The Ethics of Automation
The responsibility for regulating driverless cars should vest anywhere but with the Motor Vehicles Authority
All of a sudden, so many of our appliances — televisions, coffee machines, bathroom scales, light bulbs — have grown smart and integrated themselves into an enormous hive-mind that miraculously keeps us fit, our home temperatures controlled and our lives organised. It seems we are on the threshold of a new age, but even at this early stage, the enormous strains that this fundamental technological shift will place on the world, are already becoming visible. We are struggling to deal with the vast amounts of data that our personal devices collect and its impact on privacy. If you layer on top of that, artificial intelligence and machines that are empowered to make our choices for us, the real impact increases exponentially.
Nowhere is this autonomy more evident than in the automobile industry. Google Cars have been driving themselves around the Bay Area for years now and even though they have been generally incident-free, stray accidents have caused disproportionate anxiety. Wiredmagazine featured a story last year, in which two engineers remotely gained access to a Jeep driving on the highway and shut the engine down by hacking into its telemetry — a scary demonstration of the additional risks that a connected future could offer. But it wasn’t until Tesla released an over-the-air update that miraculously allowed its cars to automagically park themselves, that the reality of connected driverless cars really sank in.
Conventional wisdom says we should regulate autonomous cars by seeing to it that they are capable of complying with existing laws — ensuring that they are intelligent enough to abide by traffic regulations and can stick to the speed limit. This approach, in my opinion, is flawed. Our motor vehicles laws were designed to guard against human failure — essentially, to protect us from ourselves. Laws against drunk-driving, using cellphones and over-speeding exist solely to see to it that when we take control of a powerful metal capsule capable of travelling at insane speeds, we don’t end up killing ourselves. To regulate intelligent, networked cars that are perfectly aware of each other’s location, speed and direction under the same, essentially human framework is pointless.
Instead of making autonomous cars behave more like us, what we should really be concerned about is how these cars are programmed to make decisions. Whenever I think of tough choices, I am reminded of Phillipa Foot’s ethics conundrum: the Trolley Problem. It describes a situation in which a tramcar is rolling downhill towards five people, tied to the tracks, unable to move and staring at certain death. You can choose to switch the trolley to a siding but if you do so it will run over an innocent by-stander. What do you do?
One approach would be do what causes the least harm — switching the trolley to the siding would kill one person instead of five. But would your decision be any different if there was a child on the siding — and if so, how many adult lives is one child’s worth? Isn’t there a moral difference between allowing people to die by your inaction compared to wilfully switching tracks to cause the death of a human being?
Autonomous cars will be faced with decisions like these every day. And while a fallible human at the switching yard can assuage his guilt by convincing himself that he only had a split second to decide, autonomous cars will decide based on pre-meditated risk-balancing programs, consciously designed by their manufacturers. I wonder whether these moral choices should be left to corporate whim. If we don’t intervene, our cars may end up being programmed to protect their passengers at all costs — even at the cost of the lives of innocent bystanders.
It is here I believe legislators need to focus their efforts. Cars of the future will take smarter decisions and will be able to based them on inputs from millions of sensors — in their chassis, from the roads they drive on and the vehicles they interact with. They will have the benefit of machine-to-machine communication, AI and big data algorithms that will allow them to simulate millions of potential outcomes in the time they need to take appropriate action. But even with that kind of assistance, their choices will only be as sound as the programmatic basis on which they are taken.
Car manufacturers are already making these choices for us and will continue to build programming into their vehicles. We’d do far better to establish an ethical framework within which they must operate, than force autonomous cars to comply with human laws.
If it were left to me, I’d take the responsibility of regulating autonomous cars away from the Motor Vehicles Authority and hand it off to the Ministry of Robotics.
This article was first published in The Mint under a column called Ex Machina on the interface between law and technology.