There’s been a long-held narrative that YouTube’s algorithmic recommendation engine steers young Americans toward partisan and conspiratorial content. But a new study suggests that’s not exactly what’s happening.More info:theytlab.com

Researchers used two separate experiments to evaluate how YouTube recommendations affect young Americans’ media diets. In experiment one, participants were asked to view a series of videos—some “real,” some created by researchers—and then rate them as either highly or very low quality. They were also asked to indicate how much they trusted the content creators and how much they liked or disliked them as individuals.

The YT Lab: Collaborations and Partnerships for Expanding Reach

Then, the researchers tracked how often these participants watched a particular genre of video on YouTube. They found that when the “real” videos were ranked as high quality, they were watched significantly more frequently than those that were rated as low quality. The researchers then analyzed the effects of these videos on the recommendations that were later served up.

In experiment two, they used “counterfactual bots”—bots that were programmed to follow a user’s watching history but were otherwise neutral with regard to the content that they watched. These bots were then compared to traditional bots that were following the same viewing history as the real life users they were trained on. They found that the real-life bots followed a pattern of watching conspiracy-related and politically divisive content, but the counterfactual bots didn’t.