Driverless cars and the society of traffic

May 27, 2016

Last week’s Queen’s Speech set out bold new technological initiatives that seemed reminiscent more of science fiction than practical policy. In the first of our series of responses to some of the government’s proposals, Dr Eric Laurier, Reader in Geography and Interaction, and Deputy Head of RTD: Social Sciences Recruitment at the University of Edinburgh, examines the relationship between Artificial Intelligence and human interaction, and looks at the place of driverless cars in the social world of traffic.

The “driverless car” promises the appearance of fully autonomous vehicles and the disappearance of human drivers. In looking at how the mere idea of the driverless car is responded to by various social groups and institutions we can learn as much about societies’ anxieties and hopes for robots and other autonomous machines, as we learn about the realities of living with cars with increased forms and rights of agency. Yet it is the latter that the social sciences can now provide insight into by looking at the semi-autonomous cars that are already on our roads. These are cars whose technologies of autonomy vary from cruise control to autopilot.

In my field research, which is concerned with how people use technologies, we have just begun to explore how the activities of autopilots are, firstly, made sense of and, secondly, woven into existing driving practices. We are examining how autopilots are dealt with both by the persons in the car and by other drivers. By drawing upon real-time video recordings from inside cars with the new forms of autopilots, we are looking at when and how vehicles are allowed to drive themselves.

At this early stage in the arrival of a new technology on the roads, we are studying drivers establishing just what a self-driving car can and cannot do in the social world of traffic. Drivers are themselves equally fascinated by where an autopilot succeeds, where it fails and where sometimes it appears to exceed human capabilities. For example, autopilots successfully overtake cars on the motorway, they learn their driver’s route home from work and drive it for them. And yet, current car autopilots occasionally follow the wrong lines on the road, their sensors struggle with heavy rain and their complex models of driving interaction incorrectly identify the next actions of other human drivers.

The major problems that continue to be faced by semi-autonomous cars can be understood by turning to the longstanding tradition of social scientists examining how AI systems struggle with social situations where humans are trying make sense of the AI’s actions. Perhaps the most famous of these cases is the anthropologist Lucy Suchman’s study of photocopier help systems. Failures in understanding while using a photocopier with an AI system which would have been easily sorted out by human-human interaction became ‘fatal’ in human-machine interaction. While only a few people have died at the hands of photocopiers, many people die everyday on the roads through interacting with the fatal ballistics of the car, so making sense of AI action in traffic has much higher stakes. Suchman showed that a helpful way to understand the problem was that the photocopier had severely limited access to the unfolding of the social situation compared to a human participant.

It is, then, that same limited access to the social situation which continues to be a major obstacle for driverless vehicles. To a transport modeller, the birds-eye view of traffic on the roads shows flows of vehicles that appear to be obeying mechanical principles. As a result, of all our social situations, driving in road traffic seems to be the easiest environment to introduce an autonomous machine. However, mobility researchers like myself have shifted perspective down from the birds-eye to consider the view from the road itself. For the driver with their view of the vehicles alongside, ahead and behind, driving in traffic relies upon their and other driver’s (and sometimes passenger’s) mutual ongoing production and recognition of actions accompanied by the repair of misunderstandings of those actions. In the car, we show when we are about to turn across traffic by edging our car nose toward the middle of the road. Sometimes, though, we find ourselves positioned that way at the end of some other action in which case, knowing how our orientation may be misunderstood by arriving drivers, we will try and adjust the car’s nose slightly to show turning right is not our next action. Meanwhile, those other drivers will show their understanding by continuing past, or their misunderstanding by opening up a gap ahead of them. If the latter happens we will adjust our vehicle again to try and show what we are trying to do.

Social studies of car traffic on the road underline that it is anything but mechanical and is instead a moral order of the first degree where the breaking of even the more minor codes of conduct is accompanied by the sanction of car horns and angry gestures. On the moral terrain of the road, narrowing the gap behind the car in front is enough to incur the displeasure of the driver ahead. What the designers of the driverless cars can learn from social studies of driving in traffic are the details of not just how human drivers follow rules and how they break rules, but then how those same and other drivers (and passengers, pedestrians, cyclists etc.) detect rule breaking as intended or accidental, and deal with restoring social order afterwards.

While the self-driving car has much to learn from social studies of driving-in-traffic, it raises the fascinating possibility of the transformation of the social roles of the driver and their passengers. Already, in semi-autonomous cars, we see drivers puzzling over whether they have become a passenger while the system is in control and what they should then do with their time while the more humdrum driving labours are delegated to the car. A considerable part of our social lives already happens in the car: parents helping with homework on the way to school; officemates preparing for the day ahead at work during the morning commute; friends having heartfelt discussions about their relationships. These relationships have been interwoven with dealing with the work of driving the car but also drawing upon driving and the journey as a resource. A surprising transformation brought about by the driverless car might be that it creates an experience of being on the road together that is not as companionable nor caring as the time when one of us drove and the other was their passenger.

News Focus articles are the views of the author and not necessarily those of the Campaign for Social Science.

Photo source: youtube.com