Twitter Says that its Prompts on Potentially Offensive Tweet Replies Reduced Negative Interactions by 30%

It’s interesting to see what effect even the slightest level of friction can have on negative behaviors online, and how it can reduce unnecessary angst or disagreement, just through a basic heads-up.

Last February, Twitter re-launched its test of warning prompts on tweet replies which Twitter’s automated systems had determined could contain potentially offensive remarks.

Twitter launched a first iteration of the test back in May 2020, before it decided to shelve it during the US election period.

The new test, with an updated format for the alert, was released to selected users on iOS, and in a new report on the experiment, released this week, Twitter says that in 30% of cases where users were shown these prompts, they did in fact change or delete their replies, in order to avoid possible misinterpretation or offense.

Which is a significant amount – imagine if Twitter was able to remove 30% of the negativity and abuse on its platform through a simple nudge to each user.

That number’s also not fully reflective of the potential impact, as this means that, in 30% of cases when it was shown, the tweet author looked at the recommendation, then re-assessed their reply. But in many of the other 70% of examples, Twitter’s algorithms would have got it wrong, and it may not have been offensive at all. That means that the benefit here, in terms of reducing angst, could be huge – and again, it’s amazing to consider how such a simple prompt can have such a big impact.

Though it shouldn’t be overly surprising. Back in 2020, Twitter also added another pop-up alert which appears when users attempt to re-share articles in their tweets without actually opening the article link and reading it themselves first.

After three months of that implementation, Twitter reported that people who saw these alerts were opening articles 40% more often, while people opening articles before retweeting, specifically, increased by 33%.

Again, it’s just a small push, a tiny element of friction in the process. But it can have big results, with the immediacy of social media interaction often leading to less than thoughtful replies, and by simply asking people to re-assess, that, evidently, is often enough to change user behaviors.

The extension of that could also be that more users consider their overall tweet responses more deeply, and what they’re actually sharing online. Which, on a large scale, could have a big impact, and it’ll be interesting to see if Twitter does eventually roll out the alerts to all users.

Right now, Twitter’s extending the initial test, with users in Brazil now also set to get these prompts.

But it could go even further, and it could end up being another significant step in improving the in-app experience.