So… Your car wants to kill you

By Matt Tank

Technological change is rarely smooth or consistent. When the iPhone was launched in 2007, it changed the trajectory of mobile communications overnight. Even so, we don’t always realise how sudden the change will be, even when we can see it coming.

Such is the case of Autonomous Vehicles (AVs). Most experts, including CEOs and heads of innovation at GM, Ford, Volkswagen, Audi, and of course Tesla, have agreed that fully-autonomous vehicles will be available in the next two years, and we will be able to completely do away with driving by about 2025. However, many people still think that they will be driving themselves around for a while yet.

It is likely that once vehicles have the ability to drive completely independently, near-100% adoption will be reached extremely quickly. People who can afford it can buy a new driverless vehicle, and for those that can’t, legislation mandating the use of autonomous cars (for reasons of public safety) will likely push lower income earners into becoming transport-as-a-service users (similar to public transport users now). They will be able to use a driverless car service to transport them, at a fraction of the cost of the current equivalent service.

Moving forward, the primary burden of responsibility for motor accidents will move from the driver, to the automotive company, and individually, trust in the ability of the car to avoid accidents is something we will need to build over time. Common sense and statistics tell us that this will happen eventually. Driverless cars will never have momentary lapses in concentration, road rage won’t be an issue, and multiple layers of redundant systems will improve their ability to interact with their environment. Already, there is data that suggests that this is the case. Earlier this year, the US National Highway Traffic Safety Administration reported that the crash rate for Tesla cars has dropped by 40% since the introduction of their autopilot system.

Central to the safety of these vehicles will be the choices that they make. Things that may seem obvious to us may need to be planned for in car systems. Questions like “do I swerve to avoid this feather?”, “Is it OK to run over a ladybug… or a cow… or a baby?”. To a computerised system, there is no right answer until you tell it one. This will be the easy part. The hard questions will be the ones where there is no right answer. The people designing these systems will ultimately be responsible for working out what is the “least wrong” answer.

Consider the following scenario: On the footpath is a person walking their dog. An empty AV is driving in the opposite direction to your car. As you round a bend, an elderly couple step on the road in front of your car. A collision is inevitable. Does your car: a) swerve onto the footpath, hitting the dog-walker? b) brake hard, but still collide with the couple?, or c) swerve onto the other side of the road, hitting the unoccupied AV, but almost certainly killing its own occupant (that’s you!).

Many involved in the development of AVs will say that there will be no situation where “a collision is inevitable”, but regardless, the decision-making ability of car systems will need to be pushed to their absolute limit, so questions like this need to be asked. So, what factors come into play in the scenario above?:

  • Who is “at fault” in this situation? The elderly couple perhaps?
  • Is the number of people affected (injured or killed) important?
  • Is the dog accounted for? What value does it have, compared to the human?
  • Is the value of a human life constant? Is a child “worth more” than a pensioner?
  • How does the (predicted) outcome affect decision-making? Death?… disablement?… broken bones?… a blood nose?

…and rather cynically:

  • What will result in the lowest liability cost for the manufacturer?
  • Who is a customer of the car company (or transport service)? What outcome results in the most (and least) future revenue?

It’s worth noting at this point that the car will take into account situational information, but may also be able to gather other data. By the time the decision is made, the car may know who all the parties are, what their medical history is, what they spend their money on, and whether or not they have a family. So does the decision change if one of the parties is UN Secretary-General, or homeless, or a non-citizen?

As I mentioned before, there is no easy answer, and there will be many different points of view. Personally I would advocate that values need to be taken out of the decision-making as much as possible, and go with a general rule of thumb like “Do everything possible to avoid a collision whilst remaining within the boundaries of your driving path (ie, stay in your lane), unless deviating from your path carries no risk of death or serious injuries to humans (swerve if it’s safe)”.

OK, so maybe it wasn’t that hard at all (and I’m sure there are no alternative viewpoints that people could put forward in the comments section…). It does raise one last question though, which is “how on Earth do we know what will actually be programmed into our cars?”.

In the never-ending battle between profits and the public good, profits often come out on top. If you don’t think that a company will sell you a product that will kill you while telling you otherwise, the 1970s’ tobacco industry would like to have a chat with you. The systems that control AVs are likely to be closed systems, heavily protected against intellectual property theft by the companies that developed them. These organisations are unlikely to ever tell people how they work – only that they “are safe”.

I think it will eventually fall to international standards organisations, and government to develop guidelines around crash avoidance, and enact them into law. This work needs to start now, and be designed with current autonomous technology in mind, but with a view to a future where vehicles, with access to a huge amount of data, are solely responsible for your safety on the road.

Driverless cars will soon be something we take for granted, along with the convenience of being able to use your commute time more productively, the virtual elimination of car parks, and a near-zero annual road toll. Autonomous vehicles will rapidly improve their ability to anticipate issues, and the number of unavoidable accidents will decrease. Until that time comes, we need to make sure we are prepared for the changes that this technology will introduce.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s