Text messages are a minefield. They’re easily misinterpreted (a 2005 study found that message recipients routinely fail to distinguish between seriousness and sarcasm), and one misplaced word or misconstrued phrase could land an unlucky sender in the doghouse.
Kevin Knull, CEO and cofounder of Atlanta, Georgia-based Truthify, found this out the hard way. “I sent a text to my wife that was misunderstood,” he told VentureBeat in an interview. “The reason for that is because we’ve all simplified communication down to just a few words. That can cause significant problems.”
The messaging mishap set the stage for Truthify, a smartphone app for iOS (and soon Android) that leverages machine learning to interpret a person’s emotions from their facial reactions. Using a phone’s front-facing camera, it analyzes 43 facial muscles 14 times a second and evaluates the results across seven emotional states: fear, anger, disgust, happiness, contempt, sadness, and surprise.
Truthify partnered with MIT Media Lab-born Affectiva for the necessary computer vision models, which were trained on a dataset of more than 6.5 million faces of varying genders, ages, and ethnicities in 87 countries. Knull said that microexpressions — the facial movements that occur when the amygdala, the brain’s emotional center, responds to something — are nearly universal across gender, ethnicity, and nationality, making them ideally suited for sentiment analysis.
“You get a lot more information out of it than you get in an email,” he explained. “What we found is almost all people show the emotion on their face, no matter the culture or country. They’re going to show some level of emotion.”
Truthify users can record videos, snap pictures, or write comments and then send them to recipients, and they can add other Truthify users as in-app friends. As they watch, read, or view a message, the app identifies their emotions and shares them with both the sender and recipient, showing the strength of each emotion in percentages — e.g., 70 percent “happiness” or 90 percent “surprise.”
But while person-to-person messaging is where Truthify got its start, its bread and butter is brand marketing and advertising. The startup’s cloud dashboard allows advertisers to focus-test campaigns at scale and with precision — they can target viewers by demographics like geographic region and age range, for example. And it grants them analytics that pinpoint the moment at which recipients’ emotions changed and which device they used to view the content.
“[Our technology] effectively reveals how and when people react to a campaign product,” Knull said. “That’s priceless.”
Truthify isn’t the first to pair emotion analysis — which research firm MarketandMarkets predicts will be worth more than $36 billion by 2021 — with focus testing. Firms like Beyond Verbal analyze emotions from vocal intonations, and Realeyes, another competitor, claims it can determine which ads sell and which don’t with 75 percent accuracy.
What differentiates Truthify, Knull said, is its dedication to privacy. “We’re not going to know who you are — only how you felt,” Knull said. “We think that at the end of the day, data is valuable. We wanted to create some guardrails.”
The app is fully compliant with the General Data Protection Regulation (GDPR), the legal framework that sets out privacy guidelines for folks living in the European Union. And data it collects is anonymized — personally identifiable information such as names and locations aren’t shared unless users give the go-ahead. Even among Truthify users who’ve identified themselves as friends, display names can be hidden or anonymized.
“We’re turning the tables and putting the power of ownership back in the hands of the user,” Knull said.
Truthify is available for free, but Premium and Pro tiers — both of which are available as in-app purchases — unlock additional features. Premium extends the maximum video recording time from seven to 15 seconds, allows senders to filter videos by emotion and view more than five individual responses, and unlocks access to all viewer comments. The Pro tier includes all of those benefits and extends video recording time to 30 seconds.