“A lie flies,” wrote Jonathan Swift 300 years ago, “and the truth comes slowly after it,” and he means the speed with which a lie spreads. Imagine that when Dean Swift wrote this, there was no Twitter.
Researchers from the University of Sheffield have managed to get an EU grant to develop an automated system that will check the degree of trustworthiness of online social media posts.
The internet, as we all probably know by now, is an environment that thrives on everything. From the harsh truth to fraud and white lies. We have seen untruths spread like a wave on Twitter several times. Optimistic researchers probably think that there must be a way to preserve the good side of social media, such as the immediate and global distribution of information, while simultaneously suppressing the dark side.
The ambition for a computer system that will be able to sort truth from lies on Twitter or elsewhere must surely have our algorithm fetish, and the most strange thing is that the software is going to replace human judgment. Is it possible? We know that algorithms are designed by humans, and their operation depends on the human who made them. So what are the assumptions that will govern the lie detector of social media?
According to Sheffield’s press release, the system will aim to “classify online rumours into four types: speculative rumours, such as interest rate hikes; controversy, such as whether the MMR vaccine is beneficial; misinformation, where something untrue starts to spread; and misinformation, which is done with malicious intent.” If something doesn’t fit into these categories, they plan to “automatically categorise sources to assess them.” Sources can be news organisations, individual journalists, experts, eyewitnesses, the public themselves or automated bots. It will also look at history and background – we know there are Twitter accounts created solely to spread false information.”
At first glance, all of this seems like a very good idea. But there are immediate examples that would have led to misleading judgments. Authentic news agencies, for example, are sometimes complicit in spreading misinformation (see the New York Times with the alleged nuclear weapons of Iraq). Sometimes there is the lone rebel who says the right thing, but the institutional authorities may disagree (see Galileo). So you understand the dangers of the argument. Of course, there is also the other dangerous side. In an optimistic scenario, academics finally manage to create a magic lie-detecting software. Who will protect us from those who operate it?
The publication was published in TheGuardian under the title Believe it or not, a social media lie detector is being developed.
