Trending: Key Zillow Group exec Greg Schwartz leaving after 12 years, amid massive changes at real estate giant
PNNL's Svitlana Volkova
Svitlana Volkova, a data scientist at Pacific Northwest National Laboratory, is part of a team of researchers who analyzed cryptocurrency discussions on Reddit. (PNNL Photo)

Computer scientists from the Pacific Northwest National Laboratory have mapped the ebb and flow of Reddit’s discussions about cryptocurrency — not only to see how online chatter can predict market behavior, but also to gain insights into how disinformation goes viral.

“Cryptocurrency is a very good proxy program for disinformation,” said PNNL data scientist Svitlana Volkova, one of the authors of a study presented at the Web Conference 2019 in San Francisco.

The ups and downs of cryptocurrencies have been much in the news over the past couple of years, as have the controversies associated with disinformation campaigns like the ones orchestrated by Russian agents during the 2016 presidential campaign. And cybersecurity experts are seeing evidence that the disinformation battle is already ramping up for 2020.

Tracking disinformation scientifically can be a challenge, however, because the perpetrators tend to blend in with the crowd. On a broad topic like presidential politics, it’s hard to come up with an algorithm that focuses in on what’s true vs. what’s false.

It’s easier to look at how information gets passed along on well-defined Reddit discussion forums devoted to specific cryptocurrencies such as Bitcoin, Ethereum and Monero. So Volkova and her co-authors — Emily Saldanha and Maria Glenski — conducted an analysis of tens of thousands of Reddit comments made on the forums for those three crypto coins between 2015 and 2018.

The team set up parameters to measure how fast discussion threads took off, how much volume those threads generated, how many people participated and how engaged they were. They saw clear differences in activity patterns.

Bitcoin, the most popular cryptocurrency, generated the most activity: On average, there were 3,600 comments posted each day for Bitcoin, compared with 500 for Ethereum and 380 for Monero. People also tended to respond twice as quickly to Bitcoin posts than to posts about the other two coins.

Discussions about Ethereum, which is a cryptocurrency as well as a blockchain development platform, had the largest possible lifetimes. But Monero, which the researchers said is favored over Bitcoin for illegal transactions on the Dark Web, had the largest median lifetimes.

Monero posts were also five times as likely to attract follow-up commentary than posts about the other two cryptocurrencies.

“These social signals are quite useful, and by incorporating them with machine and deep learning, we intend to build predictive models that hit on causal relationships between different variables so we can explain model decision-making processes,” Volkova said in a news release.

For a separate study that’s yet to be published, the researchers devised a system to extract social signals from postings to Reddit and Twitter as well as the GitHub coding platform, and link them to a rise in Bitcoin prices.

“We were able to predict it with very decent accuracy,” Volkova said. The team is also working on models for tracking pump-and-dump investment schemes on Telegram channels (as are other researchers).

Such models will be incorporated into follow-up studies on how disinformation is spread, and what types of social-media actors are most likely to spread it.

Volkova said that PNNL will be in charge of an upcoming team challenge focusing on the role played by persistent groups on social media platforms, as part of a cyber initiative funded by the Pentagon’s Defense Advanced Research Projects Agency. (DARPA also supported the cryptocurrency study through its SocialSim program.)

Network science is making significant headway in understanding how social networks are structured, and how those structures influence the flow of information. But does that mean we’re any closer to defeating disinformation?

“Mitigation is harder,” Volkova said. “We have the models right now that can detect disinformation and can say, ‘OK, look, this is not true, this is false or misleading.’ We can make the audience aware of the way disinformation spreads. But then we have to somehow coordinate with the social platform providers to actually make a difference.”

Your move, Reddit … and Twitter … and Facebook …

Subscribe to GeekWire's Space & Science weekly newsletter

Comments

Job Listings on GeekWork

Find more jobs on GeekWork. Employers, post a job here.