“Can Robots Read Our Minds” Dr Alessandro Vinciarelli - TEDxUniversityofGlasgowSalon
Okay, I hope anybody can hear me. We are trying to answer in the few minutes we have at this position about this question; can a robot - or in general, a machine - read our mind? We start with a little game. You see a picture here; can anybody tell me what these two people are doing? Discussing, talking - what are they doing?
Shouting, arguing; any other?
If I ask you the type of relationship they have - friends, colleagues, student and supervisor - what type of-
The funny thing of this game is that there is no right or wrong answer here. What is interesting is that the picture is very poor: We don't know these people are. We cannot hear their voices. We don't know what they say. We cannot see their faces. And still we managed to give, confidently, a very high-level interpretation of the picture, in terms of social and psychological phenomena. So how this can happen? In general, our attention tends to focus pretty much on what people say, but there is a lot of nonverbal communication going on whenever there is social interaction.
Here you see examples of what we call nonverbal behavioural cues. We as humans are very much sensitive to this cues. We perceive them and we interpret them in terms of social signals; so in terms of the relational messages that people exchange with one another. Let me go a bit deeper into this phenomenon, because it is really at the core of the answer we are going to give to our question.
When we talk about nonverbal behavioural cues we talked about essentially three elements; the first are the nonverbal behavioural cues that we have just seen in the picture. Here you have a list; it is very long, but it is not exhaustive. There are so many nonverbal behavioural cues that the psychologists have grouped them into five major classes - The Codes - and this is the second important element of nonverbal communication. Let's go one by one:
- The first is physical appearance; everything contributes to the way we look like, which means:
I think we all heard at a certain point in life that "It is not important how you are outside; it is important how you are inside." It's false. It has been investigated extensively; if you look nice, if you look good, your life is going to be much easier.
- Then your face and head behaviour:
- Facial expressions. No need to explain this - you are happy, you smile.
- Use of gaze; when you want to establish social contact with someone, the first thing you do is to establish eye contact, gaze contact.
- Then all the little movements we can do it our head - shaking, nodding etc - that conveys social meaning.
- Voice behaviour. This may look contradictory; what has speech to do withnonverbal? Well in speech, there are not only words. So when we talk about nonverbal vocal behaviour we talk about everything in speech except words:
- The very sound of your voice
- The intonation
- The use of vocalisation like crying, sobbing, laughter and so on.
- Even the use of silence is a non verbal of behavioural cues in speech.
- Gestures and postures. when I talk about gestures here I'm not talking about things like this or this, that have a very precise meaning. What I'm talking about is more this type of spontaneous gesture; that I use, for example, to accompany my speech. This type of reaction like touching your head when you are discomfortable and so on.
- About posture; well I'm talking to you, I'm oriented towards you. I can give this talk in this way; you can perfectly hear me because there is a microphone; the content will be the same, the slides would be the same; but the impression you will develop will be totally different.
- Finally, space and environment. Think of this room; depending on the role you play this evening, you stay here or you stay there. Think of the place where you work; in general, the more you are important the bigger is your office. So space §has a social function.
Here we come to the third important element of nonverbal communication. Nonverbal communication is a natural phenomenon; and like everything in nature, it has a function. Here we have a list of the functions that the psychologists have identified; some of the most important. Let's go by them one by one:
- Forming impressions. You see that I'm not behaving the way I behave with my friends at a cafe. I wear my best jacket; I try to talk in a smart way, because I'm trying to convey a certain impression about myself.
- Expressing emotion, I think, no need of explanation; you are happy you smile. You look at a person, and somewhat you can manage to understand immediate if that person is sad/happy amazed/worried and so on.
- Same thing, relational messages; this is what we have seen in the picture at the beginning of the talk. It's the type of attitude that people display to one another; are you aggressive, are you §, are you friendly, are you open and so on. Very interesting - Mother Nature considers to be very important - lying effectively and effectively detecting the lies of others. Now think of nature; how many animals lie to survive? Camouflage is lying. Standing on the posterior legs to look bigger in front of a predator is lying. In our human society we have more sophisticated type of strategies to lie; however we do it, and we have to be good at doing it when you do it, and good at understand it when others lies to us.
- Similarly, power and persuasion. Look at a group of animals; there is always an alpha individual, and it is very important that everybody else understands where they are in the hierarchy. If you don't understand who is above you, it can be very dangerous; 30% of the monkeys die because they try to challenge established authority. In our human society, we do not risk to die so frequently - at least here, fortunately - however, still for us it's very much important to understand when you talk with someone that is above us.
This is a very general introduction to nonverbal communication; now the point is, what has this to do with the question at the centre of this talk; what have this to do with the question whether a robot or a machine can read our mind? Well, it has to do in this way; what does it happen to this if I replace the person on the right with a machine? A machine means anything - any apparatus that has "sensors" - and a minimum of processing power. So the technological questions are, can I automatically detect nonverbal behavioural cues and can I automatically make sense of nonverbal behavioural cues?
The answer is yes in both cases; technology produces nowadays technological approaches that can detect the nonverbal and behavioural cues that have been mentioned, and they can make sense automatically of them in terms of social & psychological phenomena. Exactly the same way as we have been doing - when we have seen the picture at the beginning of this talk and we've figured out - we have guessed that those two people are singing, laughing, arguing, were being lovers and so on.
So when we go to the question we posed at the beginning - can a robot read our mind? - the answer is yes. Why it is so? Very simply because nonverbal communication is the physical, machine-detectable evidence of social and psychological phenomena. When we think of our inner life - of our mind, of our psychology - we tend to think of it as something abstract; something that we cannot really access with our senses. This is true; however the same inner life - the same psychology, the same mind - leaves traces in the physical world in terms of nonverbal behavioural cues, and here is the key for the machines to access it. If we can perceive the cues with our senses, then we can detect them with sensors and then we can understand that with artificial intelligence approaches.
Now, what is very much cool for the people like me that work on this type of technologies is that it is about people as much as it is about machines. It is very much interesting, you don't need only to know computer science, machines, mathematics etc; you also need to learn a lot about people, about psychology, about human sciences. What is a bit frightening about this technologies is that nowadays the progress is mostly machine-oriented; what does it mean? In this moment, we exploit the knowledge of we have about people to build machines that are more and more powerful at doing what we do every day - understanding and reading the mind of others - irrespectively of any effect that this can have on the life of people. In a sense, we are exploiting only one side of the lecture; we are learning that we cannot build a social intelligence machine if we do not understand people. So now the progress is machine oriented; we exploit the knowledge we have of people simply to build better machines. It would be nice to do it the other way around.
We cannot understand people if we cannot build social intelligence machines. What does it mean? These machines are a tool of unprecedented power to understand how people work; and it is very much an important to understand how people work, because if you understand it you can improve their life substantially. Let me give you a couple of examples that involve the school of computing science where I work. We are using a robot to teach autistic adults how to develop better social skills. This is crucial for them to make their life happier and healthy; in this way we better understand autism works, and through that we manage to improve the life of these people. Similarly, we use these technologies to detect as early as possible when children are mistreated by their parents. Why is this important? If you understand it early enough, you can increase a lot the chances of saving these children and making their life as adults happier and healthy.
So from this point of view we are on a very nice way, because we have the possibility to make technology that improve the life of people; but for what concerned § both the positive sides of these technologies, and the more worrying side of this technology? We should never forget that when we talk about artificial intelligence - because this is what we are talking about -"artificial" should be that big and "intelligence" should be that small. There is still a lot of work for us that work on improving the life of people for this; and still we don't need to worry much about the very negative effects, that potentially these technologies can have. Thank you very much.
--- END OF TRANSCRIPT ---
“Can Robots Read Our Minds” Dr Alessandro Vinciarelli - TEDxUniversityofGlasgowSalon