Online outrage's feedback loop

What they did: Yale researchers measured expressions of moral outrage on Twitter studying real-life controversial events, such as the nomination of Brett Kavanaugh to the Supreme Court, using machine learning to develop a classifier that could label some 12 million tweets.

• They also studied subjects in a simulated Twitter experiment designed to test whether algorithms, which reward users for posting popular content, encourage expressions of outrage.

What they're saying: "The kinds of discussions we are increasingly having in digital public spaces are life and death kind of questions," Molly Crockett, a Yale associate professor of psychology and author of the study, told Axios.

• "A lot of it is driven by outrage, and I think we as citizens need to know how the design of the spaces that we are interacting in can impact the way we choose to express ourselves."

What they found: Users who got likes and retweets when they posted moral outrage on Twitter were more likely to express outrage in later posts.

• Reinforcement learning — if you reward a behavior, you encourage it — is a "biological fact," Crockett noted.
• "Now we show, for the first time, outrage expressions get disproportionate amounts of reward online," Crockett said. "It's not surprising that if you get rewarded for expressing outrage, you express more outrage in the future."

The researchers also found that while people in politically extreme networks express more outrage than those in moderate networks, members of the more moderate groups were more influenced by receiving likes and retweets for expressing outrage.

• "There might be something special about outrage in this online social network environment where you have these very salient rewards, likes, shares and so on," Crockett said. "It essentially sets up a competitive market for outrage expressions."

• The study was published Friday in the journal Science Advances.

Between the lines: The researchers focused on Twitter because it has the most openly available data for such research, Crockett said.

• But Crockett said she speculates that the results would be similar for any platform that has the same kind of social reward design feature.

The big picture: If design can promote outrage, it can probably turn down the volume as well.

• Twitter said earlier this year it will prompt users to rethink whether they want to send a mean or offensive tweet as a way to encourage more healthy conversations.
• "I think a lot of people recognize that these tools have tremendous potential for good, but also carry a lot of risk," Crockett told Axios. "A lot of us are worried about the potential for harm, and so I hope that we can just keep learning more and stumble forward as best we can."