The Dunning-Kruger effect is one of the most famous—and predictable—biases in human behavior. It posits that people who don’t understand a topic also lack sufficient knowledge to recognize that they don’t understand it. Instead, these people know just enough to convince themselves that they completely grasp the topic, with results ranging from hilarious to painful.

Inspired by the widespread sharing of blatantly false news articles, a team of US-based researchers looked into whether Dunning-Kruger might be operating in the field of media literacy. Not surprisingly, people do overestimate their ability to identify misleading news. But the details are complicated, and there’s no obvious route to overcoming this bias.

Evaluating the news

Media literacy has the potential to limit the rapid spread of misinformation. Assuming people care about the accuracy of the things they like or share—something far from guaranteed—a stronger media literacy would help people evaluate if something is likely to be accurate before pressing that share button. Assessing the credibility of sources is an essential part of that process.

Evaluating credibility is a skill—and it’s one that people can clearly be bad at, leaving them open to the Dunning-Kruger effect. So the researchers of the study arranged a set of experiments determine whether Dunning-Kruger was an issue.

The basic test was straightforward. Relying on a couple of YouGov panels, the researchers gave the participants a set of actual headlines and asked the participants to rate them for accuracy. Without being told the test results, the participants were then asked to rate their own performance compared to the average person.

Assuming that people could rate themselves accurately, you’d expect that about half would rate themselves above average while the other half would rate themselves as below average. But that’s nowhere close to what was seen. Ninety percent of the participants claimed they were “above average in their ability to discern false and legitimate news headlines.” The average self-reported ability outperformed 69 percent of other people.

On its own, this result could simply be representative of a general overconfidence. To determine whether the least competent were the most likely to overestimate their abilities, the researchers broke up participants into four groups based on their performance. The bottom quartile accurately judged accuracy about 10 percent of the time, and the top quartile was close to 90 percent accurate.

The top quartile also underestimated their own performance by about 15 percentage points. The above-average quartile were roughly accurate in terms of their self-assessment, and performance estimates went downhill from there. The lowest quartile showed a 40 percentage-point gap between their self-assessment and their actual performance. While the less competent didn’t rate themselves as highly as the top performers, this is clearly a case of Dunning-Kruger.

In news that should surprise no one, men were more likely to have an inflated sense of their own media literacy. Republicans also fell into this category, which is not shocking given the high levels of misinformation about the election and the pandemic currently appearing on right-wing news sites.

Big mismatch, minor effects

While those are important findings on their own, the big question is how this inflated sense of competence influences people’s decisions about consuming and sharing news reports. Here, the researchers benefitted from the YouGov panel, where several participants had agreed to share their browsing history anonymously (this was gathered by a combination of browser plugins and VPN service).

The researchers broke down visits to news and commentary sites based on whether the site had a history of spreading misinformation. In terms of exposure to misinformation, overconfidence was associated with a slight increase—in other words, the stronger the Dunning-Kruger effect, the more likely someone was to visit the sites that frequently post false stories. The effect, however, was minor. Those with the strongest misplaced confidence in their own abilities were only 6 percent more likely to view misinformation than those with a reasonable appraisal of their own skills.

A separate set of questions indicated that the misplaced confidence was associated with an increased willingness to share false stories, although the effect was fairly small. This willingness was influenced by whether the false story was consistent with people’s political beliefs. Part of the problem is that people with overconfidence in their media savvy have a harder time discerning true and false stories than people with media skills.

Overall, we shouldn’t be surprised that Dunning-Kruger applies to media literacy as well. And while the effects were small, if they replicate, they’ll help improve our understanding of the misinformation landscape. The new research makes an interesting comparison with an earlier study that indicated the average person was pretty good at recognizing misinformation but didn’t always bother to apply that skill before sharing or liking a story.

“Low performers genuinely believe in their own abilities”

The depressing part of the present research, however, is that there’s a fair bit of literature on attempts to correct for Dunning-Kruger, and most of it describes failure. “Studies suggest that low performers genuinely believe in their own abilities and are not simply making face-saving expressions of self-worth,” the researchers note, and they add that Dunning-Kruger is generally associated with “resistance to help, training, and corrections.”

So even as we get a better grip on the factors influencing the misinformation flood we’re facing, we’re not necessarily getting closer to identifying what to do about it.

PNAS, 2021. DOI: 10.1073/pnas.2019527118  (About DOIs).

https://arstechnica.com/?p=1769175