NAB 2018: Analytics Scientists Look at Social Media and Bots

Fabric Media chief executive/founder Jason Damata led a discussion at NAB with two experts in the field of social media intelligence. Dr. Indraneel Mukherjee founded LiftIgniter, which is “a machine learning personalization, recommendation and discovery engine” for websites and apps to have one-on-one conversations with users. Dr. John Kelly is chief executive at Graphika, which turns “network relationships into dynamic maps of social influence, enabling precision targeting and action to drive business results.”

Damata asked both scientists to describe what they find most challenging in today’s market, especially as data has proliferated over multiple platforms. “Our biggest challenges in the brand/agency space is conceptual,” said Kelly. “We have a technology that uses advanced machine learning to see patterns in relationships across various social media platforms, and we have to describe it to people who are using techniques that have been around since the 1930s.”


Graphika’s solution is focused not on individuals but on “communities of interest.” “Communities arise from behavior, micro-decisions, and we’re interested in mapping out these communities and targeting them as a group,” he said. “It’s a different value proposition.”

Mukherjee reported that his company does not track individuals either. “We only use data from a particular website to help improve their experience on that website,” he said. “When you go to YouTube, the job is to show you videos that you are likely to engage with next. It may not be related or make sense, but if it’s interesting to you it’s doing its job, and killing all the other video websites.”

In response to a question as to how much of this is “social engineering,” Mukherjee noted that, “it’s not just analyzing behavior but acting on that data in real time.” “Does that bias peoples’ behavior? Potentially,” he said. But, he noted, “the biggest predictor of what you did in the next five minutes is what you did in the last five minutes.”

Kelly said that recommendation engines are criticized because as “they optimize for eyeballs, they lead to more and more extremist videos.” “A lot of what happens online is a combination of robotic and other non-human activity,” he said.

Bots are a hot topic, and both Kelly and Mukherjee addressed it. “Detecting bot traffic isn’t hard, unless it’s extremely sophisticated,” said Mukherjee. “We and most analytic companies can use machine learning to filter out bot traffic. But, if all you care about is more clicks, you can tune it towards that objective, creating bubbles.”

Kelly agreed, noting that, although there is a good practice of bot identification, the second and third generations of bots have become increasingly sophisticated, with the most recent being “highly crafted sock puppet personae.” The future should be programmatic, opined Mukherjee. “The marketers are in control of what the algorithms do and that they don’t cross the boundary and get too targeted,” he said.

“The social advertising industry is facing a reckoning now,” concluded Kelly. “It’s being forced to grapple with these issues. Automation can be part of the solution and we won’t see bots go away any time soon. Intent as well as outcome is important. There will be sophisticated approaches once we settle on the outcomes we want.”