Byrony Gordon of the Telegraph wrote in frustration this week about how social media is turning us into idiots. She chronicles several tweets and trends across social media in the aftermath of the attacks on Paris as evidence that, "...social media hasn't just turned people stupid - it has also turned whole organisations into unthinking idiots whose knee jerk reaction in such situations is not to uncover the truth but get hits." Gordon's proclamation flies directly in the face of James Surowiecki's notion of the scientifically-based "wisdom of crowds."
Surowiecki draws upon science-based analysis drawn from research on the Central Limits Theorem, which explains how large probability samples produce great estimates of phenomena in the real world. Gordon's analysis draws conclusions from a far less systematically drawn sample. Her sample suggests that people feel less safe after the recent attacks in Paris, despite the reality that attacks like these have been happening around the world for a long time. Her analysis highlights specific examples of idiocy among the world of tweets and wall posts, such as the claim that the Eiffel Tower went dark for the first time since 1889 (a preposterous claim, for sure) and the massive number of "lemmings" who liked and retweeted the claim.
These examples of idiocy are indeed true and the numbers of people echoing them are indeed large and troubling. But even in the dark tower example, fewer than 60,000 people retweeted or liked it among the hundreds of millions of Twitter users and the billions of tweets they share. As Daniel Kahneman and Amos Tversky found in their seminal work on Prospect Theory, even statisticians who should know better will overvalue a recent contrary example and let it undermine grounded statistical conclusions.
Gordon draws conclusions about the idiocy of crowds based on a heuristic (mental shortcut) that privileges the most recent and extreme examples of social media posts over a systematic analysis of all relevant social media posts based on a probability sample. In a probability sample, all relevant posts have an equal chance of being selected for analysis. But Gordon analysis is based on seeking examples of idiocy and using those posts as the basis for her analysis.
As a result, Gordon is able to create a narrative that makes the crowds look like idiots because it is easier to find idiotic posts. Alternatively, a systematic analysis of social media posts is likely to find, contrary to Gordon's findings, that people are by-and-large as scared now as they were before Paris and aware of other attacks around the world commensurate with the media coverage they actually got.
But that story is boring and does not sell the news.
Don't get me wrong, I am as guilty as Gordon when it comes to spinning provocative yarns out of selective evidence, but at least I do my best to acknowledge when I am doing so. And Gordon is reporting on what some people really are saying on social media. And those people are very likely idiots, or at least acting like idiots for the moment. But to draw the conclusions that we are more idiotic now and more so because of social media is just not justified. One could just as easily construct a narrative of smart reactions to Paris on social media by applying a different filter for selecting example posts.
Perhaps this is what makes social media so compelling: it is all things to all people. People can find information that interests them, educates them, challenges them or reinforces what they already believe. But as analysts of what social media means to society and how it affects us, we really do need to be more systematic if we hope to create analyses that are more likely to be true.