This month, we spoke with anthropologist Melissa Cefkin, Principal Scientist and Senior Manager at Nissan Research, about the challenges of creating autonomous cars that can move smoothly in a human landscape.

Have you ever navigated a four-way stop in the United States, or passed another car on a narrow road anywhere in the world? Road use, whether engaged as a driver, cyclist or pedestrian, demands actions stemming from an encounter with another, however fleeting and unremarkable it seems. Further, we often try to take those actions in the “right” way for that time and place. We read the situation: “Does the other driver see me? Does she seem to be waiting? Are we both seeing the situation the same, that she arrived first and I should wait for her to go, or does she think I arrived first? She’s paused a touch longer than I would have expected. I guess I’ll go.”

How will an automated vehicle (AV) respond to such situations? Who goes first is not necessarily noteworthy in any single instance. But eventually many AV could be moving through the same places one after another. Given that movement through the streets is such a fundamental part of our daily existence, as we scale up these encounters, the implications of any given choices could matter to how familiar, efficient and navigable the roads feel.

At a more fundamental level however, the extent to which people will encounter automated vehicles as more machine-like or more human-like has yet to be determined; it will necessarily evolve over time. And let me say one other thing: we have to consider what “autonomy” really means. We have to recognize that it is an abstraction, it does not exist in an absolute sense. Autonomy only has meaning, conceptually, vis-a-vis another controlling force. People are said to be autonomous when they live free from the control of another governing entity. A machine system is said to be autonomous when it can take actions independent of the control of a person. The route towards autonomy is to enable the vehicle to act on its own goal by goal; driving down a single lane, and then adding a lane change, and then traversing an intersection with lights, and so on. Vehicles already do take actions ‘on their own’ in ways we hardly recognize, such as anti-lock braking systems. The interplay with people will remain, we just have to see exactly where and in what ways.

Who is (or will be) doing this teaching and how is it being done? Is this already a focus of autonomous car makers?

A common approach in the industry is to use machine learning or deep learning. Through the high volume of data gathered by onboard sensors and feeding that data through machine learning algorithms, vehicles are being designed to figure things out for themselves. Another approach is model-driven — making models of human behaviour based on a priori views of people’s actions. At Nissan we are using multiple approaches, including empirical studies driven through social science theories to inform our designs. Machines may become pretty good at identifying similar and different actions and even anticipating patterns over time. But they don’t have a means to understand and judge in social terms about the contingencies and meanings of the patterns, which people use to assess and anticipate actions. Looking at road use through this lens should help us both deepen and improve, and also hasten, the learning of the machines. More importantly, we are able to sensitize the development teams as to what to expect in ongoing development, and how to prioritize focus areas.

Can we actually teach autonomous cars human behaviour or do you think, as an anthropologist, that there are big challenges autonomous car-makers are bumping into?

Of course there are challenges and there is a bit of hubris in describing the effort in this way. They will never be human.

Nor should we necessarily want them to be. Hopefully they’ll never gun through an intersection when a light changes — a big waste of energy. They won’t be fumbling for their cell phones or get distracted by a crying baby.

My view is that we should be examining today’s practices and expectations not with the aim to recreate them exactly. Society is ever changing, after-all, especially as technologies are introduced and continue to be adapted. Personally I’m pretty happy about the changes afforded by graphite and the printing press but somewhat less enamored by the overwhelming effects of the attention-grabbing screen and button of the smart phone. We should be examining today’s practices and expectations to see where fissures and opportunities lie and what that could mean for development tasks. But we should also be using our analysis to raise questions about the potential broader social impacts. This is and always has been the special value of ethnographically informed approaches at work.

Human behaviour will also have to change with autonomous cars on the horizon. What changes do you foresee and how can service designers anticipate them?

I have been referring to our everyday experiences of road use, to how the feel of moving through time and space could shift, if but subtly. Indeed the automobile has been an exceedingly dominant force in shaping our urban landscape, our economic systems, even our social structures. Are we on the cusp of shifting that dominance? There are a plethora of possible implications of automated vehicles for cities, for public spaces, for government services, for family livelihoods — the list goes on. There are tremendous opportunities for rethinking the possibilities of urban spaces and how they are used, for instance, or service models for transportation broadly as time goes on. Nonetheless, given where we are today, it seems to me that most likely near term opportunities lie in services related to commercial transport and urban public transportation.

How/why will service design become even more critical in the automotive industry? What should service designers focus on in the automotive world?

I agree, it is likely to become more important. With increasing autonomy, cars have the potential to realize more fully their eponymous function, to serve as a means for the conveyance of people and things; for moving goods and people about, for providing access to employment and to friends and family, for getting from place to place whether for pleasure or necessity. Mobility as a service. The degree to which vehicles continue to be owned versus accessed remains to be seen. There is of course a great deal of focus on the increased potential for shared use and shared ownership models. Vehicles may in the process also begin to lose something of their potential to act symbolically in other ways, such as markers of status or identity. Or it could go the other way.

As to what service designers should focus on, I can tell you what I would like to see. What I would like to see is that the service design community plays a role in keeping the focus on the systemic aspects of the service systems, to focus on publics and how things work together and not just singular business opportunities, of which there will be many: services related to mobility itself, of course, but also data services and don’t forget the care side of the picture and the maintenance and upkeep of the autonomous vehicle systems.

A long-time observer and participant in the growth of anthropological research in and with business, Melissa is the author of numerous publications including the edited volume Ethnography and the Corporate Encounter. She has served as president and conference co-chair for the Ethnographic Praxis in Industry Conference (EPIC), where she will be a keynote speaker in October this year, and recently on the US National Academies of Science committee on Information Technology, Automation and the Workforce. A frequent public and academic speaker, the work she and her colleagues are doing has been cited broadly, including in PBS, Wired, the Financial Times, the Estonian national press, CNN Chile, Canadian Broadcasting, and beyond. Melissa holds a PhD from Rice University and has been a Fulbright scholar; she worked previously at IBM Research, Sapient and the Institute for Research on Learning (IRL).