This is why emotional artificial intelligence matters | Maja Pantic | TEDxCERN
Summary
TLDRThe talk delves into emotional artificial intelligence, a field with potential for immense societal good but also misuse. It discusses technology's ability to sense and analyze human emotions through facial expressions, with applications in therapy for autistic children and detecting depression signs. However, it warns of the risks of data misuse, such as personalized pricing based on online behavior, urging consumers to be vigilant about their digital footprint.
Takeaways
- 🧠 Emotional artificial intelligence (EAI) is a growing field focused on the automatic sensing and analysis of human behavior, particularly facial expressions.
- 🤖 EAI has the potential to greatly benefit people, such as in therapeutic applications with autistic children, where robots can provide consistent facial expressions for learning.
- 📈 The technology is advanced enough to detect and track faces and facial expressions in various scenes, but challenges remain with occlusions, head position changes, and small faces.
- 👶 For autistic children, EAI can help them understand and express emotions in a way that is recognizable to others, facilitating communication.
- 🤝 Robots can be programmed to display the same facial expressions consistently, which is beneficial in therapy sessions where human expressions might vary.
- 😄 Autistic children often respond positively to robots, as seen in therapy sessions where they show increased attentiveness and engagement.
- 😟 EAI can also be used to detect behavioral cues of depression, which is a significant and increasing issue, especially among teenagers and the elderly.
- 🔍 Depressed individuals may exhibit subtle and short-lived smiles and a negative bias in interpreting expressions, which EAI can help identify.
- 💡 The technology can flag these cues to healthcare providers or family members, potentially leading to earlier intervention and support.
- 🛒 However, EAI can be misused, as demonstrated by a Facebook patent for a camera that could customize prices based on consumer behavior and preferences, including medical remedies.
- ⚠️ There is a call to be aware of the data we share, especially behavioral data, as it can be exploited by companies for their benefit, urging consumers to be informed and cautious.
Q & A
What is the main topic of the talk?
-The main topic of the talk is Emotional Artificial Intelligence (EAI), also known as affective computing or human-centric AI, which involves the automatic sensing and analysis of human behavior, particularly facial expressions.
Why is the human face important in EAI?
-The human face is important in EAI because it is the only observable window of our inner selves, including our emotions, intentions, attitudes, and moods. It is used for recognizing other people and judging various attributes such as age, gender, and sometimes even personality.
What are some of the challenges in EAI technology?
-Some challenges in EAI technology include dealing with occluded faces, large changes in head position, especially fast ones, very small faces, and accurately interpreting typical facial expressions in various contexts.
How can EAI be applied in therapy with autistic children?
-EAI can be applied in therapy with autistic children by using robots to teach them to recognize and express emotions in a way that typically-developing people can understand. Robots can provide consistent facial expressions and positive feedback, which can help in the iterative loop of behavioral therapy.
Why are autistic children often not able to understand or express facial expressions correctly?
-Autistic children often cannot understand or express facial expressions correctly because they may perceive the thousands of daily facial expressions as thousands of different categories, making it difficult for them to generalize and learn the appropriate expressions to communicate their emotions.
What is one example of a positive outcome from using EAI in therapy with autistic children?
-One positive outcome is that a child with attention deficit disorder became more attentive after a session with a robot, as reported by his daily supervisor. This shows the potential of EAI to improve focus and engagement in therapy.
How can EAI technology be used to detect signs of depression?
-EAI technology can be used to detect signs of depression by automatically sensing and analyzing behavioral cues such as the quality and intensity of smiles, as well as the presence of negative bias in interpreting situations.
What is the potential misuse of EAI technology mentioned in the talk?
-The potential misuse of EAI technology mentioned in the talk is the patenting of a camera by Facebook that could recognize people and customize prices based on their online search patterns and behavioral signals in shops, leading to higher prices for those who show more interest in a product.
How can consumers protect themselves from the misuse of their data by companies?
-Consumers can protect themselves by being aware, getting informed, not clicking on all cookies, and being cautious about the kind of information they give away. They should keep their own data for their own use.
What is the speaker's final message to the audience regarding EAI technology?
-The speaker's final message is to be aware and informed about how EAI technology can be used and misused, and to take care of the personal data they provide to companies to prevent potential misuse.
Outlines
😀 Emotional AI: Potential and Ethical Considerations
This paragraph introduces the concept of emotional artificial intelligence (AI), a burgeoning field within AI research with significant potential to benefit society. It discusses the automatic sensing and analysis of human behavior, particularly facial expressions, which serve as a window to our emotions and intentions. The speaker highlights the advanced state of the technology, which can detect and analyze facial expressions in various scenarios. However, they also acknowledge the challenges, such as dealing with occluded faces and rapid head movements. The paragraph emphasizes the importance of consumer awareness in determining the ethical use of this technology, setting the stage for a deeper exploration of its applications and misuse.
🤖 Emotional AI in Therapy and Autism
The second paragraph delves into the application of emotional AI in therapeutic settings, specifically for children with autism. It explains the difficulty autistic children face in interpreting facial expressions due to their atypical generalization abilities. The speaker suggests that robots can be more effective than human therapists in teaching these children to recognize and express emotions consistently. The use of robots provides a controlled environment for learning, as they can offer the same facial expression repeatedly, which is crucial for autistic children's understanding. The paragraph provides examples of successful interactions between autistic children and robots, illustrating the positive impact of emotional AI on communication and social skills.
😟 Detecting Depression with Emotional AI
This paragraph explores the use of emotional AI in identifying behavioral cues of depression. The speaker discusses the alarming rates of depression among various demographics, emphasizing the need for tools that can detect early signs of this mental health issue. The technology can analyze the quality of smiles and the presence of negative bias in facial expressions, which are indicative of depression. The speaker contrasts the public misconception of depression with the actual behavioral signals that AI can detect. The paragraph also raises concerns about the potential misuse of emotional AI, such as Facebook's patent for a camera that could potentially adjust prices based on consumers' online behavior and preferences, including sensitive information about health conditions.
Mindmap
Keywords
💡Emotional Artificial Intelligence
💡Affective Computing
💡Human-Centric Artificial Intelligence
💡Facial Expressions
💡Autism
💡Behavioral Therapy
💡Depression
💡Misuse of Technology
💡Data Privacy
💡Cookies
💡Ethical Considerations
Highlights
Emotional artificial intelligence (EAI) is a new field in AI with the potential for immense good or misuse.
EAI, also known as affective computing or human-centric AI, involves automatic sensing and analysis of human behavior and facial expressions.
The human face is a crucial window into our emotions, intentions, attitudes, and moods.
Technological advancements have enabled the detection and analysis of facial expressions in various naturalistic scenes.
Current technology can measure the intensity of facial gestures like frowns and smiles, and analyze higher-level behaviors and emotions.
Challenges remain in dealing with occluded faces, large head position changes, very small faces, and typical facial expressions.
EAI is mature enough for applications such as therapy with autistic children, where it can help teach emotional expression and recognition.
Autistic children often struggle with the complexity of facial expressions, making communication challenging.
Robots can be programmed to display consistent facial expressions, aiding in therapy for autistic children.
Autistic children's affinity for mechanical toys makes robots particularly effective in therapy.
EAI can also be used to detect and flag behavioral cues of depression, a significant issue in the Western world.
Depression's impact is particularly severe among teenagers and the elderly, with many cases going undetected and untreated.
EAI can analyze the quality of smiles and negative bias in interpretation, which are indicators of depression.
Facebook patented a camera technology that could potentially misuse EAI by customizing prices based on consumer behavior and preferences.
The potential misuse of EAI raises ethical concerns, especially when it comes to privacy and the exploitation of personal data.
Consumers must be aware and informed about the data they share, including behavioral and facial expression data.
The talk emphasizes the importance of being cautious with the personal data shared online and the potential for both positive and negative uses of EAI.
Transcripts
[Music]
[Applause]
this talk is about emotional artificial
intelligence it's a relatively new field
in research in artificial intelligence
but it has a great potential to do
immense good to people for people
however the technology can be misused
and in this talk I want to make a point
that actually it is up to us the
consumers of this technology who will
decide whether the technology will be
used for good or for evil but let me
first explain what emotional artificial
intelligence is really about so
emotional artificial intelligence goes
by various names affective computing is
one of those another one is human
centric artificial intelligence but
actually behind all of those names is
exactly the same research topic
automatic sensing and analysis of human
behavior and in particular human facial
behavior human face is really
fascinating we use the face to recognize
other members of our species to
recognize other people use the face also
to judge various things such as age
gender beauty and sometimes even
personality don't do the last one most
importantly the human face is the only
observable window of our inner selves of
our emotions intentions attitudes and
moods it is therefore not surprising
that in the recent years we had a surge
of interest in this technology and if we
could I mean the reason is that if we
could detect the faces in various
naturalistic scenes and then analyze the
facial expressions of
those phases in terms of specific
emotions or specific attitudes we could
use this kind of technology for a very
wide number of applications the state of
the art in the field is relatively
advanced so we do have and we did
develop a number of techniques that
could detect and track the phases in
various and realistic scenes we also do
have the tools that can detect track and
measure the intensity of various facial
gestures such as frowns and smiles we
also can automatically analyze
higher-level behaviors such as certain
emotions and certain attitudes like
interest however the methodology and the
technology is not fully matured so we
still have great problems in dealing
with occluded faces with dealing with
faces which where we observe large
changes in head position especially fast
changes in head position very small
faces make problem to us and also our
typical facial expressions such as the
expressions of the lady enjoying her
skydive however the technology is
matured enough to be used in a large
number of really great applications so
let me just give a couple of examples of
those the first example I want to talk
about is use of emotional artificial
intelligence in therapies with autistic
children the human face displays 10,000
different facial expressions 7,000 of
which we display on a daily basis
typically developing people classify
those expressions in maybe 15 or 20
categories
however autistic children means this
generalization ability so 7,000 facial
expressions that we express on a daily
basis they see a 7,000 different
categories this is the reason why
autistic children usually do not look at
our faces they are too confusing for
that but this is also the reason
why they can not understand what are our
facial expressions when we try to
express that we are sad or we are upset
or we are happy this is also the reason
why they cannot learn what are those
expressions that they should display
when they want to say that they are
happier they are upset
so that typically-developing people can
understand them and hence the breakage
in communication so one of the major
goals of behavioural therapy with
autistic children is to give them the
tools the skills to actually express the
emotions in a way that
typically-developing people can
understand and also to understand what
are the emotions of others around them
and one of the problems with this is
that when you have a human teach a child
these typical expressions humans like
and usually emphasize things so when
they teach the child you know when you
are happy you smile they usually raise
the eyebrows and most of you probably
didn't notice that at all right but
actually for autistic child a smile with
raised eyebrows is a very different
expression than just a smile so this is
where we came to the idea to use the
robots instead of human therapists
because you can program the robot and
the expression that the robot will show
will always be the same expression right
so when the therapist says you know can
you display a happy facial expression or
how would you feel if you get a train
then the robot complies and shows this
smiley expression the child sees this
expression repeats the expression and
then the robot because it recognizes
that the child display the correct
facial expression gives a positive
feedback so this is the way you can use
the robots in this iterative loop of
behavioral therapy with autistic
children autistic children love robots
they actually love all mechanical toys
which are built of parts
and exactly this love is something that
can have a really amazing effect on
children so the district herb boy that
you have previously seen suffers from
rather severe attention deficit however
after this session with robot his daily
supervisor who is this guy in the
background said to us that he has never
seen the boy more attentive in a whole
year right similarly when we brought the
robot to Serbia to have a session with
Serbian autistic children we had this
boy who is a nonverbal child interacting
with with a robot when we say a
nonverbal autistic child this doesn't
mean that the child cannot speak it
means that the child chooses not to
speak but after this session you see he
was really like happy and he was really
enjoying his time with the robot and
after the session he went home and said
to his mom and tomorrow in school the
robot he said to his mom so the mother
of course immediately wrote to us and
said what did you do like this is
amazing it was it was for her an
unbelievable awakening and this kind of
awakenings is actually what makes the
robot worthy of putting into the
therapies with autistic children and
this is what makes our research on this
topic fully worthwhile another case and
a very good application of emotional
artificial intelligence technology is
the case of automatic sensing and
flagging behavioral cues of depression
depression is a huge problem currently
in Western world we are having an
increase of depressed people at all age
levels however with teenagers this is
the worst in USA and UK we currently
have 25 percent of teenage girls
suffering from this illness overall 17
percent of young adults aged 18 to 22
suffer from depression it is very
similar the case with the elderly
population in UK we have 22% of people
above 65 years of age who suffer from
depression the worst part is it actually
only one out of ten of these elderly
people are officially helped by UK
national health services so it would be
really fantastic if we would have tools
that could actually detect these signs
behavioral cues of depression and flag
them to GP or to the family members when
you talk to people about depression they
often may say to you you know depressed
people do not smile that often but it is
not at all the case it is not a
frequency it is really the quality of
the smiles depressed people feel usually
embarrassed if they smile so their
smiles are short-lived and they are
often dampened another problem is the
negative bias so whatever you say to a
depressed person they will actually
interpret it more negatively than the
intention was let me ask you something
what do you think this lady is currently
watching a very typical answer is that
she's watching some kind of horror movie
but actually we asked her to watch funny
videos she's watching the videos of kids
and cats running around but due to her
negative bias she has these very strong
expressions of fear and even disgust her
smiles are really so subtle that the
intensity of them are barely detectable
and it is exactly this the cumulative
effect of all those behavioral signals
that we plot on this
lined plot at the bottom of the vida so
emotional artificial intelligence can be
used for great purposes really to help
people to help humanity but it can also
be misused let me give you an example in
the sample 2017 Facebook patented a
camera the camera will be placed in
shops and shopping malls it will
recognize people by their profiles in
facebook currently Facebook has 1.4
billion
profits also the camera will know what
are the search patterns that the people
exhibited on Facebook and on Google so
they will know what we like or what we
are interested in
furthermore the camera will watch our
behavioral signals in the shop what do
we like to which departments do we go
most often and based on all of this the
price will be customized so the more you
search for something the more you like
something the price will be higher the
worst part is actually that neither
medicines nor medical care nor medical
services are excluded from the pattern
so if somebody is ill insert is a remedy
online and has a Facebook account that
person will unfortunately pay much
higher prices for this medical remedy
although again emotional artificial
intelligence can be used for really
great purposes and great good it can be
badly misused but it is especially the
case if we carelessly and blindly give
our data our behavioral data to certain
companies so this is exactly what's
happening nowadays I mean you go to any
website they ask you to
clicking except these cookies right what
are the cookies the cookies are actually
the search patterns of what you do what
did you search to which website you went
how many times you visited that website
what did you purchase how many times you
purchase that so they have four
behavioral pattern of your purchasing
behavior then we go to this Facebook we
open the profile we put our pictures
with P to put our children's picture we
put our friends then we tag everybody we
put videos so they give everything they
have the faces they have the behavioral
patterns they have how we smile which is
a dynamic thing right they have can we
walk what makes us happy where we go
most often so then you go back to their
part and then you realize how harmful
that can be so the message for today
really is
be aware get informed you don't have to
click on all of these cookies all the
time you can actually scroll without
accepting the cookies
take care what kind of information you
give away and keep your own data for
your own thank you
[Applause]
[Music]
[Applause]
関連動画をさらに表示
Artificial intelligence (AI) – Pros & Cons | gd topic
The Ethics of Deep Learning
The Incredible Creativity of Deepfakes — and the Worrying Future of AI | Tom Graham | TED
Will Artificial Intelligence Take Over the World? | Between the Lines with Palki Sharma
Emotion, Stress, and Health: Crash Course Psychology #26
Ethical Implications of Business Analytics | Dominic Ligot
5.0 / 5 (0 votes)