BY: Follow @LizWFB
The National Science Foundation (NSF) project designed to track “misinformation” on Twitter has removed portions of its website that monitored political users, including conservatives who used the “tcot” hashtag.
“Truthy,” the nearly $1 million research project being conducted by the University of Indiana, has redesigned its website following the Washington Free Beacon’s initial report on the study.
The service is intended to monitor “suspicious memes” and “false and misleading ideas,” with a major focus on political activity online.
Truthy has received increased media attention since Republican FCC Commissioner Ajit Pai published an editorial last week warning that the project could be misused. “The concept seems to have come straight out of a George Orwell novel,” he wrote.
Pai pointed out that Truthy devoted a section of its website to tracking the use of hashtags such as #tcot, #teaparty, and #dems, evaluating “whether accounts are expressing ‘positive’ or ‘negative’ sentiments toward other users or memes.”
The “Political Topics” page where this information was found no longer exists, and was removed around the time Pai’s editorial was published with links to the website.
Screenshots taken by the Free Beacon in August show the site in its previous form, where it monitored hundreds of conservative Twitter users when they used #tcot. The site recorded the number of retweets, mentions, partisanship of the user, “sentiment,” and language of the tweets.
Twitter accounts listed included the Drudge Report, Ann Coulter, Byron York, and Fox and Friends. The top user listed was Pat Dollard, a conservative filmmaker. The user statistics page on #tcot has been deleted.
At the time, Truthy said the “Top Conservatives on Twitter” hashtag is the “most popular meme we track.” Persons visiting the Truthy website also had the option to “tell us what you think about” any Twitter account using the hash tag.
“We are collecting your responses to build a truthiness and spam detection tool, which will detect persuasion and spam campaigns,” the website said.
Truthy used to tout that it relies on people visiting the site to report on other Twitter users.
“To train our algorithms, we leverage crowdsourcing: we rely on users like you to flag injections of forged grass-roots activity,” a web archive of the Truthy about page states. “Therefore, click on the Truthy button when you see a suspicious meme!”
This line has since been deleted.
The website now focuses on social bots, geography of Twitter trends, and a gallery that shows “diffusion networks” of memes studied in 2010 and 2011, including #usa, and @barackobama.
The Free Beacon asked Truthy’s head researcher Filippo Menczer, a professor of Informatics and Computer Science at Indiana University, why these portions have been deleted. Menczer first said that truthy.indiana.edu is not the Truthy project website, and the changes were a result of updates that happen from “time to time.”
“Actually, this is the Truthy project research site,” he said, linking to a page on Indiana University’s Center for Complex Networks and Systems Research website that includes information about the project.
“The website to which you refer (truthy.indiana.edu) has been used to present some demos, visualizations, tools, and data resources highlighting some broader impacts of the project,” Menczer said. “From time to time we update the demo site to reflect current research, add new demos and updates, and remove out-of-date information and old demos that are no longer working.”
Menczer said monitoring hash tags like #tcot were from research conducted four years ago, though they remained on the site until at least Oct. 14.
“The portions to which you refer were about work we did in the early stages of the project, providing visual analytics about memes about news, politics, and social movements, based on public tweets from around 2010,” he said. “The results of that research were published in 2011 and 2012 (you can find our papers on the project research site).”
“About specific memes, we provided analytics about *all* memes (hashtags, URLs, phrases, etc) that we extracted from public tweets,” Menczer added. “#tcot just happened to be one of millions, and since it is widely used, at one point we used its diffusion network as an illustration of normal, ‘organic’ political conversations on Twitter.”
“This was contrasted with a few examples of memes spreading spam or promoted by fake accounts (‘astroturf’),” he said.
Menczer said the part of the website that asked users to mark other Twitter users as spam was one angle that they exploring that did not prove reliable, and that Truthy is “no longer focusing on those domains of news, politics, and social movements.”
Menczer has defended the project as a nonpartisan computer system incapable of determining whether a tweet “constitutes ‘misinformation,’” and then attacked the Free Beacon in a blog post at the Washington Post for igniting a “misinformation campaign” about Truthy.
However, the “about” section of the Truthy website explains that the service’s purpose is to detect “misinformation” and “social pollution” on Twitter.
“We also plan to use Truthy to detect political smears, astroturfing, misinformation, and other social pollution,” the website says.
When asked how Truthy can detect misinformation and social pollution if those terms are not defined, Menczer said, “There is no attempt to make editorial definitions or decisions about the content of messages.”
“Rather, we are interested in exploring whether there are statistical signatures in the diffusion patterns that a machine learning algorithm could identify to differentiate between normal, organic conversations and attempts to systematically abuse social media in order to drive people out of online conversation,” he said.
The Free Beacon also asked Menczer why detecting "hate speech" was included in the original grant proposal, and how he defines the term. Menczer declined to offer his definition.
“You are referring to a sentence from the broader impact section of the abstract of the grant proposal submitted to NSF in 2010,” he said. “Taken out of its proper context, that sentence can be quite misleading. That passage refers to a proposed public and open web service to allow anyone to access information and visualizations about how memes propagate through social media.”
Menczer said the detecting hate speech line originates from when they were applying for the grant and the NSF asked the researchers to speculate potential uses for the project.
“We do believe that it could help people make better informed assessments of the information that they see online,” he said. “For instance, a user might find it easier to determine that a link will lead to malware.”
Truthy is still looking into “whether it is possible to automatically detect misinformation (rumors, astroturf, malicious social bots, and so on).”
Menczer no longer publicly advertises his support for numerous progressive advocacy groups, including President Barack Obama’s Organizing for Action, Moveon.org, Greenpeace, the Sierra Club, Amnesty International, and True Majority.
He previously linked to each of the organizations on his personal page from his bio at the Center for Complex Networks and Systems Research. The page is now password protected, with a message that says, “Menczer HomerPage is for family and friends only.”
The project is now under investigation by the House Committee on Science, Space, and Technology.
“The NSF is out of touch and out of control,” said chairman Lamar Smith (R., Texas). “The Science Committee is investigating how this grant came to be awarded taxpayer dollars. The NSF must be held accountable for its funding decisions.”
After sending additional follow up questions as to why the site was only recently changed given that Menczer said the research was conducted four years ago, the university said that they will no longer be taking inquiries from the Free Beacon.
“The timing of recently completed updates to the site is purely coincidental to any other events and surely is not in response to outside commentary on the project by any individual,” said Mark Land, Associate Vice President for Public Affairs and Government Relations at Indiana University. “The truth, which I hope you are interested in printing, is that updating the site was a long and complex process done over the better part of a year as students and others associated with the project had time. There is no way the updates released recently could have been done in a few days—or even a few weeks—given the level of staff available to handle such a project.”
“I also wanted to let you know that we will have no further comment to you on this project or the work of our faculty members in this area,” Land said.