mxx1's avatar

Epistemic Segregation

MIT researchers, Vosoughi, Roy and Aral (2018) found:

False news spreads faster, farther, deeper, and more broadly than true news.

False news was 70% more likely to be retweeted than true news.

So false news is more effectively vectorised.

True news took six times longer to reach 1,500 people.

Novelty was the primary explanatory variable.

Humans, not bots, were the primary drivers of spread.

And other academic studies have semantically assessed false news, consistently finding that it exhibits significantly higher levels of negativity, anger and moral outrage compared to factual news.

So if the point of communication is to get something you want, it’s more effective to figure out a negative false-news angle.

For example, some outrageous made-up negative claims about the competition’s products, or the people in that company is better than promoting the great features of your own product.

It’s a race to the bottom. And then of course you get selective disengagement as a rational coping response.

But disengagement isn’t evenly distributed. People who opt out of social media tend to be better educated, leaving the most manipulatable audiences most exposed, and the better educated thusly exposed to them.

And get this … as a source of information, AI chat is selectively used by the engaged, curious, higher-income demographic – the same group already disengaging from the social media noise.

So AI potentially widens the epistemic gap rather than closing it, simply because it gives the already suspicious a viable alternative.

Map all of that onto politics and we’re heading for a new class war.

This one will be epistemic – a divided electorate where one segment reasons from increasingly high-quality information and the other from algorithmically curated outrage and falsehood.

Political outcomes get decided by the latter because they’re the larger and more emotionally mobilised group.

The informed minority loses democratically to the misinformed majority.

Sound familiar?

That’s a bit alarmist that argument, full of hidden assumptions, and just one way of interpreting the data. However, if I wrote an even-keeled version of this it wouldn’t be noted or read. This blog is a victim of the very trends it describes: thus proving a point. I’m just not sure which one.