Yuriy Gorodnichenko wants you to know that he appreciates the irony of his paper on how political information spreads across Twitter becoming a victim of … how political information spreads across Twitter.
Gorodnichenko, a UC Berkeley economics professor, published a paper this month arguing that automated Twitter bots could have played a small but potentially influential role in the 2016 Brexit vote and 2016 U.S. presidential elections.
“Given the narrow margins of victories in each vote, bots’ effect was likely marginal but possibly large enough to affect the outcomes,” Gorodnichenko wrote, along with his co-authors Tho Pham and Oleksandr Talavera from Swansea University in the U.K.
The paper was picked up by Bloomberg on Monday, and from there it lit up Twitter. Users on the left pointed to the paper as proof that Russian-backed Twitter bots had elected Trump, while those on the right dismissed the paper entirely and journalists and some economists questioned the authors’ claim that bots swayed the election.
— Derek Cressman (@DerekCressman) May 21, 2018
This is just MADDENING! I to this day, don’t understand how our Intelligence agencies either #1 missed this completely #2 kept it from us #3 I don’t know what they hell happened but this SUCKS!! https://t.co/eAai2Pk7ud
— 🔥 FiredUpResistance #FBR 🔥 (@Vegas040805) May 22, 2018
The problem? That isn’t quite what the paper was arguing.
“In the paper, we never said something caused something,” Gorodnichenko said Friday after watching his research conclusions metastasize across Twitter all week. “We talk about association and predictive power, which isn’t causal.”
Instead, Gorodnichenko said he and his colleagues were trying to make a point that information posted on Twitter is spread and absorbed within about an hour, and often shared by the most ideologically polarized users. Bots work like jet fuel, helping spread headlines (real or not) more quickly, and by encouraging humans to be more extreme.
“We examined how quick the news cycle was on Twitter,” Gorodnichenko said. “Everybody who has something to say about an issue would say it within a few hours.”
That was true of his paper too.
“My lesson from this reaction was that there was something like an echo chamber for our research,” he said. “If you see X and you like X, you spread it, and if you don’t like X then you don’t spread it.”
The experience underlined why, Gorodnichenko argues, news consumption on social media needs to be studied in an academic, sober way. The next steps might include studying how information on Twitter actually influences opinions or sways elections.
“We see in the data that Twitter has predictive power, but how this works we don’t know,” he said. “We need to do more research. I think people under-appreciate the significance of new media in spreading information.”
Contact Will Kane at firstname.lastname@example.org