A car is driving when suddenly everything changes. The choice is stark – plough ahead into a pedestrian or swerve and slam into a wall. It’s a nightmare scenario. Who should die, passenger or pedestrian?
This is the kind of dilemma that is being increasingly considered as machines arrive that will take split-second decisions for us. Autonomous vehicles are of particular concern to many people. Heavy and potentially fast, they must mingle with the human world and make choices that may result in injury or death.
“For the first time in history, we are building devices endowed with the ability to make autonomous decisions that have moral consequences,” says Iyad Rahwan of the Media Lab at the Massachusetts Institute of Technology, one of the authors of a study in Science today that addresses this scenario.
Advertisement
The greater good
The team has carried out a series of surveys of hundreds of people showing that most want autonomous vehicles programmed to make choices for the greater good, that is to preserve the most lives – choosing to put the life of their occupant at risk instead of ploughing into a group of school children, for instance.
But that view changes when people are asked whether they would buy such a “greater good” car. Most say they wouldn’t. It’s a classic case of what is good for others is not good for me.
It is tempting to suggest using laws to ensure that the greater good prevails, regulating all autonomous vehicles to make decisions that preserve the most human lives. But the survey suggests that such regulation could put consumers off buying autonomous vehicles. A product that will sacrifice its owner’s life to preserve the lives of strangers isn’t something most people will choose.
The authors worry that this may end up slowing the adoption of driverless cars, ultimately limiting the promise of this technology to dramatically cut road deaths.
Black and white
That would be a bad outcome, but we should be careful before we start worrying about these hypotheticals. The scenarios posed in the study are too simplistic to reflect the real world – there are never just two options; there is always some space to aim the crashing car into, room to brake, some uncertainty about the outcome.
To present these dilemmas as black and white stokes fears that are unfounded and harmful.
While it’s useful to ponder what kind of world we want to build, we should be careful not to directly conflate hypothetical scenarios with real life. The authors do acknowledge the need for more intricate decision-making scenarios.
There are other ways to look at this problem. Those inside a car always hold the upper hand in a scenario involving potential pedestrian casualties, and so should accept greater risk in such circumstances. A pedestrian is more likely to die from being hit by a car than passengers are to die from swerving to avoid a pedestrian.
To think that we have complete control, even of something as technologically advanced as an autonomous car, is folly. Chaos and randomness will always show their hand. Even if our cars behave perfectly, pedestrians will never (hopefully) be under the control of our engineers.
There will always be uncertainty in the world. Worrying about hypothetical scenarios where all uncertainty disappears risks doing more harm than good.
Journal: Science, DOI: 10.1126/science.aaf2654
