Crowdsourced fact checking? Hmm...


Why not look at pandemics, instead?

Crowdsourced fact checking? Hmm... /img/all-your-biases-are-belong-to-us.jpg

Crowds of regular people would be as good at moderating fake news on Facebook as professional fact-checkers.

A new working paper from researchers at MIT, studying “The growing stream of reporting on and data about fake news, misinformation, partisan content, and news literacy is hard to keep up with”

has found that, in places like Facebook:

  1. third-party professional fact-checking and news organizations are too few to remove or flag enough fake and misleading news, quickly enough.
  2. a group of “1,128 [presumably random] Americans found on Mechanical Turk” could be surprisingly effective, yelding an aggregate judgment very similar to that of three professional fact-checkers doing the same job after reading the whole articles.

The conclusion would be that

  • ” Individuals still fell for misinfo- but crowds did well”
  • “Crowdsourcing is a promising approach for helping to identify misinformation at scale”

because the amount of news that the 26 full time fact checkers at Facebook (yes, twenty-six. Why should you expect that a company as poor as Facebook could afford more?) can verify is a drop in the bucket, and a counterproductive one too, with “the potential to increase belief in, and sharing of, misinformation that fails to get checked”.

Color me skeptical

The author themselves honestly warn that they “see crowdsourcing as just one component of a misinformation detection system that incorporates machine learning, layperson ratings, and expert judgments”.

Me, I would like to warn you even more, for a couple of reasons at least, and suggest a solution.

First, even if there were zero intentional abuses of the system, it would very likely have, at scale, more or less the same implicit bias problems present in places like Wikipedia. What if, just as an example, the only people with the right combination of skills, time and economic needs to be MTurkers were overwhelmingly white, christians, male of a limited age bracket?

Second, the very idea has the same problem of many other human activities: it works well until it becomes really important, that is until some heavy interest and motivation to alter its course, by infiltrating one’s supporters among the checkers, emerges in the wrong places. This approach may backfire just like some proposals to “make the internet less white and western”

Maybe the very comparison of current professional fact checkers with currently available crowds of almost surely 100% good faith “MTurkers” is not even wrong, in the sense that it cannot scale, or last as imagined indefinitely.

What then?

Why not treat fake news like COVID-19?

What we have here is something bad, highly contagious, in such quantities that it is almost impossible to spot, treat and block each case singularly, for lack of expert personnel. Does it sound familiar?

If it harms and moves like a pandemic, treat it like a pandemic. Lock down fake news. NOT with censorship, of course! Just remove instantness from social networks, just like we are intentionally applying attrition to people’s movements. If fake news behaves like a virus without a vaccine (1), contain them in the same way.

  1. Because the vaccine, in this case, would be the exactly right combination of quality education, empathy and intellectual honesty. Pretty hard to quickly produce in volumes, wouldn’t it?
Stop at Zona-M   Never miss a story: follow me on Twitter (@mfioretti_en), or via RSS