Transcript

Emotionally intelligent technology

Audio

You can also listen to the interview [MP3 9.7MB].

Transcript

Matt Smith:

Welcome to the La Trobe University podcast. I would be your host Matt Smith and the field of robotics is a vastly evolving one. My guest today has been working on one aspect of their development. He is Associate Professor Rajiv Khosla, head of the La Trobe University School of Management Research Centre for Computers, Communication and Social Innovation. Thank you for joining me today Rajiv.

Rajiv Khosla:

Thank you.

Matt Smith:

You've been working specifically on emotionally intelligent technologies. What are these?

Rajiv Khosla:

Traditionally, computers have been designed on laws of physics and they have been what we call information providers. In the twentieth century we thought we'll develop these computers and information systems, human beings will use the information to solve all their social problems. But that hasn't happened. Today, what we have are changing family structures and ageing population, environmental problems, a globalised world where the boundaries of organisations are breaking down. So, what we are motivated by is how is technology going to address these social problems? Especially problems relating to the ageing society, to health care and to sustainability of organisations. What we need today is social innovation. And in our attitude to social innovation, we have fundamentally changed the design of computers from just being information providers, to actually being social innovators. That means we actually to take the technology to the people, break down the technology barriers where technologies can actually solve social problems. And in that context, what we are doing is developing an emotionally intelligent system. Because what we are trying to do, especially in the area of advanced robotics, is that we are working with communication robots and we are embodying human characteristics and human attributes in these robots – human emotions, human voice, touch, and a range of other human characteristics, so that they can engage with, for example, elderly who have never interacted with technology, in a human-like manner or what we call an emotionally intelligent manner and what we mean by that is, while they're having a dialogue with the elderly, they can not only recognise their verbal responses but they can also recognise their emotions, and then adapt the dialogue, keeping in context their emotions and the other non-verbal responses.

Matt Smith:

So, it used to be at a point where the technology would rely on humans for the input, but it sounds like emotionally intelligent technology can detect the input itself – it doesn't need the information from us, and it's actually giving us the results.

Rajiv Khosla:

Yes, that's right. For example, when the elderly interact with Matilda, the communication robot we're working, they don't have to use a keyboard, they just engage with it as if they are engaging with another person. What we are trying to develop in Matilda is what we are calling personhood, and these human attributes are part of that personhood.

Matt Smith:

You've done some testing at the moment with the robots in hospital. How are they going, and are they proving to be effective?

Rajiv Khosla:

Yes, what we have done is, we have done trials in nursing homes, we've done it in Northern Queensland and in Melbourne, and the results have been extremely positive. The biggest achievement we've had is that Matilda and Jack have helped us to break down the intergenerational barriers. They have interacted with elderly who have never interacted to technology in their life. They have played bingo and ??(3.20), they have played quizzes, had a conversation in the persuasive dialogue with the elderly to help them improve their diet, so the feedback we have received – it has helped the people reduce social isolation, it has helped personalise care because it frees up the care givers, for example, when Matilda plays bingo, the care givers can go to each resident, many of these residents suffer from different conditions, and help them actually with the game. So it's helped to personalise care, when they have one to one interaction with Matilda, it can help them to feel a sense of usefulness – they can send emails to their relatives, Matilda can also read news – you know, it gives them that sense of independence.

Matt Smith:

When you've been using these robots in testing, are they being used for their actual healthcare, or just to give the elderly something to interact with?

Rajiv Khosla:

Both the things. What we are interested to find out, what's the impact of Matilda on the overall wellbeing of the elderly, and wellbeing includes healthy living, their healthcare, their social interaction, one of the broader aspects relating to ageing is that you want to be mentally active, and engaged. Not only that, but these games are at different levels of complexity. We can have games which are like pattern matching games, which can determine issues related to dementia and Alzheimer's, early detection of these problems, because once we can have this early detection, we can design personalised therapeutic systems. We can then determine for this person what kind of activities need to be designed into Matilda. We are working towards a whole sum approach in terms of the wellbeing of the elderly.

Matt Smith:

Where does the emotional intelligence come into that in the healthcare system?

Rajiv Khosla:

Well, Matilda is interacting with the elderly who may be suffering from obesity or diabetes or other medical conditions, while it is having the conversation or a dialogue, it's detecting their emotional responses also. So, for example, if it detects that the elderly is feeling anxious, it will come out and say "Oh, I see that you are looking a bit anxious. I'm here to help you. We can have the conversation now or we can have it another day." In other words, it is factoring in the emotional context in which the elderly is interacting, just like we do in face to face interactions. We continuously adapt ourselves based on their non-verbal responses. So that's one example. And now we are measuring emotions to use official expressions. Actually what we are measuring is changes in the facial expressions. We are not about determining whether somebody is in a happy state or a sad state, you know, those stereotypes, but what we uniquely determine is whether they are in a positive emotional space, or in a negative emotional space. We monitor those changes. Another thing we are working on now is the speech acoustics, that is from the tone, we want to make out what is the emotional state. We are trying to integrate these multiple sources of information when Matilda interacts with the elderly so that we can enhance the reliability of emotionally intelligent communication.

Matt Smith:

So what other applications of the technology are there? You've been using emotional intelligence in applications like road safety, but also just with the robots, you've been using them for job applications.

Rajiv Khosla:

That's right. The motivation behind designing and developing Matilda for conducting job interviews is related to the sustainability of organisations today. One of the important aspects of sustainable organisations today is the extent to which the organisations can retain their employees, because the cost of recruiting a new employee is sometimes six times more than the cost of retaining an employee. And if you want to retain an employee, in the first place you need to get the right employee. What's the definition of a right employee? And so what we are trying to do is, that Matilda actually finds out how much passion a person has for a job, it develops an emotional profile, and the cultural profile, and then benchmarks with the respective team the candidate will possibly work with. So in other words we are trying to define what is the definition of the right salesperson for example. Or the right employee, not only for the particular job, but how this employee will fit into the culture of the organisation. So, it's a very entrusting and a unique application of Matilda, because I mean, you may ask, why not do you conduct interviews using a PC? Why do you have to use a robot? Again, it comes down to social interaction. The rate of social interaction a candidate has when they are interacting with a robot, doing their interview, is much larger than to do it with a PC. The reason is because Matilda, when it is asking questions, it has gestures, it also has expressions. When the candidate answers the question it acknowledges with a nod, with an expression. As the interview progresses, the human attributes of Matilda come to the forefront and you have the candidate totally involved, as if it's a real interview. That enhances the whole communication and the quality and the reliability of the information which is coming in terms of their emotional and cognitive responses.

Matt Smith:

That's an interesting term that you use. You said it's a human interaction with the robot. Now the design of the robot I find, it looks very much like a child's toy almost. It's about two feet tall, it's very colourful, but at the same time, you haven't tried to make it look too human.

Rajiv Khosla:

You may be interested to know that Matilda has got a baby face and it's 40cm tall. And it weighs about six, six and a half kilos. And it's got universal acceptance across practically all the age groups – the children, the young adults, as well as the elderly. The reason is that not only because the baby face has universal appeal, but it's the whole size, structure and colour looks non-intimidating. That's very important, because traditionally people associate intimidation when they think about robots. Because of its size and appearance and the different colours, it's non intimidating. And more so, it's got all the touch sensors, so people can kind of cuddle it, it's got all the human voice and other attributes. So, the total package looks appealing. The reason we have gone with this research and we have this incredible community support for this, is that Matilda is generating a good feeling among its human partners. That is very important for our research. And similarly with the kids, we are thinking, how can we motivate them in terms of education? In the primary and secondary schools, and in the tertiary sectors. Because it's got a way of measuring the emotional responses and engaged in emotional intelligent communication, so it can help children in terms of their language, it already speaks two languages, Japanese and English, it can help them through their maths skills for example, general knowledge skills. It's a way it can take them off from the telly.

Matt Smith:

Will it do their homework for them?

Rajiv Khosla:

It won't do their homework for them. That's not the intention.

Matt Smith:

If you asked nicely?

Rajiv Khosla:

Well, over time, because we are developing personhood, you never know, because it will personalise itself to the needs of the human partner, so it depends what space you put it into and what you want out of it.

Matt Smith:

That's all the time we've got for the La Trobe University podcast today. If you have any questions, comments or feedback about this podcast, or any other, send us an email at podcast@latrobe.edu.au. Dr Rajiv Kholsa, thanks for your time today.

Rajiv Khosla:

Thank you.