Any talk of autonomous vehicles inevitably leads to a discussion of the old trolley problem. That is, if a trolley car was guaranteed to hit either person A or person B, but you could choose which person it hit (or decline to intervene), what would you do? In the hypothetical, person A and person B are usually assigned more specific traits, such as doctor, woman, man, child, dog, etc so as to induce a rationalization of the choice by considering person A and person B’s societal worth. Many believe this thought exercise is relevant because autonomous vehicles may need to be programmed to make similar decisions.

Instead of leaving this debate to ethicists and philosophers, a group of psychologists and computer scientists decided to pose the question to average people around the world. The results, published recently in the journal Nature, showed a preference by respondents for taking action versus doing nothing, saving people over animals, and saving children over older adults. Digging deeper, more chose to save a girl over a boy by a slight margin, but chose to save a male doctor over a female doctor – also by a slight margin. Dogs were slightly favored over criminals, but criminals were slightly favored over cats.

Cultural differences were also present. In eastern countries (Middle East, India, and China), respondents showed a weaker preference for saving younger people over older people while in southern countries (mainly Latin America) the preference for saving humans over animals was less pronounced.

While These Scenarios Are Unlikely, Autonomous Cars Will Have to Make Tough Choices

How relevant is any of this, though? As Kelsey Piper pointed out in a recent Vox article, autonomous vehicles are highly unlikely to ever be in a situation where they need to make this type of decision. Drivers are rarely, if ever, put in positions where they have to quickly decide whether to hit person A or person B. For obvious legal and public relations reasons, it’s unlikely any company would voluntarily introduce programming that allowed its cars to entertain most of the scenarios posed to the participants of the survey. But some scenarios are unavoidable. 

While drivers are rarely put in positions where they must choose between person A and person B, they’re often put in positions where they must take quick, reflexive action. Since the average, untrained human is often incapable of making rational choices that minimize damage in high-pressure situations, society (and thus laws and juries) is more forgiving of bad decisions.

But the same can’t necessarily be said for computers. If a deer or other obstacle suddenly appears in the middle of the road, it’s less excusable if the computer-driven car swerves into a pedestrian or comes to a sudden stop when safer alternatives exist.  Autonomous vehicles will almost certainly have to make some decision when confronted with sudden, unexpected changes to the environment.

This Could Mean Pitting Passengers vs. Pedestrians

That could mean prioritizing the driver (or passenger in the case of autonomous vehicles) over a pedestrian. In the survey, more respondents chose to save the pedestrian over the passenger, though western countries showed the weakest preference. From an American perspective, perhaps this makes sense. Americans, on average, have experienced more time as a driver or passenger than as a pedestrian on a busy road. The driver/passenger experience is therefore more relatable than the pedestrian experience and humans are generally inclined to choose self-preservation over self-harm.

Like most of the scenarios presented in the survey, the question of passenger vs. pedestrian is fraught with concerns over equity. With the possible exception of parts of New York City, most routine pedestrians tend to be lower-income individuals. Our inability to minimize costs under pressure coupled with our desire for self-preservation already puts pedestrians in vulnerable positions, but swapping a human driver for a computer changes the entire situation. No longer is the outcome preceded by the unpredictable actions taken by an irrational human actor, but by a programmed machine capable of quickly processing large amounts of data.

While respondents showed a weak preference for saving pedestrians over passengers, companies could opt to save the customer-passenger. They could argue that they’re simply programming cars to take the actions human drivers are inclined to take and have been taking. But that’s lazy calculus. A driver’s instinctive action to swerve away from a deer in the road and (unintentionally) towards a pedestrian is just the final action in a series of actions that created the danger in the first place. 

If brakes fail or a deer jumps in the road, the passenger of an autonomous vehicle is hardly to blame for harm caused by the vehicle. Neither, of course, is the pedestrian. But by introducing a heavy, fast moving vehicle, the passenger chose a method of travel that potentially produces more severe outcomes. In that sense, the passenger made the situation much riskier for everyone (including the deer) by introducing the vehicle. The passenger, while not the direct cause of the ensuing harm, is perhaps more responsible for the harm than the pedestrian.

Companies Could Avoid the Controversy By Investing in Good Urban Design

The passenger’s decision to travel by car is hardly a choice, though. Most American cities are designed for cars, not pedestrians. Many employers expect employees to have cars. Many transit-friendly and pedestrian-friendly neighborhoods are simply unaffordable for most Americans. The problem therefore lies much deeper than how an autonomous vehicle is programmed to react in high-pressure situations. Instead of attempting to deal with an ethical problem in its final stages, companies could address the base cause of the problem: poor urban design. 

Car companies now have a chance to redeem themselves from decades of blindly propelling highway-oriented urban development. They will have to deal with difficult scenarios where autonomous vehicles may have to decide the fate of innocent people. Society won’t accept these types of negative outcomes from calculated, programmed machines.

Companies could, however, make autonomous vehicles more appealing by reversing course and heavily promoting and investing in pedestrian-oriented development. Safer cities that are designed around protecting people would create space that not only ensures the safety of the public, but accommodates autonomous vehicles.