Page - (000084) - in Autonomes Fahren - Technische, rechtliche und gesellschaftliche Aspekte

Why Ethics Matters for Autonomous Cars72 In these and many other cases, there may not be enough time to hand control back to the driver. Some simulation experiments suggest that human drivers need up to 40 seconds to regain situation awareness, depending on the distracting activity, e. g., reading or napping – far longer than the 1–2 seconds of reaction time required for typical accident scenarios [38, 18]. This means that the car must be responsible for making decisions when it is un- reasonable to expect a timely transfer of control back to the human, and again braking might not be the most responsible action. One possible reply is that, while imperfect, braking could successfully avoid the major- ity of emergency situations a robot car may find itself it, even if it regrettably makes things worse in a small number of cases. The benefits far outweigh the risks, presumably, and the numbers speak for themselves. Or do they? I will discuss the dangers of morality by math throughout this chapter. Braking and other responses in the service of crash-avoidance won’t be enough, because crash-avoidance is not enough. Some accidents are unavoidable – such as when an animal or pedestrian darts out in front of your moving car – and therefore autonomous cars will need to engage in crash-optimization as well. Optimizing crashes means to choose the course of action that will likely lead to the least amount of harm, and this could mean a forced choice between two evils, for instance, choosing to strike either the eight-year old girl or the 80-year old grandmother in my first scenario above. 4.1.2 Crash-optimization means targeting There may be reasons, by the way, to prefer choosing to run over the eight-year old girl that I have not yet mentioned. If the autonomous car were most interested in protecting its own occupants, then it would make sense to choose a collision with the lightest object possible (the girl). If the choice were between two vehicles, then the car should be programmed to prefer striking a lighter vehicle (such as a Mini Cooper or motorcycle) than a heavier one (such as a sports utility vehicle (SUV) or truck) in an adjacent lane [18, 34]. On the other hand, if the car were charged with protecting other drivers and pedestrians over its own occupants – not an unreasonable imperative – then it should be programmed to prefer a collision with the heavier vehicle than the lighter one. If vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications are rolled out (or V2X to refer to both), or if an autonomous car can identify the specific models of other cars on the road, then it seems to make sense to collide with a safer vehicle (such as a Volvo SUV that has a reputation for safety) over a car not known for crash-safety (such as a Ford Pinto that’s prone to exploding upon impact). This strategy may be both legally and ethically better than the previous one of jealously protecting the car’s own occupants. It could minimize lawsuits, because any injury to others would be less severe. Also, because the driver is the one who introduced the risk to society – operating an autonomous vehicle on public roads – the driver may be legally obligated,