Automating vehicles is a solution for … what? Five theses concerning automation, public spaces and public involvement in an emerging technology
In this article, I will present five theses concerning the ongoing race for bringing automated vehicles on public roads. Some of my theses are quite disillusioning, some provide hope for a better future of urban mobility with fewer private cars, some concern seemingly “technical” details. What underlies my thinking is the fact that in a field like automated driving, where technical and regulative norms are still evolving and being recalibrated, new technical norms introduce new social norms on our roads, and they enable wholly new patterns of mobility behavior. I study technological innovations as phenomena which are developed within society, within existing power relations, and within constantly changing societal norms and beliefs of what is a “good”, “achievable” or a “bad” future. In most cases, new technologies are presented as means for achieving a better future within a society.
There are no tangible conflicts concerning automated vehicles yet, but some issues of possible contention are already foreshadowed. I will draw from various sources such as scientific discourse, public debates, and first experiments on public roads to show that automated vehicles will impact technology-society relations on various scales, and that they could expose society to various risks, especially risks related to overblown claims in public discourse and risks concerning the spatial development of urbanity. Lastly, it is not clear which societal groups would profit from automated vehicles and respective services and how they can participate in this early phase of testing and development.
The article is structured as follows. I will start by explaining why it is deceptive to speak of autonomous instead of automated vehicles (1). I will depict critiques which relate more to the technical side of automating vehicles: the standards of levels of automation, so-called SAE levels, exhibit a bias towards the perspective of engineers (2). I will clarify what “Autonowashing” means, a term coined by Liza Dixon (3). Having made these two points on standards and safety, which both relate to what is going on right now in the field of automated driving, I will shed light on the mid- or long-term consequences of automating vehicles. I will argue that the automation of vehicles, thought of in broader terms, loosens the quasi-natural connection between automation and the private car serving as the default space or spatial blueprint “in need” to be automated (4). This loosening will radically change the attachments between individuals and cars as we know them from our car-based socialization in modern societies. Finally, I will show three methods with which future urban spaces, marked by automated cars, are being imagined right now (5).
(1) Autonomous cars? Kant says no
One of the most influential definitions of autonomy, which has shaped the discursive history of the term in the modern age, was proposed by the Prussian philosopher Immanuel Kant: “the autonomy of the will is the property of the will by which it is a law to itself independently of any property of the objects of volition.“ (Kant 1996: 4:440) – a term I doubt is adequate in the case of automated vehicles. Autonomy in the Kantian sense is not the right term for such vehicles as any autonomous driving system makes up laws for itself which are highly dependent on the perception of the properties of its “objects of volition”: passengers, traffic lights, lane markings and cyclists who are being overtaken. Connected and automated cars create the objects of volition cooperatively and not autonomously.
If automated vehicles were autonomous, they would create the problem of recklessly bullying driver personalities, making their driving behavior dependent only on their own reasoning. Even cars that only rely on their own sensors or cameras need software updates, which makes them anything but autonomous. The whole idea of connected cars – cars communicating and coordinating their behavior with each other, the infrastructure and the cloud – contradicts the definition of autonomy in the Kantian sense of the Enlightenment.
In this sense, the term “autonomous cars” is not only a mockery of the history of philosophy, it is misleading for what is actually currently being developed by engineers and imagined as mobility concepts of the future.
In fact, case studies about testing automated vehicles in certain areas show how many requirements have to be met in order to run an automated shuttle on a public road, e.g., as part of the wider network of public transportation (Soteropoulos t al 2021). Is it accurate to call a vehicle autonomous if it still needs an operator on board (or remotely) to intervene with a joystick in complex traffic situations? Is the label of autonomy helpful if automated shuttles depend on the installment of additional signs next to the lane in case the buildings normally serving as orientation are not high enough or missing (see image below)? What is autonomous about an automated car that depends on regular software updates, a functioning and strong WIFI connection or good weather conditions for cameras to work? Not much, I would suggest.
A realistic view on vehicle automation necessitates a critical debate about what is achieved by naming automated vehicles “autonomous” when they in fact heavily depend on humans, infrastructure and ecology. Seen from this perspective, it is worthwhile to ask if public understanding of and dialogue about technology can be enhanced by opposing the claim of their autonomy. Would it not be better to explicate the manifold attachments and dependencies of automated vehicles in order to ground the implementation of potential use cases in transparent discussions of the new technology’s impacts on public space and digital infrastructure?
(2) SAE levels of automation and their bias
Legal scholars (Gollrad 2020) and social scientists (Stayton/Stilgoe 2020) point at the misleading character of the international standardization of levels of automation (SAE levels). SAE levels start at zero (no driver assistance) and end at level 5 (full automation, no intervention required, vehicle is supposed to drive everywhere under all conditions).
Why are SAE levels said to be in need of redefinition? Short answer: a bias towards the perspective of developers and engineers is inscribed into their logic.
SAE levels were developed based on a study of the German Federal Highway Research Institute (BASt) which was taken up by the Society for Automotive Engineering (SAE) – the latter being a non-profit association of the automotive sector which published the most important scheme for standardizing automated vehicle functions: the SAE standard J3016. This standard, regularly updated, distinguishes the aforementioned six levels of automation based on the division of labor between humans and machines.
SAE levels are particularly suited for the practices of engineers and developers as they emerge from their working practices. They are intended to support the development of a safe automated vehicle. Developers work to “find” solutions for the “task” of automating the driver. But, as the sociology of road/car culture shows, “being the driver” has always been a culturally embedded social role relating to others in the car, to the car itself, other road users and the infrastructure. The built-in logic of the SAE-Levels suggests that the task of automation is a replacement of a formerly social role taken up by humans by a technical role-script executed by artificial intelligence. And indeed, automating vehicles means replacing human drivers with AI. But as long as automated vehicles operate in mixed traffic, their behavior needs to be socially intelligible for humans. That is to say, SAE-Levels narrow down the new social roles of vehicles to mere technical questions of how to replace error-prone humans, thereby downplaying the variety and contingency of operating conditions. This informs the inadequate meaning of autonomy denoting a car independent of its factual dependencies on place, infrastructure, road culture, the complex behavior of pedestrians and cyclists, the weather and light conditions.
In this view, full autonomy means the capability of a vehicle to behave safely and properly no matter what kind of (traffic, weather, society-related) complexity it travels through. The SAE levels cement the perspective of engineers and developers, in particular downplaying one specific point: in order to reach a fully safe Level 5 vehicle, software, sensors and hardware need to record the full details of a certain area. They thereby ignore the specificity of local mobility cultures and disregard the need to both, to account in a culturally codified, yet interactively interpreted way for the actions of others and give to accounts for one’s own ongoing or next actions (Garfinkel 1967: 10-34). Orienting automated vehicle development solely at SAE-Levels risks ignoring the plurality of culturally embedded and bodily performed mobility practices positioned beyond the regular scope of knowledge of technical actors developing such vehicles (Stayton/Stilgoe 2020).
More holistic discussions about the need (or expendability) of automating mobility in a certain socio-spatial context are overridden by a seemingly linear development of successive automation as suggested by the levels of automation. This linear succession leads to automated private cars as a somewhat natural top of the ladder. An alternative would be a modular approach which does not presuppose a linear progression with full automation at its end but a context-sensitive definition of degrees of automation, depending on the needs (or expendability) of local users, the vehicle type, and the ways these vehicles fit a given socio-spatial context with its particular mobility culture. But this alternative is at odds with the interests of globally acting companies and national regulators to find standards that are applicable everywhere, given that the automation of vehicles could lead to an unforeseen global standardization of driving norms, technical norms, and of the meaning of vehicles as new social spaces within smartified cities. However, this fear of global standardization may also be exaggerated, if I may draw a parallel between the smart city discourse and the discourse of vehicle automation. When comparing actual processes of smartification of cities like Songdo in South Korea (Bartmanski et al 2021) with the corporate narratives of smartification taken at face value, one recognizes the many contradictions and inconsistencies between claims of efficiency and standardization and local social needs as well as prevailing “not so smart” political and cultural attitudes.
Scientists, citizens and regulators are troubled by exaggerated marketing claims of cars’ actual digital abilities, a phenomenon called “Autonowashing” (Dixon 2020). Overtrust in advanced driver assistance systems (ADAS) – wrongly interpreted as fully self-driving cars – results in casualties, thus undermining a major argument for automation, that is, enhanced safety.
There are some obvious reasons for the fact that vehicles with ADAS are marketed as more than they are; the most obvious reason being the financially heated competition for a fully marketable self-driving and safe car among established car manufacturers (e.g., GM, VW, Toyota), big tech companies (Alphabet, UBER), and smaller specialized start-ups as well as university labs. Interviews with researchers and developers from the field of automated driving conducted by Tennant and Stilgoe (2021) illustrate this culture of overstating the abilities with the aim of signalizing scalability and attracting more (venture) capital:
“We have people in the industry who want to pump up the value of their companies, both big well-established companies and start-ups… One company goes out and makes an aggressive claim, and all the competitors have to make sure they’re not perceived to be left behind, so they’ve got to match it.” (Tennant/Stilgoe 2021: 10 citing an interviewee)
What does overstating the abilities of cars with ADAS mean in practice? It means that cars that still need full-time supervision by a driver sitting in the driver’s seat with hands ready to take over the wheel are labeled as “fully self-driving”, as Dixon (2020) shows for the case of Tesla. Teslas have a SAE level-2 driver’s assistance system which is called fully self-driving, is marketed as self-steering in urban environments and, finally, is misinterpreted publicly as an autonomous car by Tesla’s CEO Elon Musk and lots of Tesla fans in social networks.
You can find plenty of videos on YouTube showing crashes or near-misses with Tesla cars obviously being overestimated in their self-driving abilities. Autonowashing wouldn’t be so delicate if it did not lead to deaths and injuries on public roads, presenting additional risks for other road users and Tesla fans alike. It is equally clear that Tesla and other AV-companies are in need of data generated from real-world usage of automated cars to optimize their algorithms, AI and sensors. So there is a trade-off: between the chance to improve the AI of automated vehicles with real-world data and the risks occurring when society is put to test by new types of vehicles (Marres/Stark 2020) inscribed with a so far unknown mobility culture (Hoor 2021) and marketed with exaggerated claims of their abilities. However, Autonowashing is a major source of public distrust of a technological innovation claiming to be safer than human drivers and positioned at a threshold to be marketed on a bigger economic and spatial scale (Dixon 2020: 8).
(4) A car is always a vehicle, a vehicle does not have to be a car
A prominent image circulated by the car industry advertises autonomous cars with the dream of the businessman/-woman using their newly gained in-car-time productively – but this vision of newly won life quality directly reinvested in work fits the job profile of only 13% of the population (Fraedrich et al 2016: 67). Enhanced life quality is one of the most frequently named tropes connected to autonomous driving in public discourse, along with enhanced safety, the experience of driving and data security (Diehl/Diehl 2018). Aside from working, one could also sleep, read or relax in the car while traveling somewhere, or do nothing. Logically, the imagination of newly gained in-car-time as private time is based on the car as a private space. It is hard to imagine people sleeping, relaxing or simply enjoying the same privacy in a shared automated taxi as they would in a privately-owned or individually occupied automated car. As recent studies indicate, when given the choice, private automated cars were perceived as more attractive than shared ones (Kolarova et al 2018: 45).
The first point for now is that automation of car usage perpetuates the social habit of using private cars as an exclusively private space. Private car usage, no matter if self-steered or automated, contradicts most tenets of sustainability, anti-congestion policies of cities and a socially just way of re-organizing mobility and space.
For sure, the car industry is fixated on automating cars. Huge car companiesdo not want to amplify the already ongoing devaluation of the industry’s inherited main expertise: building car bodies in different variants, combustion engines and selling those cars to individuals through a differentiated network of subsidiaries all around the world.
But looking closer at what is going on, there are plenty of prototypes of automated vehicles: some designed to be able to change their social meaning over the course of a day, some intended as conventional cars only differing with respect to the additional automated driving function. The second point on the in-vehicle space I want to make here is that the automation of vehicles changes the perception and valuation of time spent in a vehicle (Kolarova et al 2018). This is also a reason for designers to imagine and experiment with new social meanings of in-vehicle spaces, activities and new products.
The last point is that automating vehicles changes the emotional relation of a passenger to a vehicle. Less care, attention, and cognitive-emotive capacities are needed to move oneself and the car through space; repairing an automated car seems difficult to imagine when speaking about a safe, fully self-driving vehicle. If automated cars are impenetrable for their users, the emotional bond between the vehicle and its “caring owner” loosens. If automation is connected to the idea of replacing individual car ownership (which is not a given), other human-machine attachments based on less possessive ties between passengers and a car (automated buses, shuttles as part of public transport, private robotaxis) become more realistically imaginable and are already being experimented with.
(5) Regimes of imagination – future(s) of roads and public spaces with automated vehicles
Road networks for car and bus traffic are not only the main corridors for motorized traffic, their shape and structure dictate the ways non-motorized traffic can muddle through space: cycling lanes are mostly built next to roads, pedestrians can move next to roads, they cross roads at traffic lights or put themselves in danger when crossing a big street. Parking lots, something car drivers take for granted and often use without paying, enable the freedom to drive anywhere and leave your personal car-space in public space, clearly at the cost of other usages of public space. It is not radical to state that non-motorized traffic is spatially subordinated to motorized traffic and that this hierarchy is impinging upon the ways public spaces in modern Europe are perceived, planned and used.
Thus, the last critique of the socio-technical project of automated mobility concerns its impact on the built environment, land use and the design of public spaces in the medium- and long-term. To put this in a nutshell: how would the automation of vehicles change the usage and design of roads and public spaces?
But, before I share some thoughts on this, I think it is necessary to take a step back. Obviously, the question “How would public spaces change with automated vehicles?” is devoid of context. For anyone interested in mobility issues, automation is a technological innovation which crosscuts ongoing re-negotiations about a socially just and climate-friendly distribution of public space, especially in urban areas. It is more than obvious that cars with internal combustion engines consume space, depend on an infrastructure of fuel provision, pollute the air, produce noise and have to be parked somewhere close to their owners. These dependencies of this instrument of freedom produce negative externalities. As long as the exposure of humans, ecology and infrastructures to external effects of car usage is considered to be a natural condition of a modern society, a decrease of everyone’s life quality in urban areas will be the downside of democratic mass motorization. Mimi Sheller and John Urry (Sheller/Urry 2000) have already described the ambivalence of a system of urban mobility based on movement in small privatized capsules: it produces coercion and flexibility at once; it confronts us with the political opposition of location-based political association and democratic mobilization of all social classes (with cars). All in all, car-based sociality is a central (sociologically underestimated) ingredient of modernity and a major producer of negative externalities. What is changing at the moment is that these externalities are increasingly considered to be negative externalities. They become the focus of public debate, political measures of citizen’s and (climate) activists’ anger (Sayman 2020). Any use case of automatized vehicles has to prove or at least indicate in advance that its present and future advantages outweigh negative effects of private car usage in the present. If producers and developers are not able to indicate automation’s advantages for a climate-friendly and socially just urban renewal, automated vehicles endanger to impact the basic structure of public spaces and notions of citizenship in a way that will instead promote a corporate-led, authoritarian digitization of spatial relations
So, this may seem like it is demanding too much. How should anyone be able to prove that the promised advantages of an innovation which is yet to come 1) will materialize (creating safer, cleaner, cheaper and more comfortable conditions than private cars today) and 2) will be politically manageable in a way that does not worsen existing conflicts produced by the consequences of the present regime of the private car? In short: How can anyone at this stage “prove“ that automatized vehicles will be better than conventional cars and that they will interact well with other users and usages of public space? Going further, why should automated cars be the solution for the current regime of car-based mobility? Putting these questions forward leaves no place for justifying the need of automating vehicles with the argument of unstoppable technological progress solely, a frame often used in media reports on automating vehicles (Taddicken et al 2020).
It is not surprising that no one can give a clear answer to these complex questions, yet answering these questions is key to stabilizing expectations (Borup et al 2006) in relation to an uncertain socio-technical imaginary which would radically change spatial relations. So it is clear that any even remotely convincing attempt to stabilize expectations is very valuable for influencing the course of the future.
I will depict three ways in which this work of imagining the spatial effects of automatized mobility is currently done in the present: first, by depicting highly idealized images of future public spaces in renderings; second, by anticipating the spatial effects of autonomous mobility scenarios in data-intense computer simulations and, third, by confronting the socio-technical imagination of automatized mobility with the unresolved complexity of ongoing negotiations about the mobility-related order of public space.
Advertisements exaggerate, they claim more than what an actual product can in fact do, advertisements project the “„present+ a new product”“ into an unknown future and thereby perpetuate existing societal inequalities. This is all well known. In the case of automatized mobility, advertisements usually show renderings of public roads, public spaces and public life. What we see most often, is a digitally re-figurated mode of controlling movement in public space.
This visionary image of a big car manufacturer shows a re-figured world: fewer cars, more people’s space, green parking areas, almost no cyclists and different types of automated vehicles. The image shows a harmonious public life: cyclists do not disturb the traffic; there seems to be less road space seems to be less than today and, in turn, enough space for consuming pedestrians in turn; cars seem to be “downgraded“ to mere instruments of mobility. In this case, the ideal world of an automated future clearly has a different spatial form than current public spaces. Infrastructures are adapted to a fleet of automated cars, which does not necessarily have to be necessarily privately owned. All of these cars could be robotaxis, automated shuttles or rented cars. In case they are private automated cars, this image would probably underscores the amount of cars which, in fact, were needed to serve the demand of trips. Such colorful renderings are very common. They are attractive for mass media communication;, nonetheless, taken at face value, such renderings of techno-urban futures have a series of consequential shortcomings (Mélix/Singh 2021).
Renderings of automated urban futures assemble visual references from the present, the future, the local and the global, existing and yet to come technologies in one coherent image. This image, then, is circulated in expert communities, public herarings and mass media. It is used in the present to anticipate and campaign for a certain image of how the future will look like at the end of a process which is about to start. In this sense, these images are neither representations, nor fictions as the effects of the image itself are supposed to play an active part in achieving this future. In this way, renderings smooth out possible social conflicts and contradictions. Instead of asking how multi-modal mobility in urban spaces can be organized, a lot of renderings of automated mobility are solely focusing on the aesthetics of a new technology embedded in a refigured, highly mechanized urban space. Tensions and conflicts which may play out during the process of adapting existing public roads, spaces and mobility cultures to the requirements of an growing fleet of automated vehicles are ignored or treated as solvable by more technology.
The current system of mobility produces conflicts on a daily basis: conflicts about planning and financing streets and highways, subsidies for electric cars or e-bikes, about cycling lanes, fuel taxation, conflicts about what makes a good service in public transport or pricing parking lots. Against this backdrop, it is by no means realistic that western democratic societies will refigure spatially the way these images present it: without a democratic debate about the mobility needs of diverse social groups and the right way to achieve them by accommodating the existing order of public roads and spaces. In this sense, renderings of automated urban futures do not only smooth out the expectable conflicts which will occur when new technologies intervene in the present system of mobility and its contestations (Mélix/Singh 2021: 250); they also put laypeople in a passive role. It is difficult for laypeople to distinguish at first glance, whether these renderings are realistic representations or exaggerated fictions, whether they are planned realities already decided set to be realized or suggestions for future urban spaces meant as an invitation to add ideas, evaluate or criticize. If the latter was the case, then it would still be unclear how and where laypeople could bring in their voices and evaluations when confronted with the opacity of such images of the future.
Geographers, transport researchers and data scientists want to know: How would the travel behavior of people and the use of land for different purposes change, if the vehicle fleet on the roads was consisted of, let’’s say, 10, 50 or 100 percent of automated vehicles? Would people be willing to travel longer distances and/or use the automated car for trips they would not have made or they had taken the bike or bus for before (modal shift)? As there are currently neither 10, nor even 3 percent of cars are automated and as the technology is still in its infancy, geographers and other scientists use models to simulate the hypothetical impacts of automated vehicles on a city, region or even nation-state to calibrate adequate policy-measures. As such, modeling studies are another way to foresee by which dynamics will mark future spaces would be marked, if there were automated vehicles increase on the streets. But, contrary to the aforementioned renderings, which fix one desirable image of the future, models simulate spatial and social processes in time.
Far from being standardized, there is a plethora plenty of simulation software and approaches to modeling which differ in many respects (Soteropoulos/Berger/Ciari 2019: 33-38). In their review article, Soteropoulos and colleagues compare the assumptions, central use cases and results of 37 modeling studies. All simulation studies considered in this review model the link between the introduction of specific use cases of automated vehicles and corresponding changes in land use and travel behavior. Some of these studies simulate the traffic based on a realistic map of a city’’s traffic network, some just take a mathematical grid network as a spatial basis. Some consider the speed and acceleration of vehicles, some assume a constant speed for each automated car. Some integrate traffic peaks at rush hours into their models, some assume a constant traffic flow. Some determine a constant waiting time for a car- or ride-sharing service; others use varying acceptable waiting times depending on the availability of alternative modes, such as public transport.
Some studies assume the availability of automated vehicles for children, adults without driver’s license and elderly people, others assume the availability of automated vehicles only for those with driver’’s license. Further, not only are the conditions of a traffic system (traveling speed, waiting times, social structure of travel demand) flexibly modeled, it also the supply side, which still remains uncertain: private automated vehicles, automated ride pooling taxis (ride sharing), automated buses as part of public transport or automated rental cars (car-sharing)? All 37 studies Soteropoulos and his colleagues investigated, vary in this respect.
In a nutshell: every study models its own world; assumes certain behavioral patterns and habits of those demanding automated vehicles; models different products on the supply side and assumes different conditions of the overall traffic system. As a consequence, these studies are not comparable, and their results are highly dependent on the assumptions made and the city or region which is sought to be reconstructed within the model.
However, there are some trends which point to the same direction. Even though when modeled for entirely differently spatial contexts, it usually turns out that privately owned automated vehicles most often lead to an increase in vehicle miles travelled, which signifies more traffic, especially in the city centers. The same applies to shared automated vehicles.. Only in the case of a large share of travelers willing to rideshare despite the increase in travel time caused for picking up other travelers, modeling studies show that automation can reduce the overall vehicle miles travelled. Yet the authors of the study finally concede that simulations do not cover the entire complexity of socio-spatial dynamics in a city, as their models only roughly distinguish urban from suburban areas.
3. Finally, I would like to present some voices from civil society, professional associations and committed citizens, journalists or bloggers, who engage with the socio-technical imagery of automated driving from a critical point of view. From their perspective, highly idealized renderings of automated public spaces and roads represent a future, in which mobility will be left to big companies, start-ups and the car industry, all of which (supposedly) do not care much about providing a user-oriented, eco-friendly, and spatially encompassing mobility supply.
Take for example the Bundesverband Taxi und Mietwagen e. V., one of the leading associations of taxi companies in Germany. In a press release from 2016, their president Michael Müller claims, that automated taxis cannot replace the service of a human taxi driver, who is able to carry disabled, old, or sick passengers into or out of the car.
The Association of German Transport Companies (VDV) warns against an unregulated introduction of private automated vehicles as well as privately managed robotaxi fleets. Facing Confronted with a growing shortage in available manpower to drive trams, buses, and subways as well as having difficulties in sufficiently serving rural areas sufficiently, the association of public transport agencies presents itself as an institution that is ready to incorporate automation in their existing structures and businesses,; and as an institution that is ready to co-produce legal regulations and practical knowledge with other actors. Further, public transport agencies fear that a growing motorization with automated vehicles could fuel the existing spatial dominance of the car. In their opinion, this could lead to an excessively cheap supply of rides with robotaxis, more short trips with cars, empty trips (of robotaxis), more congestion, less space for non-motorized means of transport or public transport – all in all, a scenario that is marked by automated, yet car-based dominance on public roads.
Finally, let me take a short turn for the perspective of cyclists. Alike pedestrians, cyclists do not carry a protective iron cage around with themselves. But unlike pedestrians, cyclists often have to share the road with motorized traffic, which exposes them to a high risk. 90 percent of all car incidents are caused by human failure, and cyclists are often the ones who bear the consequences of these errors. Given this constellation, the interests and hopes of cyclists towards in relation to automated mobility are quite different from those of car -drivers, public transport agencies, pedestrians, and users of public transport, notwithstanding and are moreover independent from the car industry. First, automated cars lack the ability to communicate the way car drivers do. Eye contact, hand signals or nodding are just some examples of the usual ways car drivers and cyclists give an account of what they are about to do. But these indexical expressions can vary in their shape and detailed execution from cyclist to cyclist, from pedestrian to pedestrian and driver to driver. It can also happen that human road users are uncertain and do unintelligible things. This is why automated cars, in turn, can be highly uncertain in safety-critical situations when they are tasked with identifying culturally coded, yet contingent expressions with their perceptive apparatus. The same applies to lane markings or generally, bleached out lane markings or traffic signs.
Obviously, an automated vehicles still lack certain perceptive capacities for identifying non-standardized, yet, for humans, clearly legible signals, and are in need of new perceptive and communicative instruments and habits, which have to be comprehensible for cyclists, pedestrians, and drivers. The same applies vice versa. Training algorithms on public roads may optimize the planned and executed behavior of vehicles but, as Jack Stilgoe (2018) convincingly argues, machine learning has to go hand in hand in with social learning; that is, other road users learning how to understand, respond to, and communicate with automated vehicles. This adaptation of road users to the behavioral patterns of automated vehicles, in turn, should create a new input for the recalibration of their AI and so on. As we learned from the modeling studies, most experts in planning and transport expect that it will take a lot of many years until automated vehicles will dominate vehicle fleets by rates of over 50 percent. So the next years will be marked by a mix of automated and non-automated vehicles at penetration rates of 5 to 10 or 30 percent – depending on the socio-spatial context, regulations and economic feasibility of use cases. Naturally, automated vehicles communicate more effectively if they dominate the streets. If not, they can pose such a risk to other road users or show such ineffective performances that it may become expedient to reserve lanes only dedicated to automated vehicles. This in turn would minimize freedom of movement for pedestrians and cyclists.
Let me sum up the main points and end with some open questions. The automation of vehicles is a highly uncertain socio-technical project, a recently resurrected technological dream that already existed 100 years ago (Kröger 2014). We have seen that it is more appropriate to speak of automation than of autonomous vehicles. We also got an insight into what is called “Autonowashing”, a term coined to talk about overtrust of consumers in advanced driver assistance systems as a consequence of a hyped public discourse and misinformation by manufacturers. While the industry produces fancy images of future public spaces with automated vehicles, such renderings put laypeople in a rather passive role. Uncertainties continue, as researchers can only give clues as to whether automation will free up spaces in cities, since they cannot foresee the development (and adaptability) of a nascent car-technology-industrial complex. Consequently, being either pro or contra vehicle automation mainly depends on the way automated vehicles will be introduced in the future, and on path dependencies which are still to be settled and governed. So we are faced with many uncertainties which do not entirely allow us as social scientists to estimate benefits and risks of automation on secure grounds, but we can state that the public’s involvement/role in evaluating this technology is still very marginal (Marres 2020).
It is still not clear how consumers of automated vehicles should be appropriately educated, but what is even more important is finding a way to open up the process for the public’s evaluation in social, ecological and cultural terms. Mainly, social scientists should help address all the questions that still need to be answered, if vehicle automation is supposed to be a process including all those affected. Will automated vehicles mainly be private cars or shared vehicles (with or without shared rides)? Can cities decide for themselves which types of cars and business models they will allow? Will people be willing to share rides with strangers in a vehicle without driver? Will people be willing to pay more for an automated trip with no other passengers? If so, will these new services lead to a mobility culture with fewer or more cars on public streets? Will cities witness a growing fleet of automated vehicles as an addition to already existing private car fleets? How should cities proactively inhibit the growth of their vehicle fleets by, let’s say, promoting automated robotaxis? Who will pay for the huge infrastructural adaptations required to fully profit from the advantages of automation in economic, environmental and socio-spatial terms? How will public transport agencies compensate for a loss of customers in buses and trains? How will public transport agencies, cyclists and taxi drivers re-position themselves when faced with artificial intelligence as a competitor for road space, regulative privileges and financial subsidies? How can the AI of automated vehicles be opened so that laypeople can understand and evaluate it and regulators can intervene in the public’s interest?
The motivation to build and articulate stable and positive expectations for vehicle automation has come to an end with the end of the technological hype in recent years. This is also because larger parts of the public and civil society ask themselves who will profit from (or be negatively affected by) vehicle automation and in what way. It is precisely these social complexities and uncertainties that should be kept in mind for a realistic evaluation of the opportunities and risks of vehicle automation in a privately motorized society in need of a huge socio-ecological transformation. From a citizen-centered point of view, it is nothing less than the apportionment, accessibility and the cultural, social and political functions of public spaces being at stake when assessing futures of vehicle automation.
Volkan Sayman (email@example.com) is a Sociologist and Ethnographer. Using insights from Science and Technology Studies, he thinks critically about the shape, usage, and governance of the manifold infrastructures which gov‐ ern our daily lives. He is trained in a variety of qualitative methods. In the last four years, he has pub‐ lished on power relations playing out in knowledge and digital media infrastructures. Currently, he is investigating the refiguration of public and private spaces as a consequence of increasingly automated mobility systems, such as autonomous cars.
Bartmanski, Dominik, Seonju Kim, Martina Löw, Timothy Pape, und Jörg Stollmann. “Die Refiguration von Räumen durch smarte Apartmentkomplexe: Über Praktiken der Verräumlichung der südkoreanischen Mittelschicht”. In Am Ende der Globalisierung, herausgegeben von Martina Löw, Volkan Sayman, Jona Schwerer, und Hannah Wolf, 205–30. transcript Verlag, 2021. https://doi.org/10.14361/9783839454022-009.
Bieler, Cuno, und Sutter, Daniel. “Externe Kosten des Verkehrs in Deutschland Straßen-, Schienen-, Luft- und Binnenschiffverkehr 2017”. Schlussbericht im Auftrag von Allianz pro Schiene e.V., 2019.
Bischoff, Joschka, und Michal Maciejewski. “Simulation of City-Wide Replacement of Private Cars with Autonomous Taxis in Berlin”. Procedia Computer Science 83 (2016): 237–44. https://doi.org/10.1016/j.procs.2016.04.121.
Dewey, John. Die Öffentlichkeit und ihre Probleme. Philothek. Berlin: Philo, 2001.
Dixon, Liza. “Autonowashing: The Greenwashing of Vehicle Automation”. Transportation Research Interdisciplinary Perspectives 5 (Mai 2020): 100113. https://doi.org/10.1016/j.trip.2020.100113.
Diehl, Nazim, und Can Diehl. “Autonomes Fahren im Diskurs – Semantische Netzwerke und diskursive Regelmäßigkeiten“. In Kommunikation und Technik, herausgegeben von Florian U.Siems und Marie-Christin Papen, 325–38. Wiesbaden: Springer Fachmedien Wiesbaden, 2018. https://doi.org/10.1007/978-3-658-21537-8_18.
Fraedrich, Eva, Rita Cyganski, Ingo Wolf, und Barbara Lenz. “User Perspectives on Autonomous Driving A Use-Case-Driven Study in Germany”, 2016. https://doi.org/10.13140/RG.2.1.2539.4969.
Garfinkel, Harold. Studies in Ethnomethodology. Englewood Cliffs, N.J: Prentice-Hall, 1967.
Gollrad, Markus. “Baukastenmodell statt Automatisierungsstufen – Anknüpfungspunkte für die Regulierung automatisierter Fahrzeuge”. Recht Automobil Wirtschaft 8, Nr. 2 (2020): 103–9.
Hoor, Maximilian. “Öffentliche Mobilität und eine neue Mobilitätskultur – Grundlagen, Entwicklungen und Wege zur kulturellen Verkehrswende”. In Öffentliche Mobilität, herausgegeben von Oliver Schwedes, 165–94. Wiesbaden: Springer Fachmedien Wiesbaden, 2021. https://doi.org/10.1007/978-3-658-32106-2_7.
Kant, Immanuel (1996): Practical Philosophy. Groundwork of the Metaphsics of Morals. Cambridge, Cambridge University Press.
Kolarova, Viktoriya, Felix Steck, Rita Cyganski, und Stefan Trommer. “Estimation of the Value of Time for Automated Driving Using Revealed and Stated Preference Methods”. Transportation Research Procedia 31 (2018): 35–46. https://doi.org/10.1016/j.trpro.2018.09.044.
Kröger, Fabian. “Die Zukunft, die nie eintrat”. TATuP – Zeitschrift für Technikfolgenabschätzung in Theorie und Praxis 23, Nr. 1 (1. April 2014): 9–12. https://doi.org/10.14512/tatup.23.1.9.
Marres, Noortje. “What If Nothing Happens? Street Trials of Intelligent Cars as Experiments in Participation”. In TechnoScienceSociety, herausgegeben von Sabine Maasen, Sascha Dickel, und Christoph Schneider, 30:111–30. Sociology of the Sciences Yearbook. Cham: Springer International Publishing, 2020. https://doi.org/10.1007/978-3-030-43965-1_7.
Marres, Noortje, und David Stark. “Put to the Test: For a New Sociology of Testing”. The British Journal of Sociology 71, Nr. 3 (Juni 2020): 423–43. https://doi.org/10.1111/1468-4446.12746.
Mélix, Sophie, und Ajit Singh. “Die visuelle Refiguration urbaner Zukünfte: Zur Prozessualität von digitalen Architekturvisualisierungen am Beispiel von Hudson Yards in New York”. In Am Ende der Globalisierung Über die Refiguration von Räumen, herausgegeben von Martina Löw, Sayman, Schwerer, und Wolf. Re-figuration von Räumen. Bielefeld: Transcript, 2021.
Mietzsch, Oliver. “Öffentliche Mobilität finanzieren: Plädoyer für ein neues Verständnis der Mobilitätsfinanzierung im öffentlichen Verkehr”. In Öffentliche Mobilität, herausgegeben von Oliver Schwedes, 211–37. Wiesbaden: Springer Fachmedien Wiesbaden, 2021. https://doi.org/10.1007/978-3-658-32106-2_9.
Sayman, Volkan. “Problematische Themen problematischer Öffentlichkeiten. Ulrich Beck und Bruno Latour als Impulsgeber einer Soziologie öffentlicher Praktiken”. In Soziologische Phantasie und kosmopolitisches Gemeinwesen, herausgegeben von Oliver Römer, Clemens Boehncke, und Markus Holzinger, 309–31. Nomos Verlagsgesellschaft mbH & Co. KG, 2020. https://doi.org/10.5771/9783845288376-309.
Sheller, Mimi, und John Urry. “The City and the Car”. International Journal of Urban and Regional Research 24, Nr. 4 (Dezember 2000): 737–57. https://doi.org/10.1111/1468-2427.00276.
Soteropoulos, Aggelos, Martin Berger, und Francesco Ciari. “Impacts of Automated Vehicles on Travel Behaviour and Land Use: An International Review of Modelling Studies”. Transport Reviews 39, Nr. 1 (2. Januar 2019): 29–49. https://doi.org/10.1080/01441647.2018.1523253.
Soteropoulos, Aggelos, Emilia M. Bruck, Martin Berger, Alexander Egoldt, Arne Holst, Thomas Richter, und Zoltán László. “Automatisierung, öffentlicher Verkehr und Mobility as a Service: Erfahrungen aus Tests mit automatisierten Shuttlebussen”. In AVENUE21. Politische und planerische Aspekte der automatisierten Mobilität, herausgegeben von Mathias Mitteregger, Emilia M. Bruck, Aggelos Soteropoulos, Andrea Stickler, Martin Berger, Jens S. Dangschat, Rudolf Scheuvens, und Ian Banerjee, 75–106. Berlin, Heidelberg: Springer Berlin Heidelberg, 2021. https://doi.org/10.1007/978-3-662-63354-0_6.
Stayton, Erik, und Jack Stilgoe. “It’s Time to Rethink Levels of Automation for Self-Driving Vehicles”. SSRN Electronic Journal, 2020. https://doi.org/10.2139/ssrn.3579386.
Stilgoe, Jack. “Machine Learning, Social Learning and the Governance of Self-Driving Cars”. Social Studies of Science 48, Nr. 1 (Februar 2018): 25–56. https://doi.org/10.1177/0306312717741687.
Taddicken, Monika, Anne Reif, Julia Brandhorst, Janina Schuster, Morten Diestelhorst, und Lennart Hauk. “Wirtschaftlicher Nutzen statt gesellschaftlicher Debatte? Eine quantitative Framing- Analyse der Medienberichterstattung zum autonomen Fahren.“ Medien & Kommunikationswissenschaft 68, Nr. 4 (2020): 406–27. https://doi.org/10.5771/1615-634X-2020-4-406.
Tennant, Chris, und Jack Stilgoe. “The Attachments of ‘Autonomous’ Vehicles”. Social Studies of Science, 14. August 2021, 030631272110387. https://doi.org/10.1177/03063127211038752.
A similar point was made by Leonie Seng in a science blog article from 2017. Autonomes Fahren – Eine Frage der Ethik? Oder: Kant fährt Dein Auto gegen die Wand » Feuerwerk der Neuronen » SciLogs – Wissenschaftsblogs (spektrum.de)
The Berlin based Start-Up Vay uses tele-drivers at a certain location to bring their cars to their customers, who then steer the car wherever they want to drive to. Startup Tries Out Remote Operators Who Use Screens to Drive (caranddriver.com)
This is the initial study of the BASt on which current SAE-Levels were based and then, further developed: F83-Dokument (hbz-nrw.de).
A very recent and extraordinarily bizarre story happened in San Francisco, US, where Waymo trains its automated taxi fleet (with safety drivers). Inhabitants of a dead-end street mention around 50 Waymo cars a day driving the dead end until the point of no return and then making a u-turn out of the street again. The safety drivers of the car say that the AI of the car chooses this routeway and that they can’t do much about it. In a similar vein, Waymo’s statement reads: „In this case, cars traveling North of California on 15th Ave have to take a u-turn due to the presence of Slow Streets signage on Lake. So, the Waymo Driver was obeying the same road rules that any car is required to follow.” The technical script sending Waymo’s cars into this narrow dead-end street is obviously not intelligible for its inhabitants Dead-End SF Street Plagued With Confused Waymo Cars Trying To Turn Around ‘Every 5 Minutes’ – CBS San Francisco (cbslocal.com) .
A regional court in Munich forbid Tesla to use the designation of “Autopilot” or “Full Self Driving” in Germany. This is the first legal sanction for Autonowashing in Germany.
The EDAG-City Bot is a prototype of an automated vehicle platform adapting its meaning, shape and function in manifold ways over the course of a day. As such, it represents a striking example of the point that automation makes entirely new vehicle spaces possible besides beyond the standard private car. Vision – EDAG CityBot (edag-citybot.de)
 Negative externalities arise when the product and/or consumption of a good or service exerts a negative effect on a third party independently of the transaction itself. Any ordinary transaction involves two parties, i.e., the consumer and the producer, who are referred to as the first and second parties in the transaction. Any other party that is not involved in the transaction is referred to as a third party. Third parties bearing the negative consequences of transactions they are not involved in are subject to negative externalities. Popular examples for negative externalities are traffic congestion, air or water pollution.
 The private car regime, among others, obviously produces negative externalities (Bieler/Sutter 2019; Mietzsch 2021). As these negative externalities are noticed by those directly or indirectly affected and made visible in public discourse (Dewey 2001 ), negotiations and conflicts arise about who has to take responsibility concerning these externalities. In urban contexts, these points of contention are, first, the apportionment of public space for organizing traffic (competition between cyclists, pedestrians, public transport) and, second, more encompassing conflicts concerning the relevance and spatial apportionment for different usages of public space: traffic, politics, recreation, housing, the environment, children, culture, social contacts etc.
The president of the “Federal Association of Taxis and Rental Cars” refutes the claim that human drivers become obsolete with automated vehicles: BZP zu autonomen Fahren- gute Taxi-Fahrer werden immer gebraucht
Around two thirds of all bike crashes can be traced back to crashes involving cars. In 75 percent of these cases, car drivers are to blame. Unfallzahlen – das Fahrrad ist das tödlichste Verkehrsmittel | STERN.de
Prof. Noortje Marres’ project “Beyond the Lab: An Empirical Philosophy of Intelligent Vehicle Testing” starts in 2022. It will “study street trials of intelligent technologies as ‘theatres of accountability’: choreographed situations where the ability to take into account, to give an account, and to call to account, is shared, contested and negotiated between different road users.” Professor Noortje Marres (warwick.ac.uk)
See for James Bridle’s artist intervention using lane markings to trap a would-be automated car Meet the Artist Using Ritual Magic to Trap Self-Driving Cars (vice.com) and Autonomous Trap 001 on Vimeo
See for the talk of Vladlen Koltun of Intel, who frames autonmated driving as a research problem: „In the settings in which we have deployed such advanced high-dimensional perception systems based on modern computer vision, there is a second line of defense. If the system is uncertain, it can hand off its predictions, it can hand off some content to human imitators who can screen the content further and even if a mistake is made, mostly human lifes are not at stake, usually human lifes are not at stake.“ Vladlen Koltun: Autonomous Driving: The Way Forward (July 2020) – YouTube (4:31)