Self-aware Networks
Computer scientist Erol Gelenbe on the idea of AI self-regulation, the goal function of a network and whether ...
The video is a part of the project British Scientists produced in collaboration between Serious Science and the British Council.
I am an experimental psychologist at heart. It is a young subject. It started at the end of 19th century. And cognitive psychology is even later, it started in the middle of 20th century and replaced a behaviourism as the main way of studying people. The ‘cognitive’ in cognitive psychology refers to the fact we are using computers as a metaphor or information processing. So we talk about memory and storage. We talk about perception and templates. And social cognition arrived even later, at around the end of 20th century. And initially, as the name implied, it was about implying the idea of cognitive psychology to social psychology. And in the early days, this meant things like how do we recognize emotions in people’s faces.
And that has two components because we could be talking about people’s fleeting moods: are they angry, are they sad, are they disgusted. Or we could be talking about people’s dispositions: is this person trustworthy? is this a nice person? And that of course leads to things like prejudices, race prejudices, dehumanizing people, and that sort of thing. Another aspect of social cognition at that time which became very interesting to everybody with a discovery of something of biological motion which was discovered by Johansson in Sweden where he did his very clever experiment. He took people, he attached lights to the joints so you have a light on your wrist, on your elbow, on your shoulder. And then you can film people where all you see is the lights, you don’t see the people. Just from these very minimal information you can see whether person is working or running, whether they are male or female, whether they are happy or sad. And this became a technique for experimentally studying one aspect of social perception, if you like.
The key difference between cognitive psychology and behaviorism was that cognitive psychologists wanted to know what was happening in your mind, not just looking at behaviour. So, they were concerned with what people were thinking, what they we introspecting, and, for example, how they solve problems, how they thought about problems in order to solve them. Cognitive psychologists were very interested in memory attention and perception. And, for example, you can show that how well you perceive something depends on whether you are attending to it comes out of the blue. You can ask what is it that people attend to when they’re confronted with the complicated scene.
One of the things that was shown in social cognition was that the first thing we attend to is faces. If there is a face in the picture, that’s what we look at first. Indeed, I remind that there was a famous Russian psychologist called Alfred Yarbus who studied eye movements very early on in this field. And he showed how people actually look at faces in painting and so on. So, that was the way that social cognition developed.
There’re various questions that we study in social cognition which are still no really answered. So, one question is: is social cognition something special? Are the ways that we interact with other people different from the ways that we interact with physical objects in the world? Are the ways that we attend to people different from the way we attend to trees or animals or something like that?
It’s been a great deal of work in social cognition on the recognition of the emotions in faces. In the early days this depended on basically taking photographs or actors representing emotions. And it turned out that there are roughly five different emotions, that everybody can recognize across cultures, what is happiness, sadness, disgust, surprise and one other. And there are always problems with this, because, for example, we recognize happiness by a smile. And it turned out that in this research that there are actually two kinds of smile: a genuine smile and something that is known as a Duchenne smile which refers to the particular muscles involved, which is a sort of a fake smile. So people are very worried about using actors as a basic material for presenting the emotions because unnecessarily extremely good actors may be producing the fake version. Now, more recently, a chap called Alex Todorov in Princeton in the last twenty years or so has managed to produce computerized versions of faces. So, you don’t need an actor, you actually have what they call a face-space in which you can present different faces and you have dimensions. The way that faces generated is based on the real muscles in the face. So you’re actually simulating the expressions that can be produced by human faces.
One important aspect of the face is how trustworthy it is. You can show here is that you can easily generate a face that looks trustworthy and a face that looks untrustworthy. Everybody would agree, that this face looks trustworthy and this face looks untrustworthy. But this almost has nothing to do with reality. Someone with the face like this is not necessarily actually trustworthy. It;s just what we have this social consensus about what a trustworthy face is. And interestingly, you can speculate and this is something that is not yet fully conformed. If you have an advantage to be born with a trustworthy face, you can talk advantage of this.
Social cognition is very important because it directly concerns our everyday lives, and not just everyday lives, but I mean social life in general. Social cognition is not relevant simpler to how I talk to you at this moment, how I am responding to your micro expression as I talk to you. But also there is a much larger scale: how do groups interact, how do societies work. And something that’s obviously particularly concerning me this moment: why on Earth did the UK vote to leave the European Union? Social cognition will enable us to understand how some of these mechanisms actually work.
Computer scientist Erol Gelenbe on the idea of AI self-regulation, the goal function of a network and whether ...
Historian Alessandro Scafi on the Medieval knowledge about spherical form of the Earth, compass points on medi...
Neuroscientist Diana Omigie on tuning systems in different cultures, congenital amusia and aspects that influe...