Communications of the ACM
Two Apps Aim To Tell The Truth And Nothing But The Truth
As people consume more and more media online — much written not by de facto trusted sources but by "citizen journalists" unencumbered by trained editors — some believe there is a growing need for content to be evaluated — carefully.
That is why two new software tools — both works-in-progress — are being built to help readers distinguish truth from BS.
One depends on experienced journalists to fact-check content while the other uses "collective annotation" to inform readers what other readers believe is true or false.
The former, Truth Goggles, is the brainchild of Dan Schultz, an MIT Media Lab grad who created the browser bookmarklet as part of his master's thesis.
Once triggered by a bookmark or browser plug-in, the JavaScript gathers up a Web page's text and sends it to a server where the API scans it, identifies the phrases that have been previously fact-checked, and highlights those phrases on the Web page. Another click on the highlighted text gives more details about whether the phrase was deemed true or false — plus an explanation.
Currently the prototype's single source is PolitiFact.com, whose reporters fact-check and evaluate only political content.
"Subsequent versions will be connected to additional sources — other professional fact-checking organizations like Snopes.com and FactCheck.org — so we can branch out and evaluate any kind of information with the potential for inadvertent — or intentional — misinformation," says Schultz.
But first, Schultz needs to conquer the high hurdle of "paraphrase detection." Because the same statement can be written many different ways, the text on the Web page may not be word-for-word identical to the fact-checked phrase. As a result, the Truth Goggles demo
had given numerous false-positives until Schultz stripped out the paraphrase detection. He hopes that problem will be solved once he incorporates a new natural language processing platform called Luminoso. If successful, he hopes to have a functioning version in four to six months.
Meanwhile, a second "truth-seeking" app — Hypothes.is — is being developed by a San Francisco-based, non-profit venture founded by Dan Whaley, who started one of the first online travel-booking systems in 1995.
"Hypothes.is depends not on experts or professional journalists but on 'crowd annotation,'" says lead developer Randall Leeds.
He observes most blogs have a comment box at the bottom "that's often very difficult to use to have a sustained conversation. Hypothes.is is a platform that facilitates storage, search, and discovery of annotation."
Users who want more information about an article will be able to activate Hypothes.is by clicking on a browser button bringing up indicators alongside the browser — as in a Google document — that show where others have commented. Users can read the comments, respond, or vote as to whether the comments are accurate or inaccurate.
"While we are developing an interface that's browser-based," says Leeds, "we also want to build the tools to enable future e-readers, PDF readers, and so on, to allow users to discuss content wherever it appears in a way that's agnostic to the format or access mode. That's our long-term vision."
Currently there is an early developer prototype of the interface along with the open-source code. The next step will be creating a functioning peer-review system. The goal is to have a more public beta available around the end of Q1 2013.
"Hypothes.is is not a tool to tell people whether something is true or not — it's a tool to let people have their input and, by having more information, they can decide for themselves whether it is true or false," says Leeds. "But, at the end of the day, I can definitely see where it might have upward pressure on information quality."
Videos:
• Introduction to Hypothes.is
• Are Truth Goggles Too Good To Be, Uh, True?
Paul Hyman is a science and technology writer based in Great Neck, N.Y.
No entries found