By Elizabeth Broadbent
Engineers are studying human behaviour in great detail in order to make robots that not only look like us, but can also understand us and interact with us in socially acceptable ways. These studies are teaching us many things about our own human nature, as my recent paper explains.
The robots in films like Blade Runner are very humanlike, with thoughts and feelings, motives and desires. But making robots that are just like us is a huge challenge. Technical limitations make it currently impossible to make robots identical to humans, although Hiroshi Ishiguru has made a geminoid (a humanlike robot that looks like himself), and David Hanson has made a number of impressive android heads. While these heads look very humanlike, their movement is limited, and robots are even more limited in their body and limb movements.
What researchers have found is that if a robot looks only partly humanlike or moves in a non-human way, it can make us feel very uneasy – an effect known as the Uncanny Valley. This theory states that as robots look more humanlike, we feel more comfortable with them, but when robots look partly human and partly in-human, they fall into the this valley, where people feel revulsion. Research suggests that this effect may occur due to category uncertainty, evolutionary adaptation to avoid diseased individuals, or the violation of perceptual expectations. Robots currently being used for healthcare or in shopping malls often take a more simple humanoid form (with a simple plastic head and body, and usually on wheels). This form is both easier to make and it has a greater chance of avoiding the uncanny valley.
For any robots to interact with us in our world, they need to behave in socially acceptable ways. Roboticists have studied how humans interact with each other in order to program robots to act in a similar fashion. How to approach a person is one simple example. From psychological research we know about social distance, which refers to how much distance we like to keep between us and other people. Psychologists have found that gender, age, and culture, as well as how much we like a person, can all influence whether one moves away from or approaches another person who is approaching us. Yet this knowledge was insufficient to know how a robot should approach a person. Roboticists therefore observed how people performed this task. After finding that people used an angled approach, the engineers programmed these same approach paths into the robot. This increased social acceptance of the robot, compared to its original straight-line approach, when the robot approached from behind.
As well as behaving appropriately themselves, robots need to accurately perceive and interpret human behaviours. An example is how to detect whether someone is trustworthy or not. Engineers have studied human dyads for behaviours that elicit trust, using manual and automated coding schemes. Less trustworthy behaviour included face touching, arm crossing, leaning back and hand touching, whereas an open armed posture, leaning forward and having the arms in the lap indicated more trustworthy behaviour. This adds to previous knowledge about deceptive behaviour from psychologists like Paul Ekman, and enables engineers to build more perceptive robots.
Interacting with robots
Turning the tables, research on how humans react to and perceive robots can also add to knowledge of ourselves. One such topic is the perception of mind. We appear hard-wired to perceive that all creatures , even robots, have at least some aspects of mind. People think robots have a moderate degree of agency (such as the ability to plan, remember, communicate, recognise emotions, and think) but they have less ability to experience things like emotions, hunger, pain, and consciousness. The more humanlike the appearance of the robot, the more we attribute mind to it. Yet the attribution of mind is not all about humanness, as babies are perceived to have strong abilities to experience yet little agency.
Sadly, our tendency for violence and aggression also emerges in our interactions with robots. In the absence of their parents, groups of children abuse shopping mall robots – calling the robot names, punching and pushing it, and blocking its path. One positive take is that by altering the robot’s responses to abuse, we can learn more effective strategies for combating abuse. In the shopping mall example, avoidance and escape was the best strategy for the robot, as altering the robot’s verbal behaviour or making the robot gently push to continue its path did not work. Young boys left alone with a small wheeled robot also abused it by covering its eyes or verbally abusing it. If the robot responded with a sad and fearful face this increased the abuse, whereas if it adopted an angry face the abuse reduced.
On a more positive note, we also express friendship and companionship with robots, patting and hugging pet-like robots. We even experience empathy when we watch a robot cutting its finger or losing its memory. And research suggests that robots that make encouraging comments are seen as better friends than robots that make neutral comments. So robots can even teach us how to make better friends.
If you think making robots is all about the robots, you are only half right. The other half is about studying and finding out about humans.