous, be able to receive new instructions remotely in order to correspond to the current priority of the SAR team.

2.2 Human autonomy cooperation

Johnson et al. (2012) argues that governments, and the conventional wisdom praise autonomy as a given shortcut to an improvement in performance. However, the idea of this shortcut is superficial and misleading for several reasons. Autonomy in systems with complex joint activity between humans, software and other agents can result in degraded performance due to neglection of important conditions that affect the effectiveness of the management of

interdependence among humans, software, and other agents. Research is ongoing to create more independent agents and robots that need no supervision. But these autonomous agents and robots will need to collaborate in the future, they will not only need to do tasks for people, but they will need to conduct task with people and other agents. Therefore, they must be capable of interdependent joint activity when needed, which requires need for support of task handoffs and joint participation in task which in turn requires continuous and close interaction. Because of this required capability for interdependent joint activity impose demands on autonomy algorithms at a deep level, system design must incorporate the idea of interdependent joint activity from the beginning (Johnson et al., 2012).

This conventional idea that automation is a shortcut to reduced manpower needs, less training for operators and a decrease in errors is derived from an attitude that autonomous systems are be able to surpass human reasoning abilities and a rampart enthusiasm for automation. In everyday conversation, automation is often discussed as the inevitable substitute for human counterparts. This is because there is an assumption that human and machines abilities can be compared when in reality, they are different and that abilities of a machine is often

(19)

10

complementary. Indeed, it is not a surprise that machines outperform humans in certain abilities, such as data storage and processing. The machines excel humans in these particular abilities but if something goes wrong, intervention by a human is needed as the machine does not possess the human abilities of adaptability, flexibility and human reasoning. In this sense, the human abilities compensate for the lack of similar abilities in the machine, and vice versa. When automation is implemented, human operators will not have fewer things to do, but new things to do. The new tasks will revolve around monitoring and coordinating with the

technology (Hoffman, Sarter, Johnson, & Hawley, 2018).

This infers that the human operator must track, comprehend, and predict machine actions and intentions, which requires a lot of attention from the human operator. In a successful

implementation of automation, the human operator can perform task more efficient or better than they would have had without the automation. However, far from all implementations are successful, there are several examples where the operators do not understand the systems abilities and limitations, creating a dangerous reliance on automation that can degrade human cognitive and coordination skills. These new tasks require high proficiency among the

operators, meaning that they must be trained as there will always be new problems and unanticipated problems in systems, no matter how advanced the systems is. An understanding and an awareness of these issues can greatly improve implementation of automation

(Hoffman et al., 2018).

An example is put forth by Endsley (2017) where the operators find the systems difficult to understand once there is a problem. A commercial airplane crashed off the cost of Brazil because of a malfunctioning sensors which directly affected the automation in the cockpit automation. As the pilots were unable to find the underlying problem and not understanding why the automation behaved as it did, the airplane crashed resulting in the death of everyone onboard. This highlights the issue that even though the automation itself functions properly most of the time, the ability to resume manual control is essential.

Hoffman et al. (2018) also showcase an example where the role of military crew had changed from controlling the systems to observing the systems routines. The monitoring of the system was difficult as the system was problematic to understand due to poor system design and a lack of transparency as well as training. This resulted in a missile being launched targeting a friendly vessel. Neither did the crew understand how anomalies and unanticipated

(20)

11

mental models of the system needed to adapt during an evolving mission. Indeed, the

promised benefits of automation does not always correlate with the reality of implementation. Sarter, Woods & Billings (1997) present a set of promises of automation and their

implications in the actual real world when implemented. Tabel 1 presents this set, edited and expanded by Hoffman et al. (2018).

Tabel 1: The promise and reality of automation

The promise The reality

Automation as substitution, better results in the same system.

Transforms the practices and operators. Old and familiar routines and features are altered.

Automation frees up humans by offloading work to machines.

Creates new kinds of cognitive work, often at wrong times.

Automation frees up limited attention by focusing humans on the correct answer.

Creates more threads to track, making it harder for people to remain aware of and integrate all the activities and changes around them.

Less training and human knowledge are required.

New knowledge and skill demands are imposed on humans, who may no longer have enough context to make decisions because they have been left out of the loop. Machine agent will function autonomously.

Same feedback to humans will be required.

Teamplay with people and other agents is critical to success. New types of feedback, including effective attention guidance are needed to support people’s new roles. Automation results in increased and

beneficial flexibility.

Automation results in an explosion of features, options and modes which creates new demands, types of errors and paths toward failure.

Human errors are reduced. Both autonomous agents and people are fallible, new problems are associated with human-agent coordination breakdowns, agents can obscure information necessary for human decision making.

Indeed, the implementation of automation can sometimes create complex changes in the real world and a need for operators and autonomous agents to collaborate. As stated above, new problems can be associated with human-agent coordination breakdowns. In an experiment which examined human-agent teamwork, Johnson et al. (2012) found a statistical significance when comparing human acceptance, or user preference, in different ways of working with a virtual agent. In the study it was shown that humans experienced that a certain way of

(21)

12

collaborating outperformed others and that way of collaborating was also more accepted than the others. The authors believe that this is because the human holds the overall plan and the majority of the context which allows for a greater degree of creativity. When asked as to what characterized this way of collaboration as superior, the participants highlighted that the information was shared between them and the agent, the ability to anticipate and to predict, and the feeling of control while still preserving autonomy. This might infer that transparency and control may be more important than autonomy in interdependent joint activities.

This way of collaboration was also the highest ranking when user performance was measured. In the simple abstract domain where the study was conducted clearly showed that there are certain characteristics in the capabilities of autonomous systems which allows for more effective interdependent joint activity. These results do not support the conventional idea that increased performance is derived from increased autonomy, quite the contrary. The results highlight the practical implications which underscores the importance of considering interdependence when designing autonomous agents.

2.3 Integration of new technology in organizations

As technology advances, it is inevitable that novel technology is introduced and integrated in already existing work processes or organizations. The main goals of introducing new

technology is to improve organizations whether it has to do with increased efficiency, safety or earnings. However, there are fundamental challenges in successfully integrating novel technologies in preexisting organizations. Indeed, the majority of large technology integrations projects are either challenged or failed while only a small percentage is

considered successful. The Standish group conducts surveys in order to measure the success of software projects. The report states that in larger projects the success rate was only 8% while 43% were deemed to have failed (The Standish Group International, 2015). This means that the novel systems are either rejected completely by the organization or that it is partly used in ways that the system was not intended to be used and in combination with old systems that better supports the work processes than the new system. This is unfortunate not only because of the loss of potential novelty in the organization but also due to the large economic loss (Eason, 2017).

There are several factors that affect whether the integration is successful or not. One being a discrepancy between the implementers and the users, more precisely this occurs when the information that are presented by implementers are largely positive descriptive information

(22)

13

about the novel technology. This could be due to a strong belief in the perceived benefits of the technology resulting in an unintentionally disregard to operational concerns and complex operational issues. The operational concerns and issues can easily become a minor focus to the implementers.As a consequence of this, users are positively biased to the novel

technology and will encounter negative surprises such as operational difficulties and unanticipated costs (Griffith & Northcraft, 1996).

Commonly, it seems that these negative surprises and organizational consequences tend only to be discovered during implementation alone and are managed in a spontaneous and

ineffective manner. The failure is often the result of an ambitious objective to alter processes within the organization in a fundamental way and expecting the organization to adapt to the new ways of working. Typically, these are long-term and large projects that have

commissioned a technology supplier to design the system which goal is to support the new work processes. These large-scale technical integration projects are often technocentric meaning that it is assumed that organizational structures, processes and practices will fall in line with the newly implemented technical system (Eason, 2017).

The idea of integrating and implementing technology in ways that supports organizational structures, attempts to identify roles for humans and question the social aspects of the system design is not new. Clegg (1988) introduces the idea of “Appropriate Technology” and

explains how companies commonly integrates new technology. New technologies are

commonly integrated without any clear planning or strategic planning. It is also very common for the projects to be technically driven especially in high-tech companies. In less

sophisticated companies’ integration of new technology is simply integrated with the least effort possible. In technically driven projects predictability of the given system is often in focus. There is a common view of humans as unpredictable and hence a source of unexpected errors, and it is therefore not unusual that system functions are allocated to machines rather than humans. In general, limited attention is paid to human and organizational aspects during integration of new technology. Even though the paper Clegg (1988) was published 30 years ago, it would seem that the issues when integrating new technology still remains. Clegg (1988) explain that the idea of “Appropriate Technology” is characterized by that the users understand the technology, that the technology is flexible and adaptable to different

circumstances as well as supporting organizational simplicity. In order to achieve this

organizations should open up the design process and include end-users, managers, supervisors and other actors who will be associated with the system. The design work should also consist

(23)

14

of interdisciplinary teams and the teams should develop specific scenarios that will

systematically aid to consider decisions regarding allocation of function, extent of automation and organizational structures.

Similarly, Karlsson, Taylor & Taylor (2010) suggest several factors which have an effect on the success of integrating a novel technology into an existing organization. These factors are the complexity of the technology, the experience of novel technology within the organization and the experience of implementation such technology, the adaptability to existing processes and the user understanding of the technology. The information which is given to the actors within the system is also important, it must allow the actors to understand how the system not only affects them and their role, but also other actors and the organization as a whole (Pan, Pan, & Devadoss, 2008).

Even though there are ideas of including social aspects in the integration of new technologies, the technology-driven projects in techno-centric mindsets remain (Man, 2019). To include social aspects in the implementation of new technology one must have a sociotechnical perspective. The sociotechnical approach is characterized by the inclusion of human, social, organizational and technical factors (Baxter & Sommerville, 2011).

2.4 Sociotechnical systems

The idea of sociotechnical systems can be traced back to the introduction of coal mining machinery when the social and technological aspects of the workplace were clearly shown to be interconnected. The implementation of technology underlined the need to consider social aspects as well as technological aspects when attempting to implement a new technology in a preexisting organization. Organizations are in many ways complex systems with several interconnected and interdependent parts. This means that all parts should be considered when implementing a change to one specific part as the change may affect or require change in other parts of the organization. Failure in doing so may affect the effectiveness of the organization as a whole (Davis, Challenger, Jayewardene, & Clegg, 2014).

This has led to the creation of sociotechnical design methods which include human, social, organizational, and technical factors. The usual outcome in projects where these methods are applied is a clearer understanding for how human, social, organizational, and technical structures are used. Although this is known, and sociotechnical aspects considered important, the methods for sociotechnical design are seldom the chosen approach. The difficulties in

(24)

15

using the methods and the disconnect between technical engineering and individual interaction design with the systems could be an explanation for the lack of using an sociotechnical approach (Baxter & Sommerville, 2011). The sociotechnical approach is characterized by designing the system with the consideration that both social and technical factors influence the functionality, usage, and effectiveness of systems. The failure to include social and organizational aspects when implementing and designing new systems could directly increase the risk that the given system will not meet the expected goals. As stated in the previously mentioned Chaos Report, most IT-projects are deemed as failures (The Standish Group International, 2015). These projects often have no issues in meeting their technical requirements but are considered to be a failure due to the lack of support for the real work in the organization. The techno-centric approach fails to correctly map the

interdependencies between organization, people doing work and the already implemented technical systems (Baxter & Sommerville, 2011).

Several different research communities have addressed sociotechnical issues in relation to specification, design, and implementation of computer-based systems. Information systems, work analysis and cognitive system engineering are among the communities which have had mutual awareness of sociotechnical issues. However, the term sociotechnical systems are broadly used to describe many complex systems and the methods are commonly more similar to philosophies than to the more traditional methods in systems engineering. Even though the usage of the term is broad, Baxter & Sommerville (2011) presents five key characteristics of sociotechnical systems, these are:

• Systems should have interdependent parts.

• Systems should adapt to and pursue goals in external environments.

• Systems have an internal environment comprising separate but interdependent technical and social subsystems.

• System can achieve system goals by more than one means.

• System performance relies on the joint optimization of the technical and social subsystems, focusing on one of these systems to the exclusion of the other is likely to result in degraded system performance and utility.

Baxter & Sommerville (2011) explains that during the 1980s, methods for sociotechnical systems were used as a mean to keep workers during a time of labor shortages, making the work in the industries more humanistic. However, during the 1990s sociotechnical ideas was

(25)

16

put aside as re-engineering and lean production techniques dominated, only to get a revival of interest during the 21st century. Sociotechnical ideas and approaches may not always be clearly marked as such even though the characteristics appear in ethnographic and

participatory design methods. As of today, there is a wide range of different approaches and methods in social-technical system design that can inform the development of sociotechnical systems. Below is a few mentioned by Baxter & Sommerville (2011):

• Cognitive Work Analysis (CWA) is a method which was developed to analyze complex sociotechnical systems. It predicts what a system could do, instead of explaining how work should be or is done.

• Ethnographic workplace analysis aims to inform the design of sociotechnical systems from situated nature of action. The method highlights operational issues and how functionality and use of the system is affected by workarounds and dynamic process modifications.

• Contextual design concentrate on the designer’s comprehension of how the customer conducts work on which the design is based. It is assumed that the system inherently embodies a certain way of working, which dictates how the system will be used and structured. The activities highlighted by this method is predominantly front-end design and customer work.

• Cognitive System Engineering analyze organizational issues and aims to provide practical support for system design. Observation is used to analyze work in context and abstraction is used to identify patterns in the observation of work settings and situations. This results in a greater understanding of sources of expertise and failure. • Human-centered design deals with the principle of using the explicit understanding of

users, their task and the environment as a base for the system design. It aims to take social and cultural factors, such as working practices and organizational structure into consideration.

These methods all try to address real problems in complex sociotechnical systems. However, according to Baxter & Somerville (2011) these methods are yet to have had any significant impact on industrial practice. Some of the main reasons behind this could be the lack of consistent terminology, different levels of abstraction, conflicting values, fieldwork issues and the lack of a success criteria. The inconsistency in terminology creates confusion as to what is meant by sociotechnical systems. Many different fields have adopted the term with their own interpretation. Similarly, the levels of abstraction are varied when describing and analyzing

(26)

17

sociotechnical systems. The designer often has an immense commitment to improving the quality of working life and job satisfaction in the system, assuming that system efficiency and an increase in productivity will follow. While managers on the other hand view sociotechnical principles are the means to achieve the company’s objectives. When it comes to fieldwork, sociotechnical system design involve stakeholders and it can be difficult to select which type of stakeholders to focus on and to what extent one should involve them. There is also a limited evaluation of the efficacy of sociotechnical system design methods. There is no method that is specific enough to allow for empirical testing. There is also a huge difference in establishing evaluation criteria for technical and social elements within the system. The technical criteria or requirements could be easily established and evaluated while the social elements cannot (Baxter & Sommerville, 2011).

Indeed, the sociotechnical approach to system design is not without challenges but Baxter & Sommerville (2011) present some practices which could provide support. One being to sensitize different actors involved in the system to sociotechnical considerations as well as raising potential issues that could be a conflict of interest between stakeholders. Another being to clarify that any type of observation or ethnographic work is to gain a deeper

understanding of the conducted work, not to assess or audit what is done and reporting it to a potential manager. Systems engineers could also be brought to the workplace where actual work is done, to reveal the complexity and difficulties of work. This could also be achieved by having so called war stories or workplace vignettes which are story elements that

highlights the complex sociotechnical work in the system.

Furthermore Carroll (1995) points out how there long has been a missing strategy to involve user perspective and having active participants in the design processes within computer and information science. They suggest that scenario-based design can offer a strategy to

development that is more grounded in the contextual reality in which the end system will be used. End users are described as an integral part of developing scenarios that correctly represent real world work contexts. With that said, the construction of cases allows for the participants to think with external representations. Kirsh (2010) argues that external

representations allows people to make sense of situations as external representations lowers the burden of the cognitive workload, they create a structure that serve as a sharable object of though and allows for persistent referents. External representations allow for more powerful thinking, they can have people reach ideas and thought that they would not do without them.

(27)

18

Mentioning external representations Tohidi, Buxton, Baecker & Sellen (2006) present their study where they have compared the outcome of presenting one solution with presenting several solutions. They found that a single solution receives higher ratings and less critique then when presented along with three similar designs. This imply that participants give more appropriate critique when presented with multiple design. An article by Read, Salmon, Lenné & Stanton (2015) also highlights that scenarios or cases contribute towards design planning. Cases can be developed to assist the design participants when analyzing the context and domain problems, meaning it can be of use when iterating, evaluating, and re-designing. The scenarios may also serve useful when communicating with different project stakeholders and to avoid the usage fashionable or new technologies instead of conducting design driven solutions.

2.5 Summary of previous literature

As Unmanned Vehicles (UxVs) are becoming an increasingly popular tool within public safety, they have the potential to aid emergency service in detecting feasible routes and pinpointing personal electronic devices as well as providing an overview of accident areas. However, there exists a concern with operating, monitoring, and handling these UxVs as personnel trained to save lives now has these additional tasks. The fear is that this combination may steal resources and time which should be used elsewhere as lifesaving operations usually only can be conducted in a limited time window. Autonomy is thus key in order to apply UxVs to the domain of public safety. This entails that appropriate features must be developed to correspond to the tasks at hand.

However, autonomy is often associated with immediate improvement in performance. This conventional idea is superficial and misleading as autonomy in complex systems with joint activities can actively degrade performance if the other interdependent parts of the system is neglected. In fact, the development in automated systems are creating more joint activity between automation and man, meaning that automation must collaborate with human or non-human agents. This imposes demand on the deep levels of autonomy algorithms as the idea of interdependent joint activity must be considered in the beginning of development. Yet,

automation is seen as a shortcut to reduced manpower needs, less training for operators and a decrease in errors as the unpredictable human actions are removed. Automation does not mean that human operators will have fewer things to do, but new things to do and these will revolve around monitoring and coordinating with the automation. There are several more assumptions when implementing automation, such as errors being reduced but automation

(28)

19

results in an explosion of features, options and modes which all create new demands, types of errors and paths toward failures. It is a challenge to implement a novel technology, not just automation, in an organization. The goal of most novel technology implementations is to improve either efficiency, safety, or earnings. However, most large technology integrations projects are either challenged or deemed as failures whereas only as small percentage is considered successful. The factors behind this are many but the major ones are discrepancy between implementers and users, heavily technocentric projects and the idea that fewer human parts translates into less possible errors.

Most of the challenges in implementing a novel technology and the assumptions about automation are born from the neglect of thinking of the system from a sociotechnical perspective with several actors, actions, and interdependently connected parts. The idea of sociotechnical systems is not new and there exists several sociotechnical design methods which include human, social, organizational, and technical factors. Although the benefits of considering sociotechnical aspects, the methods to do so are seldom the chosen approach. The difficulties in using the methods, the inconsistency in terminology, the interdisciplinary nature and the disconnections between technical engineering and individual interaction design with the system are some of the suggestions as to why this is. A majority of large novel technology implementation projects fail. The failures as seldom due to failure to reach the technical requirements but the projects fail due to a lack of understanding and support for the real work conducted in the organization. The sociotechnical approach and methods are not without challenges but there are some practices which could provide support. Some practices are to sensitize different actors within the system to sociotechnical considerations, raise potential issues that could be a conflict of interest and having the implementors brought to the real-world workplace in order to reveal complexity and difficulties of work. Another way of doing this is by so called war stories, cases or workplace vignettes which highlight the complex sociotechnical aspects in the system.

2.6 Research questions

Given the previous literature and the apparent issues within the field, three research questions have been developed.

• RQ1: In what tasks and situations can WARA-PS autonomous vehicles be

implemented during a search and rescue operation at sea in the case of a drifting boat with potential castaways?

(29)

20

• RQ2: What values and processes in the case of a drifting boat with potential castaways are influenced by the implementation of WARA-PS autonomous vehicles?

• RQ3: Are there any guidelines to consider during the development in WARA-PS autonomous vehicles to respond to the needs and challenges of the Swedish search and rescue domain more correctly?

RQ1: Aims to provide a tangible understanding of how WARA-PS autonomous vehicles can be implemented in an already complex sociotechnical system. The intent is to create a bridge between the complex real-world work environment and the actors developing the autonomous vehicles. RQ2: Aims to highlight the potential influence an implementation of autonomous has on the described sociotechnical system. RQ1 and RQ2 provides a snapshot of how the implementation, and the consequences there of, could look like. RQ3 aims to provide a direction for future development of the WARA-PS project.

(30)

21

3 Method

To provide answers to the stated research questions several steps where conducted. See Figure 3 which describes the different steps and methodology in this thesis. Starting from the left, the construction of a case provided a verified case and insight in SAR personnel’s attitude toward novel technology. These founded the basis for the conducted analyses, CWA, and thematic analysis. The outcomes served as a base when the WARA-PS vehicles were incorporated, and when conducting the concluding workshop. The concluding workshop resulted in final

analyses featuring the WARA-PS vehicles and a direction of development which were used to answer the research questions of this thesis.

Figure 3: Steps of the thesis

3.1 Collection of data

Current and relevant literature in UxV usage in SAR operations were reviewed to gather knowledge about the current situation and advancements of UxV-use in SAR operations across the globe. This was done as a larger knowledge base before meeting with domain experts allows for more advanced understanding and more informed questions to be asked. In this thesis interviews and focus groups were used to collect qualitative data. As during all qualitative data collection methods, the focus was to retain a large amount of data in a flexible way. Qualitative interviews and focus groups are different from more structured interviews as the interest lies in the thoughts and opinions of the interviewee. Unlike structured or

interviews aiming to collect quantitative data, this approach entails follow-up questions and interviews about the certain topic and the ambition to collect as detailed answers as possible

(31)

22

(Bryman, 2011). Some of the collected data will be analyzed using thematic analysis. The qualitative data will be carefully examined to code and group the data which will uncover themes.

3.2 Analysis method: Cognitive Work Analysis

The term complexity can be used in various contexts and have different meanings. One way of thinking about complexity in systems is by considering the following aspects of a system:

• Dynamism of the system: How dynamic is the system, to what extent can the nature of the problems transform over time and can the system change states without

intervention?

• Uncertainty: How predictable are the future states and how complete and correct is the available data?

• Sub-parts, variables, and interconnections: To what extent are different parts of a system interconnected?

• Risk: What is at stake?

(Jenkins, Stanton, Salmon, & Walker, 2009)

Indeed, the domain of SAR at sea would prove to have high levels of dynamism, uncertainty, interconnections and risk, meaning that the system is highly complex. Several different methods were considered to analyze the gathered data, such as Hierarchical task analysis (HTA) as it is one of the most known analysis technique among human factors practitioners (Stanton, 2006). However, as HTA describes the chosen systems in a normatively manner in terms of goals, plans and activities it was disregarded as the level of complexity in the SAR at sea has proven to be very high. Meaning that one exact process for every emergency

operation does not exist as it is highly dependent on the current available resources, weather, and nature of emergency at hand. This is true even if a specific case or scenario is examined as an exact sequence of events is unlikely to occur twice.

In the end, Cognitive Work Analysis (CWA) was chosen as the method to analyze the

gathered data. CWA is aimed towards analyzing complex sociotechnical systems as it focuses on the limitations in the possible actions which can performed in the given system. The limitation allows for a description of the possibilities in an action and this focus allows the analysis to be adapted for unanticipated actions (Naikar, 2017). This is well suited when thinking about the nature of the SAR at sea, unanticipated actions are likely to occur. CWA

(32)

23

typically consists of five different phases, each one examining different types of limitations. These five phases are as described by Jenkis et al. (2009):

1. Work Domain Analysis (WDA)

Work domain analysis aims to define the task environment and is conducted at the functional level. WDA can represent information in a systematic manner by using an abstraction hierarchy which allows the chosen systems to be examined at five different levels. The highest level is the functional purpose of the system as a whole, second one being the values priorities and measures, third level being the purpose related functions, fourth the physical functions and lastly the fifth level lists the physical objects within the system. Depending on what system is being analyzed, the analysis can start from level five or one. When analyzing an already existing system, one can start with the overall purpose and think about how that purpose is achieved throughout the different levels. When new technologies are introduced, one may start from the lowest level and insert the new physical objects to examine what they influence.

2. Control Task Analysis (ConTA)

Control task analysis is used to understand the tasks at hand, based on the previously conducted WDA. ConTA specifies what needs to be done independently of how or by whom. To assist this, a contextual activity template was developed which can represent work characterized by work situations and work functions. The work situations are usually specific locations and the work functions are usually activities which are based from the level of purpose related functions of the abstraction hierarchy during WDA. In the contextual activity template, the work functions are along the y-axis and the work

situations along the x-axis. This creates a table which allows for the relationship between the two axes to be examined. The relationship is represented by circles indicating that the work functions within the bar typically occurs during certain situations, while dotted boxes represent where it is physically possible for the work functions to occur. 3. Strategies Analysis

The third analysis aims to examine the different strategies adopted during a particular situation, as different agents perform different tasks in different ways. This is because a different strategy may be chosen given a high or low workload at the given time. There are several ways to examine these different strategies, however a simplified flow map is

(33)

24

supported by most practitioners. This flow map describes a start state, possible strategies, and an end state. The contextual activity template can be used here to show in what situation and functions the different strategies can be considered.

4. Social Organization and Cooperation Analysis (SOCA)

SOCA allows the modelling of the allocation of function of available resources. Actors are usually mapped onto the contextual activity template to show where they have an influence on the systems, allowing an overview of which actors are capable of what. However, there is no assessment over what actors might be the optimal for performing a certain function, the focus is solely on capability. When mapped onto the contextual activity template, this analysis allows an overview of workload and how it is distributed within the team given the considered work situations.

5. Worker Competencies Analysis

The last and final analysis aims to address the limitations of actor behavior during

different situations. It examines the behaviors required to complete the tasks at hand in the given system. This is typically modelled using the Skills, Rules and Knowledge

taxonomy.

Since CWA first was developed in 1986, several additions and different usages of the

methodology has occurred. The contextual activity template was an addition to ConTA which was developed by Naikar, Moylan & Pearce (2006).According to Read et al. (2015) CWA has proven to be a tool which can support the design of sociotechnical systems, but it has been difficult to translate the outputs of CWA into design guidance. In their study they attempt to uncover what current CWA-based design practices aligns with sociotechnical systems theory. Their results show that the CWA outputs, especially the abstraction hierarchy from WDA support sociotechnical process principles. The authors have used the abstraction hierarchy and it’s five levels within to define design requirements and evaluations criteria, and to identify to tools for a CWA – Sociotechnical system design toolkit. Using the abstraction hierarchy in WDA they identify different needs for the toolkit at each level, i.e. the toolkit should allow for an evaluation criteria at the values & priority level and be flexible in the choice of proposed and current tools at the lowest level, physical objects. The authors end their notion by proposing that future applications of CWA may change as sociotechnical system theory thinking may alter the approach of CWA. Another paper by Rauffet, Chauvin, Morel & Berruet (2015) suggests that CWA is yet to address a dynamic function allocation claiming

(34)

25

that they are an crucial issue in complex sociotechnical systems, they provide additions to existing CWA tools in order to detect conflicts in allocation of work in order to provide the best configuration for the particular work. CWA has also been highlighted as a useful technique to inform early phases of design, which can improve the system design and its resilience from the core (Jenkins et al., 2009).

Indeed, there are several different perspectives on how to apply CWA and how to use the outputs of the different analyses within the methodology. Given the research questions of this paper and its scope, the work will focus on the WDA, ConTA and SOCA phases in CWA. WDA will create an understanding for how the overall purpose relates to various values, processes, and the physical objects during a SAR operation at sea. The ConTA will create an insight in the work functions and their relation to work situation using a contextual activity template. To complement the ConTA and the contextual activity template, SOCA will be conducted to provide knowledge to the limitations of the different actors and reveal which actor is capable to what work function during a work situation.

3.3 Participants

The participants were all actors working within public safety, where half of the participants were specialized in SAR at sea. The actors came from SSRS, JRCC, municipal emergency services and academia. Most of the participants were highly ranked and experienced personnel within emergency services or SAR operations. The participants and their contributing context are presented in Tabel 2.

Tabel 2: Participants

Participant Organization Contribution Occasion

#1 Academia. Domain knowledge

Human Factors Maritime. Individual interview. #2 SSRS. Domain knowledge Swedish SAR. Individual interview. #3 Municipal emergency service. Domain knowledge Swedish SAR. Creation of case. Attitude to new technology. Verification of case. Individual interview. Focus group #1. E-mail. #4 Municipal emergency service.

(35)

26 Attitude to new technology. #5 Municipal emergency service. Domain knowledge Swedish SAR. Attitude to new technology. Individual interview. #6 Municipal emergency service. Creation of case. Attitude to new technology. Focus group #2. #7 Municipal emergency service. Creation of case. Attitude to new technology. Focus group #2. #8 SSRS. Creation of case. Attitude to new technology. Verification of case. Concluding workshop. Individual interview. Individual interview. Digital workshop. #9 SSRS. Creation of case Attitude to new technology. Individual interview. #10 JRCC. Creation of case Attitude to new technology. Verification of case. Concluding workshop. Individual interview. E-mail. Digital workshop.

3.4 Procedure

This section explains the procedure of this thesis. See Figure 3 which can provide an overview of the procedure, methods, and outcomes.

3.4.1 The construction of a case

To gain more knowledge about UxV usage in SAR operations at sea peer-reviewed material was reviewed and domain experts interviewed. This allowed for a greater understanding of the current challenges and advancements in human factors related to maritime as well as greater understanding for how SAR at sea is done and the related challenges. A case was needed to establish this thesis in the practicality of work of SAR at sea. The aim was to clearly base the case and its course on the experience and expertise of SAR personnel in

(36)

27

Sweden. To achieve this focus groups and interviews were conducted. The first focus group was conducted with participant #3 and #4. To get a wide a range of missions with different challenges, the participants were told to describe what they would deem as an easy, a moderate and very difficult rescue mission at sea. The course of the missions and relevant actors were described and sketched on a whiteboard. Finally, the participants were asked about their overall attitude towards novel technology in their domain, how it usually surfaces and what influence they have over the development and implementation. The elapsed time was about 2.5 hours. Following focus group #1 an interview with participant #5 was conducted about the attitude to novel technology, the time elapsed was about 20 minutes. Focus group #2 was conducted with participant #6 and #7. The same setup as during focus group #1 was used. However, the participants were told about the outcome of the previous focus group, meaning that they were made aware of what kind of rescue missions the previous groups had described. The attitude towards novel technology of the previous group were not discussed. The elapsed time was about 1.5 hours.

Following the focus group an extensive interview was conducted with participant #8 with the same setup as the focus group. However, at this point it was clear that a certain case was more relevant and more common than the others explored. Therefore, there was emphasis on the details of that case, even though other cases were briefly discussed as well. During the final part of the interview novel technology in the SAR domain was discussed, the total elapsed time was about 2 hours. Another interview was conducted with #9, following the same setup just described, however during this interview the personnel received an alarm on which the interviewer could accompany, the reality of the situation surprisingly corresponded to the selected case. This interview lasted about 1 hour.

Finally, an interview with #10 was conducted. They were asked about the chosen case, and their role and perspective in the evolving course of the case. This interviewee had a

substantial role in the case, as they are the primarily rescue leader in the chosen case. This interview lasted about 45 minutes.

3.4.2 Verification of case

The aim was to base the case in the experience and expertise of real-world personnel working in SAR at sea in Sweden. Therefore, it was crucial to verify that the case was grounded in reality. To do this, participants #10, #3 and #8 were asked to verify whether they agreed with the described case in detail or not, or if something needed to be added or altered. This was

(37)

28

done over the phone or via email. This resulted in changes and an iteration of the case to better suit the praxis of the SAR work at sea in the real world.

3.4.3 CWA & incorporation of UxV

When the case was verified and altered, the analysis using the WDA, ConTA and SOCA of CWA was initiated. Firstly, the case was analyzed using the abstraction hierarchy of WDA. As described, the abstraction hierarchy stated the overall purpose of the system and was then broken down along the five levels of the abstraction hierarch. Secondly, ConTA was

conducted with the aid of a contextual activity template which listed work function in relation to work situations. The outcome of ConTA was a contextual activity template describing in what work situation work functions could physically and will typically occur. Thirdly SOCA was conducted to uncover what actors that were capable of doing what. This was done by using the previously created contextual activity template. This outcome complemented the contextual activity template to not only represent different work functions and situations but also which actor that is physically capable of conducting the work functions during a work situation. These composed the initial results of the analysis representing the current state of the system during a SAR operation in the created case.

The thematic analysis was conducted in order to gain more insight in how the actors thought about new technology and unmanned vehicles and how these can be incorporated and what the implications may be. The revised case, the thematic analysis and the acquired domain knowledge were used to create suggestions as to how an incorporation of unmanned vehicles could manifest in CWA. One abstraction hierarchy was complemented with one quadcopter and Piraya on the lowest level, physical objects, to see what functions and values that were influenced by their implementation. Two alternatives of ConTA combined with SOCA using the contextual activity template was created. This abstraction hierarchy and the two

alternative contextual activity templates was created to serve as a starting point of discussion during the concluding workshop.

3.4.4 Concluding Workshop

To create finalized versions of the WDA, ConTA and SOCA, a concluding workshop was conducted. The participants partaking in the concluding workshop were participant #8 and #10. Due to the restrictions and recommendations by the public health agency of Sweden at the time, the workshop had to be conducted digitally. The participants were invited by email to connect to a digital room. They were notified that the session was recorded and that it would be deleted when the analysis had been completed. The participants were briefed on

(38)

29

how work has proceeded since they last were interviewed, and the purpose and research questions were explained. For the participants to familiarize with the concepts, the initial analyses were shown while an explanation of CWA, WDA, ConTA and SOCA was given. The complemented abstraction hierarchy was shown first, the interdependencies, functions and values affect by the incorporation of the quadcopter and Piraya was highlighted. The participants were asked whether they agreed on how the two unmanned vehicles influenced the system and if there was something missing. In order to ease the comprehension of the hierarchy, the five levels were divided and only two to three levels and their

interdependencies were shown at a time. The participant’s opinions about the abstraction hierarchy was noted and as discussion came to a closure, the workshop moved on to discuss the other analyses. The ConTA and SOCA was presented and combined in a contextual activity template, but to highlight the differences between the two alternatives only one work function was shown at the time. To make it clear that neither alternative was optimal or in any way correct, there was also a blank field, which were explained to be the one the participants would create during the concluding workshop. See Figure 4 which depicts the work function “Safe route planning” and the two alternative analyses and one final blank alternative.

Figure 4: Template used during concluding workshop

When the blank field had been filled, the workshop moved on to the next work function. When all the work functions had been reviewed and finalized, the participants were asked if they had anything to add. To conclude the workshop the participants were thanked before the session ended.

(39)

30

The data collected during the concluding workshop were then finalized, refined, and used to produce the final versions of the analyses. The final analyses were in turn used to answer the stated research questions of this thesis.

3.5 Research ethics guidelines

The four main guidelines in research ethics as described by Vetenskapsrådet (2002) has been followed in this thesis. These are in short, to let the participants know that their participation is voluntary, to have the participants sign a consent form stating their rights and the purpose of the study, to anonymize data in such a way that the results cannot be connected to their person and finally that their personal information will not be used for commercial or non-scientific purposes.

The autonomous crewmate : A sociotechnical perspective to implementation of autonomous vehicles in sea rescue