Responsibility in Autonomous Driving
– The Impact of the Organisational Structure on the Sense of Responsibility across the Organisation.
What are some of the reasons why an agile team-based organisation can encourage a responsibility across organisational members developing autonomous driving systems?
Autonomous Driving, an emerging technology that used to be fiction and now becomes reality. With this technology it is desired to remove human error and, thereby, drastically reduce the number of fatalities and crashes on roads, it raises high expectations towards those developing the self-driving vehicles. Industry stakeholders who aim to maximise road safety are facing challenges on multiple levels concerning ethical, social, economic, and legal questions.
Automated driver assistant system (ADAS) is a closed system of functions. It starts with sensing the surrounding environment, localising the position of the vehicle, processing the location and sensory data, planning the reaction of the vehicle, and finally, controlling the behaviour of vehicles through actions, such as steering or breaking, summarised Jo, et al. (2014, p. 7132). If this system runs error-free, it is an automated and continuous flow of data input and automated reactions of the object as the output. It enables vehicles to move from A to B without outside control. But, what if in one of these functional steps something goes wrong, data is misinterpreted, or the car simply reacts too slowly? Who is responsible for an accident caused by an automated driving car? One of the cases of such an accident caused by an automated driving car that gained strong media coverage is the Autopilot of the brand Tesla in May 2016 (Yadron & Tynan, 2016). Industry stakeholders, like Tesla, need to design strategies to minimise such technological risks. This discourse goes beyond organisational boundaries. It involves a variety of stakeholders, such as insurance companies, policy makers, technology experts, economists, philosophers, the general public, or the European Group on Ethics in Science and New Technologies. This group stated in 2018 that “‘autonomous’ systems should be guided by an authentic concern for research ethics, social accountability of developers, and global academic cooperation to protect fundamental rights and values” (Directorate-General for Research and Innovation , European Group on Ethics in Science and New Technologies, p. 17).
Along with this group of experts, most stakeholders will argue that the company selling the automated driving car holds responsibility for a technology that “should not pose unacceptable risks of harm to human beings” (Directorate-General for Research and Innovation, European Group on Ethics in Science and New Technologies, 2018, p. 17). Taking this a step further, one can then question who specifically is the person to blame for potential crashes like in the Tesla Autopilot case (Yadron & Tynan, 2016). Whether it is the seller, the Chief Executive Officer or the mechanic physically building the car, is difficult to track or decide. Hence, rather than debating about and defining organisational roles to blame for errors, it seems to be more relevant to take a proactive approach and develop strategies to avoid errors of the system in the first place.
To engage in the discussion of the company’s responsibility in autonomous driving, this paper will argue the importance of a holistic organisation-wide internalisation of responsibility as a more sustainable and highly impactful proactive measure. One possible strategy is suggested in this paper, outlining benefits of structuring organisations in an agile short-cycled team-based way.
Partially, this paper refers to my research conducted at BMW Autonomous Driving. I analysed key factors affecting collaboration and knowledge sharing across the organisation. Amongst other aspects, the effects of team-based working on self-identification and sense of responsibility are discussed. I reviewed literature on organisational behaviour and applied psychology, and conducted focus groups and interviews. The ADAS business unit redesigned their organisation, adopting an agile working model, called Large Scale Scrum (LeSS). According to Ribel and Larman, one key benefit of the LeSS adoption was a “higher degree of whole system understanding among all developers” (2018).
Firstly, a team-based organisation helps its members to work towards collective goals and encourages individuals to acknowledge how one’s work contributes to the overall product. Keeping a common goal in mind reinforces cross-team alignment by building on a level of self-interest and commitment of individuals to personally aim for and identify themselves with what the company wants to achieve (Wong, et al., 2009, p. 2896; Johnson, et al., 1981; Ehrhart & Naumann, 2006; Deutsch, 1973; Baumgartner, 2019, p. 42). In the case of ADAS, the overall organisational goal would therefore include maximal safety. However, to make it tangible for teams to identify themselves with the product and feel responsible for it, one needs to understand how their team’s work contributes to the product. Thus, managers, software developers, user-experience experts, as well as those addressing ethical issues, can be grouped into cross-disciplinary teams. These focus on single product features. For instance, a team holding the end-to-end responsibility for a safe and error-free drive from A to B on the highway. This ‘highway-drive’ feature is more tangible and reinforces the working group to see their interdependency to other teams, all building one overall safe ADAS. In seven out of nine interviews during my research at BMW, the need for overall organisational goals was considered fundament for cross-team collaboration and a collective sense of responsibility. A supporting quote by an employee was “’The mass of people working here needs to understand why we are here and where we want to go. (…) This could counteract the separation of single teams as individuals understand how complex the product is’ and that ‘individual teams will not able to develop self-driving vehicles on their own’ (Participant F, personal communication, February 11, 2019).” (Baumgartner, 2019)
Secondly, a short-cycled, incremental, and experimental approach to product development implies knowledge of results of a teams’ action and work quickly. Several organisational design experts, like Craig Larman (2004), Hackman & Oldham (1976) or Gartzen, et al. (2016), support this argument. The extent to which individuals see results of their work personally, significantly relates to the one’s internal job motivation, researched Hackman & Oldham (1976, pp. 258 – 273). In the example of a secure highway drive, cross-disciplinary teams are able to test their work and development progress promptly. One way to test and validate features is by simulating test cases vehicles encounter, explained Sippl, et al. (2016). Hence, the team, instantly becomes aware of potential risks and errors. If their written code runs error-free, they can approve and validate the code. In case an error occurs, teams can directly remedy faults. One software developer at BMW explained that it “helps to ‘see the processes that we’re making (…), see achievements and (…) feel responsible for [it]. So I have a sense of pride, a sense of ownership, a sense of co-creation, and simply an enjoyment’ (Participant D, personal communication, February 7, 2019).” (Baumgartner, 2019, pp. 40 – 41). Conceivably, teams vary between testing their feature incrementally with a simulation tool and, parallelly, test developed features by physically sitting in the vehicle. This makes it more tangible for individuals to internalise their responsibility as they feel effects on road safety and their personal lives themselves.
Thirdly, through a high degree of autonomy, members of self-organising teams experience a stronger sense of responsibility for the outcomes of their work. Hackman and Oldham (1976, pp. 272 – 273) suggested in one of their studies that “the desirability of high autonomy and high experienced responsibility for achieving beneficial work outcomes” correlate significantly. Examples for the desired autonomy may be that team members collect ideas on how they develop certain features, how they structure their working day, or how they group with other teams around work in a beneficial way.
To conclude, this paper suggests organisations to adopt an agile team-based organisation when developing autonomous driving cars to deeply root organisational-wide responsibility. It is argued as a strategy encouraging responsibility in teams through collectively aiming for overall organisational goals, allowing immediate knowledge of results, as well as through delegating a high degree of autonomy to individuals. To recommend implications for further research on this topic, I suggest to identify what role the managers play to support a strong sense of responsibility in teams. Additionally, it is notably important to consider and research cultural dimensions, personal preferences and needs, and individual characteristics of employees on the sense of accountability. Nevertheless, when adopting an agile team-based organisational working model, it also remains to be clarified how to change from a personal liability model to a model in which a team is legally accountable for their work.
After all, though, there is one fundamental issue that lays out of the control of the single company and is not related to the individuals’ sense of responsibility – it is that ADAS are not solely a matter of an error-free software, but also heavily dependent on an error-free hardware. Potential risks are as basic as if sensors are soiled or wet, or if the vehicle loses its connection to the internet in rural areas – no matter how safe the software is being built – the external dependencies are out of control of single employees. Hence, companies selling automated driving cars need to develop strategies determining how the system should react in such cases. Thus, further research should therefore, too, consider possibilities of sudden failure, defect or flaw of sensors. For instance: How can the company ensure that the vehicle including its passengers and load will come to a full stop and a safe state?
Copyright by Claudia Baumgartner.
Literature
Baumgartner, C., 2019. A Change Communication Strategy to Improve the Collaboration and Knowledge Sharing in an Agile Multi-Team Software Development Project: Business Unit Autonomous Driving at BMW Group. The Hague: The Hague University of Applied Sciences.
Deutsch, M., 1973. The resolution of conflict. New Haven, CT: Yale University Press.
Directorate-General for Research and Innovation , European Group on Ethics in Science and New Technologies, 2018. Statement on artificial intelligence, robotics and ‚autonomous‘ systems, Brussels: European Commission.
Ehrhart, M. G. & Naumann, S. E., 2006. Organizational citizenship behav-ior in workgroups: A group norms approach. Journal of Applied Psychol-ogy, pp. 960 – 974.
Gartzen, T., Brambring, F. & Basse, F., 2016. Target-oriented Prototyping in Highly Iterative Product Development. Science Direct – 3rd International Conference on Ramp-up Management (ICRM) , pp. 19 – 23.
Hackman, J. R. & Oldham, G., 1976. Motivation through the Design of Work: Test of a Theory. Organizational Behavior and Human Performance, August, Volume 2, pp. 250 – 279.
Jadhav, A., 2018. Autonomous Vehicle Market Outlook – 2026, s.l.: Allied Market Research.
Johnson, D. W., Maruyama, G. J., R. T., N. D. & Skon, S., 1981. Effects of cooperative, competitive, and individualistic goal struc-tures on achievement: A meta-analysis.. Psychological Bulletin, pp. 47 – 62.
Jo, K. et al., 2014. Development of Autonomous Car—Part I: Distributed System Architecture and Development Process. IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, 61(12), pp. 7131 – 7140.
Kröger, F., 2016. Automated Driving in Its Social, Historical and Cultural Contexts. In: G. J. L. B. W. H. Maurer M., ed. Autonomous Driving. Berlin, Heidelberg: Springer, pp. 41-68.
Larman, C., 2004. Agile and iterative development : a manager’s guide. Boston: Addison-Wesley: Agile Software Development Ser.
Larman, C. & Vodde, B., 2009. Scaling Lean & Agile Development Thinking and Organizational Tools for Large-Scale Scrum. 1 ed. Upper Saddle River, NJ: Addison-Wesley Professional.
Ribel, K. & Larman, C., 2018. LeSS Huge at BMW. Munich, LeSS Conference.
Sippl, C. et al., 2016. From Simulation Data to Test Cases for Fully Automated Driving and ADAS. Graz, Austra, 28th IFIP International Conference on Testing Software and Systems (ICTSS), pp. 191 – 206.
Wong, A., Tjosvold, D. & Liu, C., 2009. Cross‐Functional Team Organizational Citizenship Behavior in China: Shared Vision and Goal Interdependence Among Departments. Journal of Applied Social Psychology, 09 December.
Yadron, D. & Tynan, D., 2016. Tesla driver dies in first fatal crash while using autopilot mode. TheGuardian, 1 July.