How to design an emotionally intelligent machine: in terms of visual aspect considering ‘Uncanny valley’ theory.
As an exchange student, I may have awkward expressions in English writing. I would appreciate it if you consider this point while reading.
Considering the recent technological developments and industrial trends, there is a movement to create emotionally intelligent machines. For example, there are Replika, Amazon Echo or Google Home.
Then, when we create intelligent machines that interact with people, what do we have to think about in terms of actual visual design rather than just a response? In cognitive science, studies have been made that most of human cognition comes from visual stimuli. Vision is the most dominant sensation compared to other senses such as audition and somatic sense. Therefore, it is important to design machine that can emotionally interact with people in terms of visual experience.
In many cases, however, people are reluctant to interact with intelligent machines. Let’s look at Sophia, the most recent humanoid robot.
Do you really want to be friends and emotionally interact with Sophia? Of course, there is an idiom, ‘don’t judge a book by its cover’, but at least I can not do it. But even though you can’t do either, it does not mean that we are bad people. This is a phenomenon that people feel naturally.
This can be explained by the theory in robotics and psychology about the emotions that humans feel about robots or non-humans, “Uncanny valley” . To briefly summarize this theory,
“If you try to imitate it so equally, you will have more discomfort”.
Let’s look at the graph above. The empathic response increases as the character resembles a human. However, when it reaches a certain degree, the response declines rapidly. And once you reach a level that is indistinguishable from humans, the empathic response will increase again. Thus, we instinctively feel rejection of objects that have a similar appearance to humans but perform other behaviors.
In fact, when the 3D animation Polar Express, which described characters similar to real human beings through motion capture, was released, many of the children who watched the movie cried because they were scared.
On the other hand, The Incredibles or Monsters, Inc., which appropriately deformed or used non-human characters even in the same 3D, received good evaluation, because of this Uncanny valley phenomenon. Polar Express’s characters were created by attaching trace sensors to the entire body of an actual actor, and the facial expressions, movements, and reality of the characters appeared to be more real than the other works, but it was rather a fatal weakness.
Given this theory, what should we keep in mind when developing an emotionally intelligent machine? If you cannot imitate perfectly, you should not try to be too close to humans. And since it is impossible to imitate humans perfectly with our current technology, it is necessary to deform and dehumanize properly so that people do not feel displeased with the intelligent machines.
I want to show you an emotionally intelligent machine that I think is appropriate. The video is about JIBO, a social robot for the home, which has a design that reflects the uncanny valley theory with proper deformation and dehumanization, so that people do not feel uncomfortable.
Let’s consider companion animals such as dogs or cats that we are currently interacting with emotionally. Even if they are not very similar to humans, we can interact emotionally with them. Therefore, when developing an emotionally intelligent machine, it is important to create a machine suitable for the human cognitive system, which has no sense of discomfort considering the Uncanny valley theory.
