During Waterloo’s True North conference last week, Affectiva CEO Rana el Kaliouby spoke about emotion AI, the next frontier of artificial intelligence, and its applications.
True North’s three-day event in May highlighted the intersection of humans and technology, providing an opportunity to imagine and re-imagine the impact of technology — the good and the bad. It was also an opportunity to examine the values that guide technology innovation in Canada and to redefine tech as a force for good.
“There is an inevitable merger of IQ and EQ in technology.”
– Rana el Kaliouby, CEO of Affectiva
Our emotions influence every aspect of our lives – how we learn, how we communicate, how we make decisions. As a result, advanced AI systems and robots interacting with humans will require social and emotional intelligence, not just computing power. Scientists are in the process of humanizing AI and moving it from a transactional interface to a real conversation interface. To that, these devices need to have some sort of understanding of the social and emotional state of the person they are interacting with.
El Kaliouby imagined what it would be like to have a computer that understands human emotions just as well as we do. She said that people who have higher EQs tend to do better in life; they are more persuasive and are more liked. Current AI technologies are more or less oblivious to our emotional stance.
What if computers could understand the difference between a smile and a smirk, or laughter and frustration? What if our learning apps could understand the emotional engagement of a student and adapt to it, just the way a good teacher would do in a classroom? What if doctors could quantify our mental health state just the way they do with other vital signs.
— Launch Lab (@launch_lab) May 30, 2018
El Kaliouby invented Affectiva’s award-winning, emotion recognition technology that is built using deep learning and the world’s largest emotion data repository of nearly 6.5 million faces, analyzed from 87 countries, and amounting to more than 50 billion emotion data points.
“If you look at howpeople interpret each other’s mental states, 55 percent of the signal is in your facial expressions and gestures.”
“There is an inevitable merger of IQ and EQ in technology,” said El Kaliouby. “There are two sides of the new emerging field of emotion AI. One is the real-time interface and the other side of it is analytics – you collect a lot of data about how people are feeling about all sorts of things, and that drives really unique and interesting insights. Collectively, we call this field Emotion AI and it powers and transforms a number of industries.”
It takes about a 100 hours of training to become a certified face reader, and if you are analyzing people’s reactions, it takes about five minutes to code every minute of video. This is a slow and laborious process, and Affectiva is using computer vision and machine learning to automate this process. The technology can read 20 different facial expressions and about seven different emotions. It can also distinguish between age, ethnicity and gender.
“If you look at how other people interpret each other’s mental states, 55 percent of the signal is in your facial expressions and gestures. Thirty-eight percent is in the tone of voice – how fast you are talking, your pitch, and your tone, and only seven percent is in the choice of words. The entire industry of sentiment analysis is just focused on that seven percent, and the remaining 93 percent is essentially lost in cyberspace. Affectiva is trying to change that.”
The first application that Affectiva brought to market is to test how people engage with content online. The technology is used by over a third of the fortune 100 companies around the world to test ads before they go live. Affectiva has tested about 20,000 ads worldwide.
If you ask people how they feel about an ad after they have watched it, they tap into the cognitive part of their brain, whereas Affectiva taps into the visceral part of the brain that they may not even be aware of.
Affectiva vowed to not collect any data without consent.
Affectiva sends a survey on the participant’s phone and asks for their permission to turn the camera on and if they say yes, the camera turns on, the participant watches an ad, and their reaction is recorded. The data is then aggregated anonymously and the technology can provide a moment-by-moment response of how people engage with a certain piece of content.
Their research has also shown that socially progressive ads are 25 percent more engaging and more effective on a lot of perimeters (though this is obviously highly dependent on the market).
Affectiva has also partnered with Hirevue, a company that is trying to humanize the hiring process. Instead of sending a resume and a cover letter, candidates send a short video answering a few questions about who they are, what excites them, and their story.
Using Affectiva’s technology and some natural language processing, Hirevue is able to sort the candidates based on their soft skills, which has reduced hiring time by 90 percent. The company claims that these algorithms are blind to gender and ethnicity, and companies have ended up hiring 16 percent more diverse populations than they did before.
The University of Rochester is piloting this technology to help people with their public speaking skills and to practice job interviews. The technology enables a person to track their progress over time.
As we think about self driving cars and ride sharing, a technology like this can improve the experience. For instance, if a person is drowsy or texting, the vehicle can sense this and take over, thus reducing the number of accidents on the road.
Applications in mental health space
Affectiva is using their technology to build interfaces that allow for social inclusion. They have deployed it to kids on the autism spectrum and essentially it gives the kids real time feedback on what others are doing, to make social interactions easier.
Work is also being done to understand depression, as there are clear facial and vocal markers of depressive patients.
What is going to happen when the technology is more emotionally intelligent?
This type of technology can augment what a human does and make their jobs easier. In fact, AI is creating four out of five new jobs that have not existed before, such as robot trainers.
Affectiva is aware of biases in AI and privacy concerns with such a technology, and vowed to not collect any data without consent. They are also working to counter bias in AI by ensuring that the people who are designing these systems are diverse.
Photo via Twitter.