Machines Becoming Moral

Roboethics - Learning from Human Moral Development

In the last ten years, the challenge of equipping machines with some form of moral competence has developed into a whole new field of scientific enquiry called Roboethics. Leaving aside for the moment what is motivating this explosion of interest in this subject, a key question for many researchers in this field is what can we learn from human moral development that can be translated into the moral development of machines.

If we are going to look at human moral development as a model for robot moral development, then perhaps we should think about the starting point for humans. In particular, we should seek to understand how much, if anything, of our human moral competence we are born with, and how much we learn as we grow into adult human beings.

Humans are Born to be Pro-social

Child moral development has been studied extensively by developmental psychologists, and what they tell us is that we are born both with pro-social tendencies and a basic ability to differentiate between goodness and badness. Both of these innate skills constitute the elementary foundations on which a child’s subsequent moral development is based.

It is well known that babies are drawn to other people, that they are naturally pro-social. They like the look of human faces and the sound of their voices. It is also known that from an early age babies have expectations of their interactions with other people, and that they can become agitated or upset if those interactions don’t meet their expectations. Moreover, modern psychology has shown that babies can have a surprisingly sophisticated understanding of the minds of other people.

Babies are Mind-readers

Experimental psychologists Renee Baillargeon, from the University of Illinois and Kristine Onishi from McGill University demonstrated that fifteen month old babies can anticipate a person’s behaviour based on that person’s false beliefs. In their experiment, a baby would observe one person putting an object into a box, and then another person removing the object from that box and putting into another box whilst the first person wasn’t looking. When the first person turned back to pick up the object, the baby expected them to reach into the box where the object was originally placed (i.e. according to that person’s false belief that the object was still in that box), rather than the box it was now in.?

You might ask how do psychologists know that babies have this expectation? How do they know that this is what the baby is thinking? They know it through what are called ‘looking time’ studies. What babies look at is an indication of what they are attending to or focussed on. The longer a baby looks at something, the greater the significance of the thing they are looking at for that baby. As one psychologist described it, the?movement of the eyes are the “windows to the baby’s soul.”

In Baillargeon and Onishi’s study, the babies consistently looked at the box where the object was originally placed when the person comes to pick it up, indicating that that’s the box they expected the person to reach into, even though the babies knew the object was no longer in that box. These studies demonstrated that babies already have a remarkable level of understanding of the minds of other people from a very early age. And it reveals that they have an astonishing ability to recognise at some basic level the intentions of others.

Babies are born with an elementary moral compass

Psychologists also tell us that babies are born with a basic awareness of wrong and right, or goodness and badness, and are disturbed when things don’t align with their expectations. In other words, babies are born with an elementary moral compass. To study these early moral foundations in babies, psychologists devised a series of looking time studies based on moving geometrical objects. In one scenario, a red ball is shown trying to go up a hill. Then either a yellow square or a green triangle appeared. If the yellow square appeared, it would get behind the red ball and gently push it up the hill – this was the helper. If the green triangle appeared, it would get in front of the red ball and push it back down the hill – which was the hinderer.

Next, the children would be shown the red ball either approaching the yellow square, or the green triangle. What the researchers noted was that three and nine month old babies looked longer when the ball approached the green triangle (hinderer), than when it approached the yellow square (helper). The conclusion of these and other studies was that very young children show a sensitivity to goodness and badness, and that the?sensitivity to badness develops earlier and is more powerful than the sensitivity to goodness.?

A Tough Act to Follow!

It is clear, therefore, that very early on in our development we have quite a sophisticated understanding of our social environment together with a basic awareness of goodness and badness. Both of these traits are foundational to our moral capacity. These traits are already far in advance of any robot or AI system that has been developed to date by Roboethicists. It feels like we need to go back to the drawing board.