A new “lie detector” for Twitter is currently in development, and while the prospect of knowing whether your colleagues really enjoyed the delightfully Instagrammed salad may seem exciting, its true benefits could actually solve one of the biggest problems public social media platforms cause society: malicious untruths.
In the 2011 London riots, for example, rumours began to spread of tigers having escaped from London Zoo causing an uncomfortable mix of confusion, terror and humour. Could a machine-based lie detector, analysing language and context, have saved us from this bizarre state of affairs? According to the University of Sheffield’s Kalina Bontcheva, lead researcher on project PHEME: perhaps.
There are four categories of tweets that misinform people, according to the project team:
- Speculation – such as financial guesswork
- Controversy – unproven accusations
- Misinformation – accidental untruths
- Disinformation – malicious untruths
The technology would take into account a number of tweet characteristics, including the authority of the user and their history on the site. A well-respected, verified journalist would be more authoritative than a brand new account which is spamming scandalous political rumours, for example.
There is no word on whether the analysis extends to metadata such as the location from which the tweet was sent, and the researchers currently have no plans to include media such as images in the analysis, which often form a key part of corroborating or dispelling rumours.
The results of searches looking at current events would then display on a “visual dashboard” to let the user know whether a rumour was likely to be true or not.
It’s an interesting project which is expected to take three years to come to fruition. It would be reasonable to expect Twitter is doing exactly the same thing as it looks to serve one of its most active userbases: journalists and organisations like emergency services and charities.