[Applause]
there are many types of intelligence but
when we think of intelligence we
typically think of our ability to
perform verbal or logical reasoning the
type of skills that we measure in IQ
tests but there’s a intelligence can be
much more than that particularly there’s
a type of intelligence that we can all
agree that is quite useful in our
everyday lives I’m talking about social
intelligence people with high levels of
social intelligence are usually very
good at communicating with others and in
noticing and even anticipating the other
people’s intentions and emotional states
they are quite good at empathy and in
putting themselves in the perspective of
others they are good team players and
nice people to be around with my job is
to make robots more socially intelligent
robots are becoming part of our lives
although not always in the forms that we
picture from science fiction but because
these machines are operating and
assisting us in environments that are
inherently social they need social
intelligence to
let me start by giving you a concrete
example this is I cut a robotic chess
companion for kids when you begin
learning how to play chess the more you
play the faster you’ll earn up to a
certain level so together with a number
of colleagues we wanted to investigate
how we could use robots to help kids
become motivated and wanting to play
chess more so we programmed this robot
with the ability to play chess and
provide feedback on children’s moves
throughout the game through facial
emotional expressions and from what I
told you so far I guess you must be
wondering by now that the hardest part
of this project was not to program the
robot with the ability to play chess at
the level of an eight-year-old beginner
child in the initial experiments that we
conducted with the ICAT in schools and
chess clubs
we found that the robots emotional
behavior was in fact helping children
gain a better understanding of the chess
game so our setup was quite promising
however in a long-term study where the
same group of kids play chess with his
robot over five weeks we started
noticing that children started losing
interest into the robot after a couple
of weeks this happened because they
realized that their own behavior an
effective state was not having any
impact on the robots own behavior the
robot was simply responding to the moves
on the chess board for example there
were cases where the ICAT had a clear
advantage in the game and it started
getting happier and happier even if the
child was a little bit frustrated about
it in a nutshell a robot was missing one
of the key aspects of social
intelligence empathy so to address this
we started working on an effect
recognition system that would enable the
robot to detect in real-time the
effective state of the child and use
this information not
only to provide emotional responses more
in tune with the effective state of the
player but also to provide at certain
moments supportive behaviors that would
make children shift from a negative to a
positive effective state if that was
necessary a new repeated interaction
study with this improved version
empathetic version of the ikat showed
that children indeed became more
motivated into playing chess over time
when compared to the version of the
robot without empathetic capabilities
this is just one simple example of how a
social competency can be added to
technology in order to be more effective
in achieving our goal in this case was
to keep children motivated to play chess
so why do we need to build socially
intelligent machines there are still
many computational challenges let me
give you just an example of one of these
challenges that I’ve been working for
quite a while now the typical human
machine interaction paradigm is to have
one machine and one person or the user
we tend to sit alone in our computers or
when we interact with our phones we
forget about the social world around us
but robots
are a very special type of device
because they are situated in our social
world and even the ones that don’t look
human-like at all
they evoke in you a sense of social
presence and so when we deploy these
robots in social environments around
people it’s not uncommon to see more
than one people person around robot at
the same time and in a not so distant
future we can also envision to have more
than one robot added to the group so one
of the challenges that we have when we
have robots interacting in real world
social environments is that we cannot
really anticipate how many people will
be around the robot at the same time
this is challenging for us robot
programmers not only because there are
many situations that we need to account
for but also and most importantly
because the way we behave when we are in
a group is very different than the way
we behave when we are interacting with
one of the projects that I’ve worked at
Yale University before joining kth was
showing exactly this our goal was to
build a system that would detect
automatically whether children were
engaged or disengaged while they were
receiving some educational intervention
from these two little social robots when
we started looking at the data that we
collected with these robots in schools
we noticed that in fact the way children
behaved when they were alone with the
robots was very different than the way
they were when they were in a group for
example if children were alone with a
robot when they were engaged they were
making eye contact with the robots
almost a hundred percent of the time and
when they started looking away that was
a clear sign that they were disengaged
from the task however when children were
interacting in a group setting quite
often when they looked away from the
robots they were addressing their
colleagues to discuss something related
to the task itself so it didn’t
necessarily mean that they were
disengaged with robots so the way a
robot needs to interpret a glance to the
side is different depending on the
social context that the robot is
experiencing in that moment and this is
difficult because the same algorithms
that we develop in this case we need to
detect disengagement for example and
work for one social context might not
work for all the other contexts that we
robots today are still very still
struggle with understanding all these
different social cues that we humans can
this is just one of the challenges that
we are facing in this area that I worked
on social robotics but progress is
happening quite fast and let me stop
here by saying that I’m not arguing that
all robots need to be social but instead
there’s a special place for socially
intelligent robots in our future society
advances are being made and are driven
by economic and demographic trends for
example here in Sweden 25% of our
population is above 65 years old like in
many developing countries as life
expectancy is increasing people also
wish to live longer independently in
their homes socially intelligent robots
can enable older adults to live longer
in their homes independently acting as
cognitive assistants or providing help
another example where robots are
expected to play a key role in our
future society is in personalizing
education in the next five years
Sweden is expecting almost a million new
immigrants these immigrants are often
families and their children are being
integrated in our public schools these
children quite often have different
learning paces
not only because they are coming from a
different educational system but also
because they probably don’t master the
language spoken by their teacher or by
their new colleagues
so here socially intelligent robots can
complement teachers by providing
personalized instruction to these
children so that they are able to keep
up with the learning content that they
need or by acting as language learning
companions in fact studies have shown
that the physical presence of a robot
plays a very important role on how
students can recall information for
example when compared with exactly the
same content being delivered by a
virtual character on a screen so there’s
a great potential for social robots to
finally we will see more and more robots
assisting us in workplaces either by
collaborating with workers in assembly
lines or carrying things around
hospitals so that nurses and other staff
can have more free time to provide the
necessary supports to patients even when
the main purpose of the robot is
non-social if these robots will be
operating in environments surrounded by
people they will need to be socially
intelligent today robots still struggle
with things that we people can do
without almost any effort like the
ability to understand when someone is
actively listening to us or the ability
to intervene in at the right moment in a
group conversation so even though there
are already many examples of commercial
robots in factories or even exploring
other planets we don’t see many robots
around us yet and when we do their
functionality is still quite limited in
order to change this we need robots that
are not only good at what they do and
useful but also that are socially aware
and can take into account the needs and
intentions of people around them robots
need to go from an individualistic to a
collectivistic mindset like socially
intelligent people do
the goal of many social robotics
researchers myself included is to build
and evaluate systems that enable robots
to capture respond to and learn from the
subtle social dynamics these complex
real-world situations the better robots
will be at understanding human social
behavior and the better they will be at
expressing their own intentions to
people in other words the more socially
intelligent they are the more natural
and effective will be the communication
between these robots and people and
therefore the better they will be at
assisting us in a variety of domains I
believe this is the power of socially intelligent robots thank you
there are many types of intelligence but
when we think of intelligence we
typically think of our ability to
perform verbal or logical reasoning the
type of skills that we measure in IQ
tests but there’s a intelligence can be
much more than that particularly there’s
a type of intelligence that we can all
agree that is quite useful in our
everyday lives I’m talking about social
intelligence people with high levels of
social intelligence are usually very
good at communicating with others and in
noticing and even anticipating the other
people’s intentions and emotional states
they are quite good at empathy and in
putting themselves in the perspective of
others they are good team players and
nice people to be around with my job is
to make robots more socially intelligent
robots are becoming part of our lives
although not always in the forms that we
picture from science fiction but because
these machines are operating and
assisting us in environments that are
inherently social they need social
intelligence to
let me start by giving you a concrete
example this is I cut a robotic chess
companion for kids when you begin
learning how to play chess the more you
play the faster you’ll earn up to a
certain level so together with a number
of colleagues we wanted to investigate
how we could use robots to help kids
become motivated and wanting to play
chess more so we programmed this robot
with the ability to play chess and
provide feedback on children’s moves
throughout the game through facial
emotional expressions and from what I
told you so far I guess you must be
wondering by now that the hardest part
of this project was not to program the
robot with the ability to play chess at
the level of an eight-year-old beginner
child in the initial experiments that we
conducted with the ICAT in schools and
chess clubs
we found that the robots emotional
behavior was in fact helping children
gain a better understanding of the chess
game so our setup was quite promising
however in a long-term study where the
same group of kids play chess with his
robot over five weeks we started
noticing that children started losing
interest into the robot after a couple
of weeks this happened because they
realized that their own behavior an
effective state was not having any
impact on the robots own behavior the
robot was simply responding to the moves
on the chess board for example there
were cases where the ICAT had a clear
advantage in the game and it started
getting happier and happier even if the
child was a little bit frustrated about
it in a nutshell a robot was missing one
of the key aspects of social
intelligence empathy so to address this
we started working on an effect
recognition system that would enable the
robot to detect in real-time the
effective state of the child and use
this information not
only to provide emotional responses more
in tune with the effective state of the
player but also to provide at certain
moments supportive behaviors that would
make children shift from a negative to a
positive effective state if that was
necessary a new repeated interaction
study with this improved version
empathetic version of the ikat showed
that children indeed became more
motivated into playing chess over time
when compared to the version of the
robot without empathetic capabilities
this is just one simple example of how a
social competency can be added to
technology in order to be more effective
in achieving our goal in this case was
to keep children motivated to play chess
so why do we need to build socially
intelligent machines there are still
many computational challenges let me
give you just an example of one of these
challenges that I’ve been working for
quite a while now the typical human
machine interaction paradigm is to have
one machine and one person or the user
we tend to sit alone in our computers or
when we interact with our phones we
forget about the social world around us
but robots
are a very special type of device
because they are situated in our social
world and even the ones that don’t look
human-like at all
they evoke in you a sense of social
presence and so when we deploy these
robots in social environments around
people it’s not uncommon to see more
than one people person around robot at
the same time and in a not so distant
future we can also envision to have more
than one robot added to the group so one
of the challenges that we have when we
have robots interacting in real world
social environments is that we cannot
really anticipate how many people will
be around the robot at the same time
this is challenging for us robot
programmers not only because there are
many situations that we need to account
for but also and most importantly
because the way we behave when we are in
a group is very different than the way
we behave when we are interacting with
one of the projects that I’ve worked at
Yale University before joining kth was
showing exactly this our goal was to
build a system that would detect
automatically whether children were
engaged or disengaged while they were
receiving some educational intervention
from these two little social robots when
we started looking at the data that we
collected with these robots in schools
we noticed that in fact the way children
behaved when they were alone with the
robots was very different than the way
they were when they were in a group for
example if children were alone with a
robot when they were engaged they were
making eye contact with the robots
almost a hundred percent of the time and
when they started looking away that was
a clear sign that they were disengaged
from the task however when children were
interacting in a group setting quite
often when they looked away from the
robots they were addressing their
colleagues to discuss something related
to the task itself so it didn’t
necessarily mean that they were
disengaged with robots so the way a
robot needs to interpret a glance to the
side is different depending on the
social context that the robot is
experiencing in that moment and this is
difficult because the same algorithms
that we develop in this case we need to
detect disengagement for example and
work for one social context might not
work for all the other contexts that we
robots today are still very still
struggle with understanding all these
different social cues that we humans can
this is just one of the challenges that
we are facing in this area that I worked
on social robotics but progress is
happening quite fast and let me stop
here by saying that I’m not arguing that
all robots need to be social but instead
there’s a special place for socially
intelligent robots in our future society
advances are being made and are driven
by economic and demographic trends for
example here in Sweden 25% of our
population is above 65 years old like in
many developing countries as life
expectancy is increasing people also
wish to live longer independently in
their homes socially intelligent robots
can enable older adults to live longer
in their homes independently acting as
cognitive assistants or providing help
another example where robots are
expected to play a key role in our
future society is in personalizing
education in the next five years
Sweden is expecting almost a million new
immigrants these immigrants are often
families and their children are being
integrated in our public schools these
children quite often have different
learning paces
not only because they are coming from a
different educational system but also
because they probably don’t master the
language spoken by their teacher or by
their new colleagues
so here socially intelligent robots can
complement teachers by providing
personalized instruction to these
children so that they are able to keep
up with the learning content that they
need or by acting as language learning
companions in fact studies have shown
that the physical presence of a robot
plays a very important role on how
students can recall information for
example when compared with exactly the
same content being delivered by a
virtual character on a screen so there’s
a great potential for social robots to
finally we will see more and more robots
assisting us in workplaces either by
collaborating with workers in assembly
lines or carrying things around
hospitals so that nurses and other staff
can have more free time to provide the
necessary supports to patients even when
the main purpose of the robot is
non-social if these robots will be
operating in environments surrounded by
people they will need to be socially
intelligent today robots still struggle
with things that we people can do
without almost any effort like the
ability to understand when someone is
actively listening to us or the ability
to intervene in at the right moment in a
group conversation so even though there
are already many examples of commercial
robots in factories or even exploring
other planets we don’t see many robots
around us yet and when we do their
functionality is still quite limited in
order to change this we need robots that
are not only good at what they do and
useful but also that are socially aware
and can take into account the needs and
intentions of people around them robots
need to go from an individualistic to a
collectivistic mindset like socially
intelligent people do
the goal of many social robotics
researchers myself included is to build
and evaluate systems that enable robots
to capture respond to and learn from the
subtle social dynamics these complex
real-world situations the better robots
will be at understanding human social
behavior and the better they will be at
expressing their own intentions to
people in other words the more socially
intelligent they are the more natural
and effective will be the communication
between these robots and people and
therefore the better they will be at
assisting us in a variety of domains I
believe this is the power of socially intelligent robots thank you
Please follow and like us: