Australia's Chief Scientist

OPINION: If we want a future with driverless cars in it, we need to learn to trust the technology

“We as a community need to be confident we’ll be protected from unintended harm. Technology developers need to be confident they understand what communities expect, so they can work to make the better, safer, cheaper products we want.”

Dr Finkel has written a piece for the ABC on the ethics of programming driverless cars, and the decisions that will need to be made to ensure their algorithms reflect community values. The article was published on 6 July 2018.

The full text of the article is below.

 

Humans make a lot of bad decisions. We’ve baked it into the standard definition of human: “To err is human”.

So why do you walk so confidently across a pedestrian crossing in the expectation every single human driver will stop?

It’s not because you expect the people around you to be perfect. It’s because you trust in the customs and laws that restrain our bad behaviour.

And you also know that, for all our faults, there’s something important about our human capacity to feel.

We’ve all felt pain. We’ve felt the regret that comes from inflicting pain on others. We see another human, and we care.

Now, replace a human driver at the pedestrian crossing with a self-driving vehicle. What changes?

There are a lot of positives. A self-driving car can’t be drunk or distracted.

It’ll be a stickler for the road rules. You can be certain it’s gone through far more rigorous testing than the average P-plater.

But you won’t know how it’s been coded and trained to make decisions.

You won’t understand its processes of reasoning in the way you understand the thinking in a human brain.

The car is not a caring being: its sensors can register your presence and its algorithms can activate a learned response.

And if the algorithm determines knocking you down is the correct response, you know the car will do it without a twinge of regret. It can’t regret. It’s a technology.

So can you learn to trust it? Part of the answer lies in good regulation.

The purpose of regulation is ultimately to build trust in our ability to innovate safely.

We as a community need to be confident we’ll be protected from unintended harm.

Technology developers need to be confident they understand what communities expect, so they can work to make the better, safer, cheaper products we want.

Good regulation meets both aims: safety and progress.

We rightly look to our experts in government to specify the requirements for product standards and safety testing, just as we trust in government today to ensure the safety of everything from medicines to electrical appliances.

But the values underlying those standards come from the community. Do we agree with the Germans that all human life should be weighted equally?

Or do we think the owner of a self-driving car should have the ability to decide in advance what happens if a crash is inevitable: in effect, to prioritise their lives, or the lives of their children, over the lives of others?

No responsible technology developer wants to see the real promise of self-driving cars squandered by a lack of attention to safety, or a misinterpretation of community values.

Trust has got to be earned. We ought to trust in our capacity to harness this technology safely — and if history is our guide, we will.