For Chevrolet and Hollywood, IBM’s Watson is learning how to understand emotions and personalities

The movie poster for “Morgan”

IBM’s Watson cognitive computing system is learning all about human personality and emotions.

In particular, about positive outlooks and about fear.

Recently, the Jeopardy-winning electronic brain has worked on projects to score the positivity of users for Chevrolet’s “Find New Roads” campaign and to create a movie trailer for a new science fiction/horror film from 20th Century Fox.

Chevy employed Watson to power its Global Positivity System, a recently-launched mobile and desktop website to assess participants’ Facebook or Twitter posts in order to score their positive outlook and recommend related activities.

The basic marketing idea is that Chevrolet’s “positive, never-give-up attitude” has backed up the brand since the days of its indefatigable founder, Louis Chevrolet. Through this online scoring mechanism, participants can evaluate their personality and, hopefully, find a sufficient enthusiasm to welcome “new roads.”

“Because of your positivity, Chevy enables you to find a new road,'” is how Chevrolet Social Media Marketing Manager Karen Toor explained the connection to me.

This is the first time that the ebullient Chevy has worked with the expressionless Watson. The web test employs Watson’s APIs for Personality Insights and Alchemy Language to assess sentiment and personality from the sampled social posts.

Personality Insights identifies positive keywords, personality characteristics and individual traits, and the APIs that IBM calls Alchemy Language develop a social personality profile based on its sentiment analysis. Together, they build a Global Positivity System score based on 200 possible points.

That score, which can be shared or compared with others, also includes your top personality traits, like excitement or self-expression. The participant is encouraged to undertake activities relating to those traits, like learning a musical instrument if you test as being self-expressive.

As part of this positivity campaign, Chevy went to three gas stations — one each in Buenos Aires, Cape Town and New Orleans — where patrons were given the opportunity to take the Find New Roads online test and pay for their gas with their positivity scores. The higher the score, the more gas they got for having that attitude.

https://www.youtube.com/watch?v=vKd9Oq9xgzA&feature=youtu.be

To me, the FindNewRoads.com website and test seemed no more accurate or perceptive about my traits than any number of Watson-less personality tests or other pop assessments to determine who you really are. Additionally, there is no shortage of firms that conduct sentiment analysis of social posts.

But IBM Watson Ecosystem Director Steve Abrams told me that the real uniqueness here is how Watson is learning to make judgments about emotions.

He pointed out that Watson’s win on Jeopardy was a tour-de-force of factual recognition, but it didn’t involve the kinds of emotional judgments that humans constantly make. And if computer intelligence is going to really power customer service or other human-facing tasks, he said, it needs to understand the emotional content of what a human is saying.

“Not just the text,” he noted, “but the subtext.”

Sentiment, Abrams said, indicates “thumbs up or down,” but there’s a person’s tone under that and, below that, personality. Abrams said Watson can now conduct “incredibly deep analysis” for emotional content, plotting 52 different vectors from text.

He added that this can become very useful in marketing, particularly for micro-segmentation.

“Imagine someone is trying to [rent] you a vacation home,” he said. If they know you’re adventurous and not easily afraid of new challenges, they might suggest a thatched hut in Thailand, for instance.

This side of Watson is also visible in other collaborations. Influencer network Influential, for example, is employing Watson to find out which of its influencers are best suited to a particular brand campaign, based on Watson’s emotional and personality profiles. Kia Motors used this system to identify influencers with personality traits related to their brand, as did Vogue, Vanity Fair and GQ.

Scaring you

And today, Influential is getting Watson’s help in launching their #HugForPeace campaign to support the International Day of Peace. They will help to promote the occasion with influencers who have such traits as altruism, self-transcendence, emotionality and dutifulness.

Similarly, social data company StatSocial is using Watson to assist in personalized campaign targeting based on social and blog content.

To get a sense of Watson’s abilities beyond social posts, you can test out his analysis of something personal you’ve written, at IBM’s developers site.

In two to five years, Abrams posited, human-facing systems of all kinds will have a more natural interaction with humans that responds to their emotions and their personalities.

But Watson isn’t just interested in your outlook on life. He also wants to figure out how he can scare you.

20th Century Fox got in touch with IBM before the release earlier this month of its Ridley Scott-produced suspense/horror/sci-fi film “Morgan,” a tale of “a genetically engineered child gone wrong.” (Apparently, the genetic engineers in the movie story didn’t think of employing a system like Watson to give the kid a sunny personality.)

IBM Research sat Watson down, so to speak, to watch and visually analyze via APIs and machine learning more than a hundred horror/thriller movie trailers. Watson’s task: figure out what creates suspense, fear and horror in terms of sound, speech, picture, scene and emotion.

With that cram course in thriller films, Watson picked nine shots or scenes in “Morgan,” each ten seconds or longer, and a human film editor at IBM cut them together into the Watson-inspired trailer. IBM said it’s being shown on YouTube, but not in theaters:

And here’s the “real” trailer that the studio did on its own:

IBM Research Scientist John Smith told me that the Morgan experiment was the first of its kind, an effort to assess how well a computer can analyze video for emotional content, select shots or scenes and assist a film editor.

Watson was trained for that task, he pointed out, not programmed. Even before Fox approached IBM, the tech giant had been busy educating Watson with hundreds of thousands of still images that had been tagged as being primarily happy, sad, suspenseful, scary and so on.

Watson analyzed the visual and audio elements in the 100+ trailers for each of those emotions, and its machine learning found patterns that were then used for the “Morgan” project. There was not, however, any training for Watson to discover what a “story” was, so narrative structure was left to the editor.

Smith said that the turnaround time for pulling the “Morgan” shots and editing the trailer was a day, as opposed to weeks or months for a regular trailer. IBM doesn’t have any metrics on whether Watson’s trailer is successful in creating the intended emotions or encouraging moviegoers to check it out.

With this kind of emotional training and experience, sooner or later Watson will find a personality he can call his own.