Finding differences in perspectives between designers and engineers to develop trustworthy AI for autonomous cars ↔ Exploring Ethical Considerations in AI-driven Autonomous Vehicles: Balancing Safety and Privacy

EARLIER¶20-24

3 What is the difference in perspective between designers and engineers with regard to ethics when developing autonomous vehicles? What strategies can be implemented to bridge the gap in perspectives between designers and engineers regarding the development of autonomous cars? To avoid confusion, it is crucial to define the terms engineer and designer in the context of our study. According to Charisi et al., (2017), intelligent autonomous systems involve three parties: designers (including developers and engineers), end-users (owners or customers), and various government and trade regulatory bodies and insurance agents. Our study used these parties as a guideline but delved deeper into the designer category by distinguishing between designers and engineers of AI systems. Specifically, we use the term engineer to refer to individuals who specialize in back-end development, such as AI coders or researchers focused on the technical aspects of AI. In contrast, designers are more oriented towards front-end development and user-centered design in the context of UX-researcher, UX/UI-designer, while researchers within these fields are considered professors specialized in system science and informatics. 4 2.0 Theory 2.1 Trust & disrupting the business model The traditional business model of selling cars is being disrupted by new mobility solutions. One of the most innovative companies in this space is Tesla Motors, which offers supplementary services through a modular approach. For instance, the Model S, Tesla's flagship car, is designed as a platform that can accommodate additional services. These include autonomous driving capabilities that can be uploaded and updated over the air for a fee. To make this possible, the company must develop cars with the services in mind and manage the development process accordingly (Mahut et al., 2017). In short, this means that car manufacturers are moving towards selling cars as a service, with smart features instead of selling just a vehicle with the only purpose to transport the driver from A to B. A survey conducted in Germany 2015 revealed that user acceptance of autonomous driving is low. Nearly half of the 1003 car buyers surveyed were skeptical, with only 5% preferring fully autonomous driving, approximately 49% preferred traditional driving and 43% preferred semi-autonomous driving (Hengstler et al., 2016). Building trust in autonomous driving is crucial for reducing perceived risk and encouraging the adoption of such technologies. At its core, trust is a fragile and evolving phenomenon that requires the willingness to be vulnerable to the actions of another. As automation technology continues to advance, trust evolves across different dimensions, including predictability, dependability, and faith (Hengstler et al,. 2016). According to Hengstler et al. (2016), trust initially stems from predictability, where users develop trust based on the system's consistent and expected behavior. As the system demonstrates dependability and reliability over time, trust deepens. Ultimately, faith emerges when users rely on the technology with confidence, knowing that it will perform as expected.

EARLIER¶21-26

To avoid confusion, it is crucial to define the terms engineer and designer in the context of our study. According to Charisi et al., (2017), intelligent autonomous systems involve three parties: designers (including developers and engineers), end-users (owners or customers), and various government and trade regulatory bodies and insurance agents. Our study used these parties as a guideline but delved deeper into the designer category by distinguishing between designers and engineers of AI systems. Specifically, we use the term engineer to refer to individuals who specialize in back-end development, such as AI coders or researchers focused on the technical aspects of AI. In contrast, designers are more oriented towards front-end development and user-centered design in the context of UX-researcher, UX/UI-designer, while researchers within these fields are considered professors specialized in system science and informatics. 4 2.0 Theory 2.1 Trust & disrupting the business model The traditional business model of selling cars is being disrupted by new mobility solutions. One of the most innovative companies in this space is Tesla Motors, which offers supplementary services through a modular approach. For instance, the Model S, Tesla's flagship car, is designed as a platform that can accommodate additional services. These include autonomous driving capabilities that can be uploaded and updated over the air for a fee. To make this possible, the company must develop cars with the services in mind and manage the development process accordingly (Mahut et al., 2017). In short, this means that car manufacturers are moving towards selling cars as a service, with smart features instead of selling just a vehicle with the only purpose to transport the driver from A to B. A survey conducted in Germany 2015 revealed that user acceptance of autonomous driving is low. Nearly half of the 1003 car buyers surveyed were skeptical, with only 5% preferring fully autonomous driving, approximately 49% preferred traditional driving and 43% preferred semi-autonomous driving (Hengstler et al., 2016). Building trust in autonomous driving is crucial for reducing perceived risk and encouraging the adoption of such technologies. At its core, trust is a fragile and evolving phenomenon that requires the willingness to be vulnerable to the actions of another. As automation technology continues to advance, trust evolves across different dimensions, including predictability, dependability, and faith (Hengstler et al,. 2016). According to Hengstler et al. (2016), trust initially stems from predictability, where users develop trust based on the system's consistent and expected behavior. As the system demonstrates dependability and reliability over time, trust deepens. Ultimately, faith emerges when users rely on the technology with confidence, knowing that it will perform as expected. In the case of radically new technologies like autonomous driving, establishing initial trust is crucial in overcoming perceived risks. Perceived risk is defined as the level of uncertainty surrounding the potential failure of a new product or the probability of the product not functioning as intended. In the context of AI, perceived risk is related to the fact that control is delegated to a machine, making the driver feel uneasy about allowing a machine to make decisions and take actions on its own, without human intervention (Hegstler et al., 2016). 2.2 Human involvement In an article from Hjetland (2015) she describes former research where the developers/ engineers have concluded that the human is the weakest link in the ecosystem. This is mainly in the order where the machine or in this case the vehicle can perform tasks more precise than humans can. Because the machine doesn't have human attributes (hunger, tiredness, etc.) which makes it more efficient and reliable. But there is a possibility that the machine breaks down or fails which requires human interaction (Hjetland, 2015). With the possibility of

EARLIER¶23-27

A survey conducted in Germany 2015 revealed that user acceptance of autonomous driving is low. Nearly half of the 1003 car buyers surveyed were skeptical, with only 5% preferring fully autonomous driving, approximately 49% preferred traditional driving and 43% preferred semi-autonomous driving (Hengstler et al., 2016). Building trust in autonomous driving is crucial for reducing perceived risk and encouraging the adoption of such technologies. At its core, trust is a fragile and evolving phenomenon that requires the willingness to be vulnerable to the actions of another. As automation technology continues to advance, trust evolves across different dimensions, including predictability, dependability, and faith (Hengstler et al,. 2016). According to Hengstler et al. (2016), trust initially stems from predictability, where users develop trust based on the system's consistent and expected behavior. As the system demonstrates dependability and reliability over time, trust deepens. Ultimately, faith emerges when users rely on the technology with confidence, knowing that it will perform as expected. In the case of radically new technologies like autonomous driving, establishing initial trust is crucial in overcoming perceived risks. Perceived risk is defined as the level of uncertainty surrounding the potential failure of a new product or the probability of the product not functioning as intended. In the context of AI, perceived risk is related to the fact that control is delegated to a machine, making the driver feel uneasy about allowing a machine to make decisions and take actions on its own, without human intervention (Hegstler et al., 2016). 2.2 Human involvement In an article from Hjetland (2015) she describes former research where the developers/ engineers have concluded that the human is the weakest link in the ecosystem. This is mainly in the order where the machine or in this case the vehicle can perform tasks more precise than humans can. Because the machine doesn't have human attributes (hunger, tiredness, etc.) which makes it more efficient and reliable. But there is a possibility that the machine breaks down or fails which requires human interaction (Hjetland, 2015). With the possibility of 5 failure it reduces the trust from humans to the machine which can lower the adoption rate. Hjetland (2015) also presents the importance of the trustworthiness towards an autonomous system, it could be misused in the course of human interaction. With the underestimated- or overestimated trust towards an autonomous system it can lead to accidents where human interaction is or isn't required. This clarifies the importance of trust and safety when human interaction is involved with an autonomous system. Hjetland (2015) also examined various articles and found that one of the major concerns with autonomous vehicles is the level of trust that people have in them, which aligns with the results from the survey conducted in Germany where only 5% prefer fully autonomous driving (Hengstler et al., 2016).

EARLIER¶28-33

In contrast to viewing humans as the weakest link in the ecosystem, Shneiderman (2020b) presents that both the human and the machine are valuable in different fields. The machine is more valuable when it is reliable and the human when the system requires human control, which machines regularly do. If the human is the weakest link in the ecosystem then the systems should not only be viable when it's reliable, but instead all the time. Aurenhammer (2020) highlights the importance of involving humans in the development of AI, to make it human centric. To involve humans in the development it requires a human-centered approach in both the philosophy and the designing of AI. The article presents two different views on AI which is rationalistic and design, the main differences between them is that one is more user-centric while the other one focuses more on the achievements of the machine. Aurenhammer (2020) describes that human centered design (HCD) creates an opportunity for humanity to see the benefits of involving a human-centric design approach when developing AI, to create a society that originated from human needs. 2.3 Artificial intelligence The idea of ​artificial entities, robots or creatures with their own intelligence is ancient and has existed as a human storytelling tradition for millennia (Wärnestål, 2021). Technical advances in algorithm research, the availability and large amounts of data, the internet as infrastructure, increased processing power and the spread of mobile devices have all contributed to the AI ​revolution that we are in the midst of in the early 2020s (Wärnestål, 2021). AI is a technology that can learn and make decisions in complex situations, even interact with people and maintain social relationships. This kind of intelligence mirrors the decision-making process observed in humans and some animals. Yampolskiy. (2020) defined AI as a system that can adjust to its surroundings and make decisions even if it doesn't have all the information it needs. This definition is helpful when we think about how autonomous cars work, because they need to be able to navigate different environments without a lot of training or knowledge. There are two main types of AI: narrow AI (also known as weak AI) and general AI (also known as strong AI). Narrow AI refers to AI systems that are designed to perform a specific task or a narrow range of tasks. These systems are usually based on data-driven machine 6 learning algorithms and can perform their task with high accuracy and efficiency (Wärnestål, 2021). In contrast, general AI refers to AI systems that exhibit intelligence across a wide range of tasks, similar to human intelligence (Wärnestål, 2021).

EARLIER¶36-40

The distinction between the transparency and explainability needs of end users and developers is crucial. Developers require in-depth understanding of the underlying attributes and algorithms, while end users need more comprehensive and simplified explanations about the system itself. To develop systems that satisfy both groups is a big challenge, and most interfaces tend to focus on either developers or end users (Ozmen et al., 2023). There is a high need for explainability based on interpretable transparency in case an AI system fails or makes ethically relevant decisions. The individual affected would want to ensure that mistakes that were made are not repeated and that distribution of outcomes are based on fair, legal, and ethical decision-making mechanisms in society. The importance of explainability becomes even more pronounced in light of possible responsibility gaps that may arise (Ozmen et al., 2023). 2.4 Definition for autonomous cars Autonomous driving is at the moment one of the most intensively researched technologies in the transportation domain (Hengstler et al., 2016). To the layperson, the definition of autonomous vehicles may appear self-evident, characterized simply as a vehicle capable of self-driving. But going in-depth to its actual meaning, it refers to that the vehicles are equipped with Advanced Driver Assistance Systems (ADAS) that help the driver make decisions and maneuvers in day-to-day driving as well in critical situations like a collision (Rödel et al., 2014). This definition might seem quite broad, as there are different levels of autonomy. Therefore, an automotive standardization for autonomous cars was defined by the Society of Automotive Engineers (SAE). These levels range from no autonomy (Level 0) to full autonomy (Level 5), and each level represents an increasing degree of automation and decreasing reliance on human intervention (Ondruš et al., 2020). Level 0: No automation: The driver is in control of the vehicle at all times. 7 Level 1: Driver assistance: The driver is still in control, but the car can assist the driver with steering or braking. Level 2: Partial automation: The car can control both steering and acceleration/deceleration, but the driver is still responsible for monitoring the environment and taking over control when necessary.

EARLIER¶43-47

Level 5: Full automation: The car can handle all driving tasks in all conditions and environments, and no human intervention is required. 2.5 Ethics Matters for Autonomous Cars The field of ethics has been studied from various perspectives where all focused around logical reasoning and ethical evaluations. In other words, ethics explains the differences in right or wrong from the humans actions, what is morally right to do. It is concerned with determining what is right or wrong, good or bad, and just or unjust in human behavior and decision-making (Chonko, 2012). However, as we shift from human decision-making to system-based decision-making, it introduces an additional layer of unexpected behaviors, dilemmas and scenarios that increase the complexity. A study by Massachusetts Institute of Technology (MIT) on the “MIT Moral Machine” offers valuable insights into the influence of cultural and societal factors on moral decision-making in autonomous vehicles, especially in death-threatening situations (Wärnestål, 2021). Results of the study indicated that individuals from Southeast Asian countries were more prone to prioritize the well-being of an older individual over a younger person in a moral dilemma, as compared to those from Western cultures. Conversely, individuals from Hispanic and French cultures were more likely to spare an athletic individual over someone overweight. In regions characterized by high economic inequalities, such as Colombia, participants tended to prioritize the well-being of a businessman over a homeless individual (Wärnestål, 2021). Having this context in mind, what would be the appropriate course of action? Let us imagine a scenario where you're driving an autonomous car and your car is faced with a terrible choice: it must either swerve left and hit an eight-year-old girl or swerve right and hit an 80-year-old grandmother. Due to the current speed of the car, either victim would most likely not survive the impact. If the car does not swerve, both individuals will be hit and most likely killed, making it imperative to swerve one way or the other (Maurer et al., 2016). The question arises, what is the ethically correct decision in such a scenario? If you were responsible for programming the self-driving car, how would you program it to behave? It may seem tempting, based on the previous research from MIT, to program the car to hit the younger 8 person in East Asian countries and the older person in Western countries, but this approach is ethically questionable. Moving away from the ethical perspective and into the perspective of the AI, let us imagine an autonomous car facing an imminent crash where it must make a terrible choice: it must either swerve a motorcyclist wearing a helmet or a motorcyclist without one. The safety of the car and its occupants may not be significantly affected by whether the motorcyclist is wearing a helmet or not, but the impact of a helmet hitting the car window may cause more damage to the car. Therefore, we must consider that if we optimize the algorithm without ethical considerations, the program may become selfish and choose to swerve towards the person without a helmet to reduce the damage done to itself (Maurer et al., 2016). It is crucial to acknowledge that the presence of a helmet is of importance to the safety of the motorcyclist. In this dreadful scenario, it seems reasonable to program a good autonomous car to prioritize the safety of the motorcyclist and swerve into the one wearing the helmet. These scenarios and findings underline the substantial impact that cultural, environmental and context differences have on ethical decision-making in autonomous vehicles. It can be argued that these differences and development in regards to these concerns may be one of the most important factors of ethics, as they provide answers to the question of "whose ethical behavior?", as ethical behavior is different to different cultures, people and may differ in different contexts and raise important considerations regarding the adaptability of autonomous cars to cultural variations when operating in different countries (Wärnestål, 2021). Hence, the examination of cultural and societal factors in ethical decision-making is crucial in the development of ethics for autonomous vehicles and opens up hard questions such as “how should a self-driving car actually behave in traffic, and should it change as soon as you want to drive abroad where cultural perceptions may look different?” (Wärnestål, 2021).

EARLIER¶81-85

Designer C Working with user experience at leading transportation company, to streamline the use of vehicles. Have practical experience with UI/UX for AI systems. Engineer A Expertise in Cyber Physical Systems (CPS). Background in developing active systems for trucks, cars, and race cars, along with experience in connecting mechanical and electronic components. 16 4.0 Results & Analysis In enlightening of an ongoing shift in the field of AI and Industry 5.0 (Leng et al., 2022; Xu et al., 2021; Shneiderman, 2020; Wärnestål, 2021; Shneiderman, 2020a; Shneiderman, 2020b) we assumed that people that have worked in the industry for a longer period of time has a less holistic view in regards to ethical dilemmas as compared to newly graduated engineers within the field. The reason for our assumption is due to the fact that there is an ongoing shift within the field of AI, and that this shift is researched and studied within universities and academic people. On the other hand, those who work in the industry may not always be aware of the changes taking place in their environment. While we were not able to confirm our theory regarding newly graduates having a more holistic view, we were able to see that engineers are more prone to the rationalistic view and tend to follow regulations or frameworks, such as EU legislation. Whereas designers follow user-centered approaches and overall have a more holistic view of the ethical and external implications of poorly designed AI systems. We also draw a parallel between the ongoing shift with industry 5.0 and AI where both previously were defined with cutting-edge technologies that had high promise with impressive improvements to effectiveness, efficiency, and economical benefits. But not until recent times is it adding a new dimension; having a system/holistic view by having external and ethical considerations when developing the product or service. Our analysis revealed that there are notable differences in how participants viewed the importance of transparency, reliability, and safety. While some participants prioritize transparency as the most crucial pillar, others place more emphasis on reliability or safety. Additionally, we found that some participants view these pillars as interdependent, whereas others see them as separate and distinct concepts. The results of our analysis are presented in separate chapters, each focused on a specific subfield related to transparency, reliability, and safety.

EARLIER¶120-124

It's important to note that Figure 2 illustrates the varying levels of knowledge and importance of designers and engineers in different fields of AI, with the aim of achieving a trustworthy AI. The triangles in the Figure represent each group and highlight the areas where the differences between them are the greatest. To facilitate comprehension of the Figure, we have established interwalls consisting of values ranging from 0 to 1. The values indicate the level of importance assigned to each field by both designers and engineers. To clarify the parameter, we have set the range of 0.8 - 1.0 as high importance, 0.4 - 0.79 as medium importance, and 0 - 0.39 as low importance for both groups. Being positioned in the 0.8 - 1.0 indicates that their perspective is better understood by the end-users and has a greater potential to achieve trust. For instance, the emphasis on transparency by designers means that they are more likely to involve users in the transparency process, which is an approach we encourage. Conversely, engineers have a better understanding of the underlying technology and can determine when the software is reliable enough to support innovation, which also is an approach we encourage. It's important to emphasize that effective communication between designers and engineers is crucial to achieve transparency and reliability in the development of autonomous cars. As we mentioned earlier, transparency should be designed together, with the end-user involved in the process to determine which information should be visible. This ensures that the end-user can understand the technology, its limitations, and its potential, which is essential in building trust. Therefore, it's important for designers and engineers to collaborate and communicate effectively to develop a transparent and reliable framework for autonomous cars. By working together, they can leverage their respective strengths to create a comprehensive and trustworthy system that meets the needs and expectations of end-users. The interviewed respondents were asked if they considered humans to be the weakest link in the ecosystem in relation to self-driving cars. The respondents from both the designer and engineering profession were very united on that issue. Although most participants in our study 23 agreed that humans were the weakest link of AI systems, there was one notable outlier. This participants argued that if an AI system is unable to react or collaborate effectively with human drivers in traffic, then the system may actually be the weakest link.

EARLIER¶146-150

Discussion is an alternative measurement in order to foster deeper understanding and collaboration. Involving participants in discussions about ethical issues opens up for critical thinking and exploration of different perspectives. Having someone able to monitor and evaluate the discussions enables the participants for a deeper ethical understanding which can be used to evaluate the participants progress in ethical development. 29 6.0 Conclusion The research questions for this thesis were as follows: 1. What is the difference in perspective between designers and engineers with regard to ethics when developing autonomous vehicles? 2. What strategies can be implemented to bridge the gap in perspectives between designers and engineers regarding the development of autonomous cars? To answer these research questions, our research has found that there are differences in perspectives between designers and engineers when it comes to ethics in developing autonomous vehicles. While both groups acknowledge the importance of safety and security, designers tend to prioritize transparency, while engineers prioritize reliability. However, effective communication and collaboration between designers and engineers can lead to the development of a comprehensive and trustworthy framework for autonomous vehicles that meets the needs and expectations of end-users. Our research also highlights the importance of considering the interaction between humans and AI in evaluating the effectiveness and reliability of autonomous vehicle systems. Additionally, we recommend ethics to be incorporated into education of designers and engineers to ensure a shared understanding of existing frameworks and legislation and to enable the creation of products that meet customer expectations while incorporating ethical considerations. While it’s important for developers and engineers to have education in regards to ethical AI, it should also be noted that if AI continues to develop in the same phase as it has in the last years, AI will have the potential to become as much of a necessity to our daily lives as the internet. Thus, we believe that education of ethical and responsible AI should be implemented in the school system at an early age e.g., elementary school to make sure that the country stays or becomes at the forefront of an evolving and impactful technology. The responsibility for ensuring ethical considerations are incorporated into AI should be shared among all parties involved, including universities, workplaces, and individuals working on AI projects. By doing so, a common ground for ethical practices can be established, making it an essential aspect of future AI development. By reducing the gap in perspective between designers and engineers it is essential to develop a transparent and reliable framework for autonomous cars that meets the needs and expectations of end-users. Collaboration and effective communication between designers and engineers are critical to achieve transparency and reliability in the development of autonomous cars. It's important to involve the end-user in the process of determining which information should be visible and to consider the interaction between AI systems and their human users when evaluating the effectiveness and reliability of these systems. Furthermore, incorporating ethics into the education of designers and engineers can help ensure that they have a shared goal of developing trustworthy AI for autonomous cars. By leveraging their respective strengths and working together, designers and engineers can create a comprehensive and trustworthy system for autonomous cars. 30 While our findings enabled for theoretical strategies to bridge any discrepancies in ethical views, these strategies have not been field tested, thus we recommend for future research to apply our recommendations in workplaces and real life projects in order to see if the gap is reduced to improve the trust importance of autonomous cars. In addition, we have noted that ethics is not limited to transparency, reliability and safety and researchers could explore other factors that influence trust in AI systems, such as security, integrity and responsibility. By analyzing the interplay between these factors and the recommendations proposed in this study, researchers could provide a more comprehensive understanding of how to build trust in AI systems. Ultimately, the goal would be to identify the most effective strategies for increasing trust in AI systems and to promote their widespread adoption in the development of future systems.

EARLIER¶124-128

23 agreed that humans were the weakest link of AI systems, there was one notable outlier. This participants argued that if an AI system is unable to react or collaborate effectively with human drivers in traffic, then the system may actually be the weakest link. This insight highlights the importance of considering the interaction between AI systems and their human users when evaluating the effectiveness and reliability of these systems. While it's true that humans may introduce errors or biases and not be able to do tasks as efficiently or precisely, it's also important to recognize that effective collaboration between humans and AI is essential for ensuring the safe and efficient deployment of these systems. In addition, we found that both engineers and designers often disagree on how ethics are implemented in their projects. They typically rely on existing guidelines and frameworks from the EU legislation and the project's customer to resolve these issues. Although both groups believe that there are already enough frameworks and legislation in place, they do agree that these guidelines should be taught as part of their education, either in school, or at the company since the projects tend to disregard these guidelines or not have a full understanding of them. By incorporating ethics into their education, team members can better understand the existing frameworks and legislation, which can help ensure that they have a shared goal of developing trustworthy AI for autonomous cars. Early education in ethics would emphasize its importance and enable individuals to create products that meet customer expectations while incorporating ethical considerations. Figure 2, Perspective Comparison 24 While collecting data, we received various insights on how ethical considerations could be improved in AI-related projects for autonomous cars. These insights are summarized in Table 1. By utilizing these findings, we were able to establish recommendations for both designers and engineers to incorporate ethical considerations into the development process. Table 2, summary of differences and recommendations that can be used as strategies Subfield of ethics Differences Recommendation Transparency Engineers prioritize access to technical attributes and information about the car's decision-making process, while designers are primarily concerned with presenting this information in a digestible and user-friendly format that is accessible to non-technical individuals. The engineers do not believe that this information is crucial for the driver, while the designers are not particularly interested in the underlying technical aspects of the decision-making process.

EARLIER¶126-130

In addition, we found that both engineers and designers often disagree on how ethics are implemented in their projects. They typically rely on existing guidelines and frameworks from the EU legislation and the project's customer to resolve these issues. Although both groups believe that there are already enough frameworks and legislation in place, they do agree that these guidelines should be taught as part of their education, either in school, or at the company since the projects tend to disregard these guidelines or not have a full understanding of them. By incorporating ethics into their education, team members can better understand the existing frameworks and legislation, which can help ensure that they have a shared goal of developing trustworthy AI for autonomous cars. Early education in ethics would emphasize its importance and enable individuals to create products that meet customer expectations while incorporating ethical considerations. Figure 2, Perspective Comparison 24 While collecting data, we received various insights on how ethical considerations could be improved in AI-related projects for autonomous cars. These insights are summarized in Table 1. By utilizing these findings, we were able to establish recommendations for both designers and engineers to incorporate ethical considerations into the development process. Table 2, summary of differences and recommendations that can be used as strategies Subfield of ethics Differences Recommendation Transparency Engineers prioritize access to technical attributes and information about the car's decision-making process, while designers are primarily concerned with presenting this information in a digestible and user-friendly format that is accessible to non-technical individuals. The engineers do not believe that this information is crucial for the driver, while the designers are not particularly interested in the underlying technical aspects of the decision-making process. Engineers and designers should collaborate throughout the development of transparency features in autonomous cars. Including end-users in the process can ensure that transparency is developed with all stakeholders in mind, and that the technology is not limited in the future due to any specific considerations made during the development process Reliability Engineers tend to have more trust in the technology of autonomous cars, especially after relevant tests have been done. Designers, on the other hand, are more likely to view the reliability of autonomous cars with greater caution, preferring to wait until a larger percentage of users have tested the technology before fully trusting it.

EARLIER¶126-131

In addition, we found that both engineers and designers often disagree on how ethics are implemented in their projects. They typically rely on existing guidelines and frameworks from the EU legislation and the project's customer to resolve these issues. Although both groups believe that there are already enough frameworks and legislation in place, they do agree that these guidelines should be taught as part of their education, either in school, or at the company since the projects tend to disregard these guidelines or not have a full understanding of them. By incorporating ethics into their education, team members can better understand the existing frameworks and legislation, which can help ensure that they have a shared goal of developing trustworthy AI for autonomous cars. Early education in ethics would emphasize its importance and enable individuals to create products that meet customer expectations while incorporating ethical considerations. Figure 2, Perspective Comparison 24 While collecting data, we received various insights on how ethical considerations could be improved in AI-related projects for autonomous cars. These insights are summarized in Table 1. By utilizing these findings, we were able to establish recommendations for both designers and engineers to incorporate ethical considerations into the development process. Table 2, summary of differences and recommendations that can be used as strategies Subfield of ethics Differences Recommendation Transparency Engineers prioritize access to technical attributes and information about the car's decision-making process, while designers are primarily concerned with presenting this information in a digestible and user-friendly format that is accessible to non-technical individuals. The engineers do not believe that this information is crucial for the driver, while the designers are not particularly interested in the underlying technical aspects of the decision-making process. Engineers and designers should collaborate throughout the development of transparency features in autonomous cars. Including end-users in the process can ensure that transparency is developed with all stakeholders in mind, and that the technology is not limited in the future due to any specific considerations made during the development process Reliability Engineers tend to have more trust in the technology of autonomous cars, especially after relevant tests have been done. Designers, on the other hand, are more likely to view the reliability of autonomous cars with greater caution, preferring to wait until a larger percentage of users have tested the technology before fully trusting it. It is essential that designers are well-informed about autonomous vehicle technology and are actively involved in the testing processes. This would increase their knowledge of the technology and create more trust. Furthermore, experts and professionals working in this field should take a proactive role in shaping regulations and laws related to AI technologies.

EARLIER¶128-133

Table 2, summary of differences and recommendations that can be used as strategies Subfield of ethics Differences Recommendation Transparency Engineers prioritize access to technical attributes and information about the car's decision-making process, while designers are primarily concerned with presenting this information in a digestible and user-friendly format that is accessible to non-technical individuals. The engineers do not believe that this information is crucial for the driver, while the designers are not particularly interested in the underlying technical aspects of the decision-making process. Engineers and designers should collaborate throughout the development of transparency features in autonomous cars. Including end-users in the process can ensure that transparency is developed with all stakeholders in mind, and that the technology is not limited in the future due to any specific considerations made during the development process Reliability Engineers tend to have more trust in the technology of autonomous cars, especially after relevant tests have been done. Designers, on the other hand, are more likely to view the reliability of autonomous cars with greater caution, preferring to wait until a larger percentage of users have tested the technology before fully trusting it. It is essential that designers are well-informed about autonomous vehicle technology and are actively involved in the testing processes. This would increase their knowledge of the technology and create more trust. Furthermore, experts and professionals working in this field should take a proactive role in shaping regulations and laws related to AI technologies. 25 Safety Both designers and engineers are committed to advancing the field with good intentions. The perspective of safety is aligned in both fields and is also reflective in the use-case of autonomous cars where Tesla autopilot is ten times safer than the average American driver. Although their perspectives are aligned towards safety, it is crucial for designers and engineers to remain up-to-date with the latest advancements to ensure that the technology continues to progress and become even safer and make noticeable adjustments if needed, such as bad reports from the Tesla autopilot.

EARLIER¶130-135

Reliability Engineers tend to have more trust in the technology of autonomous cars, especially after relevant tests have been done. Designers, on the other hand, are more likely to view the reliability of autonomous cars with greater caution, preferring to wait until a larger percentage of users have tested the technology before fully trusting it. It is essential that designers are well-informed about autonomous vehicle technology and are actively involved in the testing processes. This would increase their knowledge of the technology and create more trust. Furthermore, experts and professionals working in this field should take a proactive role in shaping regulations and laws related to AI technologies. 25 Safety Both designers and engineers are committed to advancing the field with good intentions. The perspective of safety is aligned in both fields and is also reflective in the use-case of autonomous cars where Tesla autopilot is ten times safer than the average American driver. Although their perspectives are aligned towards safety, it is crucial for designers and engineers to remain up-to-date with the latest advancements to ensure that the technology continues to progress and become even safer and make noticeable adjustments if needed, such as bad reports from the Tesla autopilot. 26 5.0 Discussion 5.1 Insights & Reflections Effective communication is a critical aspect of any project, particularly when it comes to developing AI systems. As these systems continue to be ubiquitous in our daily lives, it's essential to prioritize the needs of end-users. AI is intended to make our lives easier and more convenient, and to achieve that goal, we must prioritize communication and collaboration in the development and deployment of these systems. In this thesis project, we explored the importance of adopting a united view when creating trustworthy AI. Our study results highlight the pivotal role of communication and end-user involvement in ensuring successful AI deployment. While our study's reliability may be affected by the sample size of our interviews, it's important to note that, toward the end of our interview process, we began to receive similar responses from participants with similar backgrounds. Despite this limitation, we believe that the insights gained from the interviews provide sufficient information to make generalizations about the study's findings. We believe that in order to start receiving more valuable responses, the sample population needs to increase and involve more participants from different companies within the industry. With gained insights from people working on projects where designers and engineers collaborate together and use different guidelines would be valuable to get a better understanding of the differences.