The social network said that it’ll display a prompt on Android and iOS for conversations it thinks can get intense. The prompt seems to remind you of basic moral values like “Remember the human (in the conversation) and “diverse perspective have values.”
This is a work in progress as we learn how to better support healthy conversation. pic.twitter.com/x6Nsn3HPu1 — Twitter Support (@TwitterSupport) October 6, 2021 The prompt doesn’t come as a surprise to me. I’ve seen people fight on topics ranging from environmental awareness to the best flavor of a snack. ANY discussion can get heated on this bird site. But I’m not completely convinced that reminding trolls about human values will stop them from slinging someone’s name in the mud. Effectiveness aside though, there are also plenty of technical difficulties in detecting these discussions. If an AI is at work, it will need to understand the correlation between words and cultural connotations of sentences used in a particular conversation. In the past, trolls have fooled algorithms to believe a conversation is non-toxic by changing some words, inserting positive words in sentences, or using different spellings of a swear. Also, at the moment, this feature seems to be limited to English. In a country like India, where a large set of the population is multilingual, someone can easily slip in a few words from another language and trick the algorithm into thinking a conversation is going well. We’ve asked Twitter to share more details on how the algorithm works, and we’ll update the story if we hear back. Twitter started using machine learning to reduce toxicity on the platform last year. Its newest experiment of intense conversation prompts is perhaps one of the more ambitious endeavors in that category that’ll test its AI might on multiple levels.
