Originally Published in Techinasia.com
Ranjan Kumar and I sit in an office above the streets of Bangalore. The golden afternoon light on the city outside contrasts with the neat blue-white of the space. I’ve just asked him about feelings, something that’s made him leap out of his chair in excitement and begin drawing on a board.
“I happened to be a teacher for a while in my career, so I love to do whiteboards,” he tells me, and draws a two-axis diagram which looked like something out of my high school geometry class.
Ranjan heads up Entropik Technologies, a startup which seeks to make interactions between humans and machines seamless. The tech he’s building reads your interactions and gauges how you feel while you’re performing them.
I happened to be a teacher for a while in my career, so I love to do whiteboards.
Ranjan’s diagram – called a circumplex model – explains their first product, Chromo. It reads the smartphone user’s screen touches and quantifies that into different emotions, complete with degrees of positivity or negativity as well as levels of intensity.
He marks emotions for me on the graph – happiness, which is intense and positive; relaxation, which is positive but with low intensity; anger, which is negative with high intensity; and boredom, which is negative with low intensity.
Depending on how hard and fast you touch your phone screen, Chromo takes a reading of valence – whether the emotion is positive or negative – and intensity. The product also has motion sensing abilities. It can tell if a user is stationary, walking, or running. It knows the position of the phone (portrait or landscape, whether the user is holding the phone with the screen facing up or down).
“We utilize that stuff to analyze everything around a user: where you are, what you’re doing,” Ranjan explains. In turn, fashion ecommerce and research companies can use Chromo to collect data. Knowing how a person’s emotions change while watching a video, for example, can help a content site know what kinds of stories to publish.
Standing out in the emotion recognition race
Tracking customers’ opinions and engagement is nothing new. A few companies have ventured into tracking customers emotions. Instead of Entropik’s focus on touch, they tend to focus on reading people’s facial expressions in real time. Emovu uses deep learning to detect emotions, and MIT spinoff Affectiva tracks customers’ reactions via facial recognition AI and big data.
Sensaura is a company that focuses on touch – it tracks customers’ feelings via their wearables, not their smartphone screens.
In November last year, Facebook acquired AI facial recognition startup Faciometrics.
The idea for Entropik came when Ranjan was working on projects with Oxford University researchers. A chemical engineering graduate of IIT-Kharagpur, he found that of the five senses, sight and touch were the best ones for emotion recognition. He found touch easier to deal with than sight and began looking into a machine that would predict human emotions. The Chromo prototype was released in August last year, and the final product was officially released last month.
We utilize that stuff to analyze everything around a user: where you are, what you’re doing.
Ranjan and his team continued to work on the product and build up its accuracy level. 75 percent accuracy is good enough for the market, and Chromo now tests at 84 percent accuracy, he claims.
“We are getting away from how I was growing up. I don’t see any kids being fascinated these days about playing sports outside,” Ranjan tells Tech in Asia. “That led me to think that the one thing I believe has happened as part of this AI development is we have lost our emotional sensitivity because machines just understand actions and triggers. It doesn’t understand the emotion piece of things.”
Entropik currently has two clients in the US and five in India. It charges a monthly fee of US$5 per 1,000 users and raised US$200,000 in seed funding last month.
Investors shy away from R&D
Because Entropik’s products are fueled by research and development, Ranjan says that it’s a little harder for it to communicate with investors. First, it takes a longer time to roll out a product as compared to an app.
Investors then want to see “traction” – company value and number of clients. In order to get to those places, companies like Entropik need money first, which means that it’s more difficult to prove that it’s worth taking a chance on such a company.
It’s more difficult to prove that it’s worth taking a chance on such a company.
“I think the whole investor community and the whole ecosystem around it focuses on traction. Traction is valued a lot more than fundamental research or a patent value,” he says.
The company is working on an augmentation to traditional lead tracking software. Instead of just knowing the number of people who click on an article about Donald Trump, for example, the software will show whether that click was fueled by a positive or negative emotion.
The company is testing how well it works with reactions to articles and videos. When Ranjan conducted tests on reactions to wedding dresses, he used his fiancee as one of the test subjects.
“It was kind of fun,” he chuckles.