DR RAJIV KHOSLA sees a time in the not-too-distant future when
robots will be helping students learn, robots that are 'emotionally
intelligent', that can recognise and respond to the visual clues
that humans display in their reactions.
Khosla believes the robots that he and his team at La Trobe
University in Melbourne are developing with colleagues at Kyoto
University in Japan will soon be able to assume the role of a tutor
to students. Or, as La Trobe is establishing a special centre for
autistic children, the robots might help these small, disabled
beings overcome the social isolation that many feel because they
cannot identify human emotions the way other children do.
Emotional intelligence (EI) is the capacity of a robot to
interact meaningfully with humans, Khosla explains. Although he is
a senior academic and management expert in the faculty of law and
management at La Trobe, he is also a computer scientist and an
electrical engineer.
But, as he says, he did not take up computer science "because I
loved programming … rather, I thought these were tools that could
be used to solve some real-world problems and make a difference in
meaningful ways".
The Khosla team has received a $1 million, three-year grant from
the giant Japanese NEC Corporation, Kyoto University's graduate
school of informatics and La Trobe to develop EI robots capable of
the interactions that only humans possess at present.
"We have established a centre - an international collaborative
network of excellence - connecting institutions and researchers
from around the world," Khosla says. "Initially we are working with
Kyoto and NEC to change the focus of information communication
technology design from becoming increasingly powerful to offering
convenience in areas such as education, health care and aged care,
the latter because of changing social structures."
He says the Japanese are intensely interested in the health and
aged care field because Japan is an ageing society and within 30
years, 40 per cent of the population will be 65 years old or more.
Australia is experiencing a similar situation, although not to same
extent.
The NEC Corporation is well known for its work in ICT and has
developed the robots. But to make them responsive in an emotionally
intelligent way, the Japanese researchers are relying on Khosla and
his team.
"The philosophy here in our centre is to incorporate the human
senses into the design of computers so they can interact with us in
an emotionally intelligent manner," he says. "We are trying to
develop verbal and emotionally intelligent information and
communication technologies that integrate and correlate verbal,
non-verbal and written forms of communication, what NEC calls
mind-to-mind communication."
Although the Japanese researchers have designed robots that can
communicate with humans, the machines do not possess the ability to
understand human emotions and respond accordingly. This is where
the La Trobe team comes in because it is developing technology that
Khosla calls "emotionally intelligent dialogue systems".
"The robot we are designing can understand human emotions
through the person's facial expressions and integrate them with
other forms of communication such as speech, just as we humans do
in speaking with others," he says. "The robot does not just pick up
the verbal signals but also tracks the visual responses with its
camera so it can communicate in an emotionally intelligent
manner."
But don't these emotionally intelligent creatures already exist
and aren't they called human beings?
"Of course," Khosla says. "But we are facing changing structures
in society and in schools where, with both parents working, they
barely have time to look after their children much less their own
parents - and that is a problem facing Australia as well as
Japan."
He says that an ageing population and spiralling health care
costs mean the support of ICT will increasingly be needed: systems
that will play a more constructive role in society and are intended
to enhance and develop the sustainability of human society.
In one trial in Japan, a robot was assigned to care for a child
while the mother was at work. The robot could communicate with the
mother by email or by calling her mobile and showing with its
camera the child was well and was eating a snack.
In the realm of education, Khosla says the researchers are
investigating the correlation of eye gaze and facial expressions of
students in an e-learning situation and correlating that with their
understanding of the content.
"We want to determine the correlation between their eye gaze,
for example when they are looking at the screen or the content that
is being presented, or at the presenter. We want to discover the
correlation between their eye gaze and their facial expression
while they look at the screen, with the aim of building proactive
EI systems and thereby enhance the quality of delivery of content
and the style of delivery."
A proactive system could tell from the student's expression
whether he or she was comprehending what was being shown.
Alternatively, in a lecture theatre, the information could be sent
to the computer screen of the lecturer so he or she could see how
the students were responding and change the pace of delivery or
take some other action. As Khosla says, it would be like the 'worm'
used during televised political debates except that at university
the worm would provide the emotional responses of the learners.
"We need to be more student-centred or what I call
emotionally-centred. Most students prefer to be in contact with the
lecturer rather than learn via the computer so we need to humanise
the technology as far as possible and make it more emotionally
sensitive to the needs of the student."
In teaching a robot about human responses, a camera at the top
of the computer screen the student is looking at assesses the
student's facial expressions as he or she works on a problem.
Khosla says computers are very good at monitoring subtle changes in
expression and these can be correlated with the student's cognitive
responses.
"We are talking about a fundamental change in communication. We
use a webcam to track the user's facial expressions and correlate
those transient changes in expression with other inputs either
through the keyboard or via some other cognitive responses.
"The health care robot we are designing can understand human
emotions through the person's facial expressions and integrate them
with other forms of communication, such as speech, just as we do in
speaking with others," he says. "The robot does not just pick up
the verbal signals but also tracks the visual responses with its
camera so it can communicate in an emotionally intelligent
way."
The robots the La Trobe team are working with are only half a
metre high, friendly-looking and non-threatening. Khosla imagines
them being used in a pre-operation room where a nurse may only have
a minute with the patient before leaving, so she would be supported
by a nurse robot that would study the patient through its camera
'eyes' and monitor their anxiety level. If the patient appeared to
be distressed, the robot would signal the nursing station so a
nurse could provide more personalised care.
He has plans as well to introduce EI robots into the homes of
elderly people where they could assist with looking after their
health and taking care of other issues. Other possibilities include
using emotionally intelligent ICT systems to help HR managers
recruit and benchmark new staff "with the requisite emotional
intelligence".
Then there are passenger robots that could be located in cars to
monitor driver fatigue and help reduce road fatalities, or
web-based intelligent robots in e-tourism to match people's
emotional preferences to their choice of holiday destination, and
even emotionally intelligent 'security' robots to assist with
criminal investigations and anti-terrorist detection.
"The idea is to facilitate the well-being and sustainability of
human society by improving the quality of life at work and in
various lifestyle situations," Khosla says. "It would completely
alter the way ICT processes data and information in organisations
and lifestyle situations in which people are involved."
Just as a human child learns more about the reactions of others
as they grow, so the EI robot acting as a tutor or in health care
or in a pre-op situation would have records of the emotional
responses of hundreds of patients. Khosla says it would develop a
huge knowledge base and would have learned the correspondence
between anxiety levels and the sedation requirements and type of
care so that it could be used in turn to train new nurses.
But is all this some sort of fanciful science fiction scenario?
Not at all, he says, declaring that within three to five years,
such robots will be a common sight in schools, universities,
hospitals and aged-care homes.
"The whole point of this research is to design these for social
innovation, to be EI partners with humans. There is no way the
robots will try to invade, dictate, manipulate or replace humans -
that is the last thing on our mind. What we are responding to are
the changes in the social structure of society; the whole idea is
to improve the quality of life at work and at home."