CBS Turns to Artificial Intelligence to Glean Viewers’ Emotional Responses to TV Shows
CBS’s research team has resorted to using the dispassionate logic of machines to get a faster read on how viewers respond emotionally to its TV shows.
The broadcaster started using the data-analytics platform developed by New York startup Canvs, which uses proprietary artificial-intelligence processing (AI) to parse natural-language comments.
CBS first started using the Canvs Surveys tool to automate the coding of open-ended responses to surveys fielded by its research team starting in the fourth quarter of 2018. Now it uses Canvs to process feedback on its entire slate of programming and tentpole events, including viewer response to this month’s Super Bowl LIII and Grammy Awards broadcasts.
The tool measures and categorizes consumers’ responses to characters, plot lines and other topics (like related shows), using a standard set of emotional tags such as “love,” “excited,” “bored,” “sad” or “anxious.”
According to Radha Subramanyam, chief research and analytics officer at CBS, the real power of the AI system is its ability to crunch unstructured data far more efficiently than human researchers can.
“When you are running thousands of surveys, humans can’t get through all of this – you end up doing things like skimming the results,” Subramanyam said. “This saves us a lot of time, and it helps us get better insights.”
More than 6,000 individual surveys are currently conducted at the CBS Television City Research Center in Las Vegas during its pilot-testing season alone.
The Center then shares its research with execs across the company, and also furnishes the findings to showrunners and producers to help inform business decisions. For instance, if fans of a show express a strong passion around a particular character, creative execs may choose to focus more on that character’s storylines.
With Canvs Surveys, CBS can now field more open-ended surveys and also expand the set of questions it includes on each one.
In 2014, when Canvs first launched, it was focused predominantly on analyzing social-media chatter about TV shows. Last year the company developed the new Surveys application to let researchers apply emotional-response analytics to their own data sets.
There’re now more than 1 trillion words, phrases, slang terms and idioms in a lexicon, which the company maps to standardized emotion-based categories. “We have literally billions of comments that reflect the ways people feel,” said Jared Feldman, founder and CEO of Canvs.
Over the last five years, industries like financial services have embraced AI, whilst the entertainment industry is still in the early days of exploiting the technology’s full potential, Subramanyam said.
“In media, we’re in the beginning of doing all we can do with AI and machine learning,” she said. “We’re evolving CBS analytics and research for the 21st century.”
Besides speed, CBS’s switch to AI-based survey processing also promises a more standardized analysis of viewer feedback.
“There’s been a ton of standardization around closed-end surveys but there has been no standardization in the open-end, stream-of-consciousness survey space,” said Subramanyam. “Now with AI we definitely have an opportunity to have a more rigorous way of understanding emotion and language that we didn’t have before.”
Other Canvs clients have included Comcast/NBCUniversal, Fox, Turner, Sony Pictures Television and Viacom. The company is also Facebook media solutions measurement partner and has an exclusive deal with Nielsen to provide qualitative insights against Nielsen’s social data.
Canvs has raised $7.1 million in financing from investors including KEC Ventures, Rubicon Ventures, Gary Vaynerchuk’s Vayner/RSE, Social Starts and Milestone Venture Partners.