A conversation with the CEO of FilterLabs about its Filter Russia tool

5 Min Read

A conversation with the CEO of FilterLabs about its Filter Russia tool

Even though we say otherwise, most of us like to put populations in generalized boxes:

- All Canadians are pleasant, funny and spend winters in Florida.

- All Germans are precise and formal.

- All Americans despise mass transit and require 15 square feet of personal space at all times.

Fun and games aside, if you’re a member of one of these populations, you know simplifications miss the cacophony of voices, personalities, very nuanced attitudes and splintered interests that make up your population. It makes sense that attitudes and feelings towards policies and issues are rarely monolithic, and it is difficult for outsiders to understand all the nuanced, localized feelings about a policy or issue.

And yet, countries have fought many a war based on poorly informed assumptions about populations. For example, U.S. intelligence agencies still use the Cold War-era practice of relying on U.S.-based analysts to remotely interpret public attitudes in hot-spot markets.

Social media monitoring: not so great

A possible improvement on the analyst approach to understanding populations is AI-driven social media monitoring. Companies such as Mediatoolkit and Brand24 leverage NLP, machine learning and sentiment analysis to understand a population’s attitudes and feeling about an issue or person to help guide political and product marketing campaigns. However, social media monitoring can be problematic as well:

  • While social media is scalable data, there is an overreliance on it to monitor populations. In geopolitical situations, governments can simply shut social media down.

  • Social media monitoring can’t easily discern local differences. For example, social media monitoring can’t easily parse the differences between attitudes towards Putin in St. Petersburg vs. attitudes towards him in Siberia.

  • Fake accounts cause significant bias in results. Meta deletes an average of 1.5 billion fraudulent accounts each quarter, according to Statista

  • Sentiment is a vague sign and it can only be useful in the context of specific use, you can’t compare sentiment across domains. In other words, there isn’t a one size fits all ML model for sentiment, therefore automated conclusions about sentiment are usually too broad.

Russia-Ukraine War: frontline for AI for Peace

Which brings us to our geopolitical issue of the moment – the Russia-Ukraine War. Given that all Western social media channels have been silenced in Russia, how can anyone outside of Russia understand what everyday Russians are thinking and feeling as the Ukraine war rolls on?

Enter FilterLabs.AI. Founded in 2021, the Cambridge, Massachusetts startup has brought a different approach to understanding a population’s attitudes, and are showing that in Russia with their project, a free tool called Filter Russia.

“The key to our products is we use our proprietary AI to seek out a range of hyperlocal data sources to monitor and analyze,” said Jonathan Teubner, Founder and CEO of FilterLabs.

“In our work analyzing Russian attitudes towards the war, the type of content ranges from message boards for hobbies, such as auto forums and Reddit-like discussion threads to community level sources – organizations that embed YouTube videos into their sites, and the messaging platform Telegram. We are interested in indirect speech more than direct speech, and these discourses contain very rich signals in that regard; there is so much to be learned from them.”

This focus on hyperlocal data sources is a unique concept that makes particular sense in the Filter Russia project – while Russia’s authoritarian government can shut down well-known social media channels and websites, it would prove to be nearly impossible to cut off the potential millions of webpages that reach micro audiences without completely shutting down the internet in Russia.

Teubner said the world is increasingly transitioning toward bimodal warfare – both kinetic and informational warfare – being the norm.  It is in this informational warfare where AI has an opportunity to help governments find paths to peace, or at the very least, not be so quick to jump into kinetic warfare.

“It doesn’t matter whether a country or group is in a conflict stance or a peace building stance, you have to understand the nuances of the population,” said Teubner. “Russia’s greatest failure was a failure of understanding the Ukrainians. We, too, will fail if we overlook the tremendous internal diversity to Russia, or any country we come into conflict with or partner with.”

”We simply can’t afford to treat Russia as an undifferentiated blob of pro-Putin sentiment. We will fail in all our foreign policy goals if we think of these populations as monolithic. There is a significant difference between being pro-war and being ready to act violently on that commitment.”


It makes sense that FilterLabs’s brand of AI for Peace will grow and perhaps displace social media monitoring as the best way to understand narratives within a specific population. It is narrower AI in the sense that AI is used to learn the types of media to crawl, reducing “thousands” of man-hours in pinpoint-specific media, according to Teubner. With AI focused on finding the right media, analysts can avoid the weaknesses of social media’s localized content and too-broad sentiment analysis.

“We are getting calls from intelligence services and governments around the world, both for interest in peace-building and security. And the great thing is, it’s public content – no one’s privacy rights are being exploited,” said Teubner.

Mark Beccue is a principal analyst contributing to Omdia’s Artificial Intelligence practice, with a focus on natural language and AI use cases.


About the Author(s)

Mark Beccue, Omdia principal analyst

Mark Beccue is a principal analyst contributing to Omdia’s Artificial Intelligence practice, with a focus on natural language and AI use cases. Based in Tampa, Beccue is a veteran market research analyst with 25 years of experience interpreting technology for business. He is a frequent speaker, panel moderator and conference chair.

Prior to joining Omdia | Tractica, Beccue was an independent consultant/analyst who provided custom and syndicated qualitative market analysis, with an emphasis on mobile technology. Previously, he was a senior market intelligence analyst at Syniverse with responsibility for identifying trends and opportunities. Beccue also served as a senior analyst at ABI Research, where he concentrated on mobile consumer technology. He has been cited by international media outlets including CNBC, The Wall Street Journal, Bloomberg Businessweek, and CNET. Beccue holds a Bachelor of Science degree in Journalism from the University of Florida.  

Keep up with the ever-evolving AI landscape
Unlock exclusive AI content by subscribing to our newsletter!!

You May Also Like