Twitter has shared new research on the effectiveness of its offensive response alert

Twitter has managed New research Among the potentially offensive tweet response alert work, which it first launched in 2020, was then re-launched last year as a way to add a level of friction and consideration to the tweet process.

Twitter’s alert prompts use automatic detection to sort out any potentially offensive terms in tweet replies, triggering this warning to add a moment of hesitation to the process.

In February, Twitter reported it In 30% of cases where users were shown these prompts, they actually changed or deleted their answers to avoid possible misinterpretations or offenses.

Now, Twitter has plunged deeply into the real pricing process of caution.

Such as Twitter:

“While it was clear that prompts force people to reconsider their answers, we wanted to know more about what happens when a person sees a prompt. To understand this, we conducted a follow-up analysis How Affects positive results on Twitter over time. Today, we are publishing a peer-review Study More than 200,000 prompts operated by the end of 2021. We have seen that prompts have a positive short-term and long-term effect on Twitter. We are the people that are exposed to a prompt Less likely to compose The offensive answer of the future. “

It’s amazing what a simple step between thinking and tweeting can do.

According to a Twitter study, For every 100 events where these prompts are displayed (on average)

• 69 tweets sent without revision
• 9 Tweets were not sent
• 22 have been revised

These results are consistent with the 30% figure above, but it is interesting to note the more granular detail here and how user behaviors have changed the prompts as a result.

But more than that, Twitter has further found that prompts can have an ongoing behavioral effect on the app.

“We’ve found the effect of being presented with a prompt even beyond the moment of posting. We found that, after only one prompt exposure, users were 4% less likely to compose a second offensive answer. Requested users were 20% less likely to compose five or more prompt-worthy tweets. “

So, although 4% may not seem very significant (although on a Twitter scale, the actual number may be large in this context), the ongoing effect is that users become more considerate in their responses.

Or they may be smarter about using terms that do not trigger Twitter alerts.

In addition, the researchers found that persuasive users themselves found less offensive answers.

“The response rate to prompt-eligible tweets that were offensive decreased by 6% for prompted users. This represents a larger and lasting change in user behavior and suggests that receiving prompt can help users be more aware of potentially offensive content as well as post future tweets. “

Again, 6% may seem like a small fraction, but with some 500 million tweets are sent every dayRaw numbers can be significant here.

Of course, this only applies to tweets that trigger an alert, which will be a small amount of actual tweet activity. But it is interesting to consider the effects of these warning prompts and how such small pushes can change user behavior.

At face value, the results show that Twitter’s offensive response alerts can serve as an educational tool for further consideration, which can help improve on-platform discourse over a wide range, over time.

The biggest way, however, is to help re-align user behavior towards more positive engagement, which can be a key step in reducing frustration and division, as it is often unintentional, or lost in translation, through textual communication that lacks conversational subtlety.

This is an interesting consideration for future platform updates on this topic. And while such prompts may be difficult to expand into new areas, or make them more sensitive, it shows that misunderstandings are a common element of online debate.

The truth is, personally, many of those with whom you disagree online will not be anywhere near as argumentative or controversial. If only we could translate some more of the personal features in the online chat – in the case of immediate feedback and action, it is worth taking a moment to consider whether the sender of this tweet was, at least in some cases, not intentionally intended to annoy or confront you in this way.

In other words, Twitter is not real life. People like debates and get involved in emotional debates. But really, it’s probably just some lonely person trying to find a connection.

The less you take it personally, the better for your mental health.

You can read the full study on Twitter Here.