Truth Goggles: digital tools for truth vigilantes, part 2
News from anywhere, even the New York Times, always warrants a skeptical eye. Could digital tools help alert news consumers when sources quoted in news stories might be bending the truth? On Tuesday I offered a roundup of ideas from digital media experts. Today we’ll take a trip to technology’s bleeding edge. Dan Schultz of MIT explains his ambitious plan to build an “automatic bullshit detector”...
News from anywhere, even the New York Times, always warrants a skeptical eye. Could digital tools help alert news consumers when sources quoted in news stories might be bending the truth?
On Tuesday I offered a roundup of ideas from digital media experts. Today we’ll take a trip to technology’s bleeding edge. Dan Schultz of MIT explains his ambitious plan to build an “automatic bullshit detector”...
Schultz is a graduate student at that prestigious, creative and slightly crazy institution, the MIT Media Lab. His thesis project, Truth Goggles, is intended to be “a magical button. When pressed, it will tell you (an average person who just wants to know what is happening in this crazy world) what is true and what is false on the website you are viewing.”
That’s a bold goal, but Schultz has already made considerable progress toward it. ...Okay, to a non-programmer, so far it may not look like much progress. In fact, Schultz points out that at this time the Truth Goggles demo isn’t functioning. But Schultz is working on it, so over the next few months this will be an interesting project to watch.
“I graduate in May, so for the next dozen or so weeks I’ll be working on it hard.”
If Schultz can get this working—even modestly—election coverage could get a little more interesting, engaging, and useful.
Here’s his original vision:
Truth Goggles would be an optional plug-in for your web browser, or a bookmarklet that you click on when viewing certain web pages. When activated, it would highlight potentially dubious claims in news coverage, to trigger users’ critical faculties before they believe what they see. It also would provide access to some of the evidence which calls that claim into question.
To support this, Schultz arranged access to private application programming interfaces (APIs) from the fact-checking site Politifact. So in that sense, his software is a bit of a Mechanical Turk. Truth Goggles doesn’t automatically distinguish lies from truth; it looks for words and phrases contained in the Politifact database (which is created by humans) and then makes assessments about which claims to highlight.
...At least, that was the original plan. Since he got started on this, Schultz starting thinking more long-term.
“Repurposing the work of a site like Politifact became a little bit kicking the can down the road. What happens when those guys are wrong? The whole point is to try to get the audience to think harder. So the system is now less deterministic. I’m trying to encourage readers to be more critical at strategic points.”
In the real world, truths and falsehoods often are separated by a large gray area, so Schultz is also reconsidering the binary nature of highlighting text. “Maybe text that is less believable gets a little grayed out,” he ventured.
Then there’s the issue of how to help readers dive deeper into potentially dubious claims. Users might hover over a bit of highlighted or grayed text and a popup window might appear listing the results of fact-checking efforts on that claim. Or perhaps a list of links to fact checking organizations or news outlets which have investigated that claim. Schultz is considering these and other options as he refines the user interface.
Also, it might be possible to implement Truth Goggles for a particular news site or app, for competitive advantage or as a premium service.
The code base for Truth Goggles is open source, so any programmer can take it and modify it. This could be done in good or evil ways.
For instance, a crowdsourced fact checking project like Truthsquad from Newstrust might use Truth Goggles to extend the reach of their work beyond their site, and perhaps grow their base of volunteer fact checkers.
But someone might just as easily roll a version of Truth Google that gauges facts based on a skewed set of sources (say, APIs from the government of North Korea), or even sources selected by the individual user. But Schultz doesn’t like those ideas.
“What I don’t want is a system that lets you, as someone who’s from a particular political persuasion, only get exposed to likeminded statements,” he said. “The whole point of fact checking is to help people question the messages that are presented to them.”
But if Truth Goggles gets up and running, eventually it (or something like it) might serve as a platform to support new business opportunities for journalism.
“The big thing is that this kind of credibility layer interface offers a good way to repurpose the ingredients of an article,” said Schultz.
Schultz acknowledged that good fact checking takes considerable effort, so the problem of scale needs to be addressed. “This tool is only as useful as the data that’s fed to it,” he said. “Politifact only checks about 40 political claims per week. That’s impressive, but it pales in comparison to the need. There are thousands of political claims made every day.”
So if news organizations, technologists and journalists collaborated to create a standard format for fact checking claims, this would yield databases that would generate standard APIs that could feed into Truth Goggles. This would distribute the fruits of journalistic labors in a useful new and highly visible way across the internet. That could drive traffic back to the original fact checking or news sites, supporting the conventional ad-based news business model.
Or perhaps entrepreneurial journalists and librarians might create specialized fact-checking services that could be delivered as premium content through Truth Goggles as a platform. This could be done for any topic (maybe finance or health news), not just political coverage.
Applied internally to a content management system or blogging tool, Truth Goggles also could become a support tool for journalists—the fact checking analogue of a spell checker, calling attention to points which might warrant further investigation before filing a story. News organizations, individual journalists, and others might be willing to pay for such a tool, if it worked well enough.
So where does Truth Goggles stand? Schultz says the back-end source code is “pretty much put together. It’s not necessarily usable at consumer scale, but it might work for R&D, something a news outlet might want to experiment with. It’ll be ready to rock at some scale before the election happens in November, though.”
The News Leadership 3.0 blog is made possible by a grant to USC Annenberg from the John S. and James L. Knight Foundation.