There’s a lot of noise and very little signal on Twitter, and sometimes it can be hard to know what to pay attention to. A team of scientists might be able to help with that, though, because they’re developing algorithms to sort the truthful tweets from the lies.
Slate reports new research, due to be published in the journal Internet Research next month, which uses a series of tests to predict whether tweets are true or otherwise. It looks for obvious clues which humans spot instinctively: messages are more likely to be true if they come from a well-followed source, are longer, or contain URLs, for instance. Language is important, too: question marks, exclamation marks, and first- or third-person pronouns all hint that a tweet shouldn’t be trusted.
Roll that all together, and the researchers — Carlos Castillo, Marcelo Mendoza, and Barbara Poblete — have developed an algorithm that can tell if a tweet’s truthful 86 per cent of the time. That’s not a perfect 100, but it’s a damn sight better than a 50-50 guess.
Of course, a human could perhaps do better right now, but the algorithm can crunch through mountains of tweets in minutes. Perhaps Twitter clients should come with an optional truth filter soon? [Slate via Verge]