Our collective ability to sift truth from untruth is obviously important. If we can't agree on the barest facts, it makes it harder to do politics, business, or even transact as neighbors. And you might say our ability to fact-check isn't that great at the moment. We're inundated with un-truthiness every day—political, commercial, and from the media itself—and we only seem to bat back a limited amount.
Could automating fact-checking increase our capacity to cope with this bullshit force field?
About half a dozen groups are working on systems that automatically track the news and decide if it's factually accurate. They do this by cutting up text into sentences and putting it through a process of spotting and checking claims. Full Fact, Britain's "independent fact-checking charity," describes the process in a new report. It says that "we can scale up and speed up fact-checking dramatically using technology that exists now." And, that we needn't wait for artificial intelligence before making big inroads on the problem.
"A successful automated fact-checking system is one that saves fact-checkers and journalists time, and makes fact-checking more effective at limiting the spread of unsubstantiated claims," the report says. "Perfect accuracy is not achievable with existing technology."
Full Fact says it's months away from putting tools in journalists' hands, including Hawk, "a claim recognition engine" that monitors debate and spots previously identified statements, and Trends, which tracks how commonly claims appear and who makes them.
Full Fact is looking to collaborate with other auto-fact-correcting projects including Emergent, Storyzy, LazyTruth, Chequeado in Argentina, Les Décodeurs at Le Monde in France, and Politifact. It hopes to standardize approaches, so progress can be made quicker, and it suggests shared monitoring around 14 sources worldwide, so it's possible to cover more ground.
"We cannot suddenly achieve the artificially intelligent holy grail of a machine that can replace everything human fact-checkers do. That is a very long way off. But there is still a lot we can do," the report says.
Read the roadmap here.
Have something to say about this article? You can email us and let us know. If it's interesting and thoughtful, we may publish your response.