-
Who would you save, the pedestrian in the
road or the drivers in the car?
-
It's not easy, and yet that's the kind of decision which millions of autonomous cars
-
would have to make in the near future. We programme the machine but who do we tell it to save?
-
That is the set-up of the moral machine experiment.
-
There are so many moral decisions that we usually make during the day we don't realise.
-
In driverless cars, these decisions will have to be implemented ahead of time.
-
The goal was to open this discussion to the public.
-
Some decisions might seem simple - should the car save a family of 4 or a cat?
-
But what about a homeless person and their dog instead of a businessman?
-
Or how about two athletes and an old woman instead of two schoolchildren?
-
The problem was that there were so many combinations, so many possible accidents,
-
that it seemed impossible to investigate them all using classic social science methods.
-
Not only that, but how do people's culture and background affect the decisions that they make?
-
The only option we had really was to turn it into a viral website. Of course, it's easier said than done, right.
-
But that is exactly what the team managed to do. They turned these situations into an online task
-
that people across the globe wanted to share and take part in.
-
They gathered almost 40 million moral decisions, taken from millions
-
of online participants across 233 countries and territories from all around the world.
-
The results are intriguing. First, there are three fundamental principles which hold true across the world.
-
The main results of the paper, for me, are first, the big three in people's preferences
-
which is save human, save the greater number, save the kids.
-
The second most interesting finding was the clusters, the clusters of countries with different moral profiles.
-
The first cluster included many western countries, the second cluster had many eastern countries
-
and the third cluster had countries from Latin America and also from former French colonies.
-
The cultural differences we find are sometimes hard to describe because they're multidimensional,
-
but some of them are very striking, like the fact that eastern countries
-
do not have such a strong preference for saving young lives.
-
Eastern countries seem to be more respectful of older people, which I thought was a very interesting finding.
-
And it wasn't just age. One cluster showed an unexpectedly strong preference for saving women over men.
-
I was also struck by the fact that French and the French subcluster was so interested in saving women.
-
That was, yeah, I'm still not quite sure what's going on here.
-
Another surprising finding concerned people's social status. On one side we put male and female executives,
-
and on the other side we put a homeless person. The higher the economic inequality in a country,
-
the more people were willing to spare the executives at the cost of the homeless people.
-
This work provides new insight into how morals change across cultures
-
and the team see particular relevance to the field of artificial intelligence and autonomous vehicles.
-
In the grand scheme of it, I think these results are going to be very important
-
to align artificial intelligence to human values. We sometimes change our minds.
-
Other people next to us don't think the same things we do. Other countries don't think the same things we do.
-
So aligning AI and human moral value is only possible if we do understand these differences,
-
and that's what we tried to do. I so much hope that we can converge, that we avoid a future
-
where you have to learn about the new ethical setting of your car every time you cross a border.
