World Science Festival: Robot dilemmas and driverless cars that could kill you
Source: Amy Mitchell-Whittington

The panel looked at what our future may look like in terms of robot ethics. Photo: Supplied

There are four people in a trolley inside a mine hurtling towards a wall. They will all die when they hit the wall.

You can either do nothing and let them die or you can flick a switch and make trolley change course and save the four people, but kill an innocent person standing on the new track.

        The key problem is that we actually have an urgent need for these systems to be able to make sound decisions.



        Matthias Scheutz

What would you do, and is there an outcome that is more morally right?
The Moral Math of Robots at the World Science Festival.

The Moral Math of Robots at the World Science Festival. Photo: Abbey Goodman

This is the kind of moral dilemma that philosophers have looked at for a long time and was asked of the audience at The Moral Math of Robots signature event on the World Science Festival's final day on Sunday.
Advertisement

Tufts University School of Engineering's Computer Science Department Matthias Scheutz said he has applied this trolley dilemma to groups of people and found there was a differing of opinions, depending whether a human or robot was at the switch.

"If a robot is at the switch, people find it more morally wrong for the robot to not do anything than to act, but when it is a human on the switch it is the opposite," he said.




Engineer professor Mattias Scheutz said decision making capabilities in robots was necessary to prevent human harm. Photo: Supplied

"This result has very interesting implications because it means that we actually have to do something about the robot.

"The easiest decision would always be to do nothing then you don't have to think through, you just don't act but because there is an expectation of the robot to act it will actually have to think through the consequences and understand the moral dilemmas."

The question then of how, why or even if a robot should understand moral dilemmas and make decisions on the 'best' course of action formed the basis of the panel's discussion, led by ABC's Catalyst presenter and science communicator Graham Phillips, which included engineers, philosophers and scientists.




Rob Sparrow said we should start to think about the humans responsible for creating robotic ethics. Photo: Supplied

Georgia Tech College of Computing regent's professor and associate dean for research Ron Arkin was on the panel and spoke of the inability for true morality and moral reasoning in robots and said they were "made to behave" morally or ethically, consistently with our own expectations.

"There is no approach that any lab is currently using that is trying to endow robots with moral agency so that they become fully responsible for their own choices and their own actions," he said.

"That is not feasible now or the future, certainly due to levels of complexity.



Ron Arkin discussed the use of robots at The Moral Math of Robots panel. Photo: Josh Meister

"The responsibility for these decisions always rests on human beings, whether that is the designer, or the user or the commander or whoever it is there is always a human being responsible, the robot didn't do it.

"Even the one that pulled the lever on that trolley problem, it is not culpable even if it makes a decision that you think is morally incorrect."

Monash University philosophy professor Rob Sparrow was also on the panel and was emphatic to point out the human responsibility for the decision making of robots, which was a complex issue that needed to be seriously considered.

"Ethics is a matter of personal responsibility it is a matter of reflection, it is something that can cripple a human life and it is something that occurs between people," he said.

"It is not a nice little rule book you can look up and think 'Oh yeah, this is what I have to do' and put it into a device.

"I also think we should recognise this as a surveillance technology, which it is, this is a device where people are exerting power over you.

"Other people, perhaps people you have never even thought of, are making choices about how these machines should respond and they are shaping your life.

"One thing we need to be thinking about is our ethics and the relations between engineers and the populations."

Professor Scheutz said having a decision making capability in a robot was a necessity, pointing to the issue of driverless cars, which are not too far away from becoming a reality.

"Imagine a self driving car on our streets that has to face decisions where it makes a quick decision how to act," he said.

"You might have someone crossing the street in front of it and it cannot break in time, what should it do? Should it still attempt to break or should it swerve to the side where in this particular case swerving to the side may cause it to crash into another car?

"Potentially regardless of what you do, the outcome could be bad so the question is how could it assess which action it should perform."

This assessment process would be based on algorithms that operate and allow the robots to make decisions formed by principles inserted into their system, Professor Scheutz said.

"We can try and encode legal principles to make sure the robot doesn't commit any transgressions on that end, for example if you have a self-driving car it has to obey the speed limits and so forth, that is programmed in," he said.

"Those are not invented by the engineers, those are just put in the system.

"The key problem is that we actually have an urgent need for these systems to be able to make sound decisions, the goal for us is not to develop moral persons, that is a discussion we will leave to the philosophers, the key is to create a system that in a scenario makes a reasonable decision doesn't freeze and stop acting."

Professor Sparrow was quick to point out the flaws involved in driverless cars and other robotic devices that would involve decision making where human safety was concerned.

"I have been wondering for a while whether people will get into a driverless car when they realise that if the car has the choice of crashing into another car killing four people or crashing into a tree and killing you that the machine is programmed to kill you, that is not really good advertising," he said.

"Notice what we are really talking about is a human decision. It is not really the machine that is killing you, there is an engineer somewhere who said 'well look if it is a choice between the sedan or the driver of the vehicle' then bad luck for the driver."