Is this guy bothering you? Tinder wants to know.
Last week, the dating app announced it would be launching a variety of new safety features, including a new AI-powered system to help flag potentially offensive messages. The new program will use machine learning to auto-screen for messages that may be harmful or offensive and encourage users to report bad behavior on the app. The system will ask recipients of any flagged messages: “Does this bother you?” and will direct users who respond “yes” to a report form.
The new safety measure is similar to other AI-powered anti-harassment systems in place on other social media and messaging platforms, such as Instagram’s recent anti-cyber-bullying feature that flags potentially harmful comments and encourages users to think twice before posting.
As Wired noted, however, it’s not a flawless system, with the contextual nuance of conversations on platforms like Tinder making it especially difficult for AI to accurately decipher between normal back and forth and potentially harmful messages. An early attempt to flag inappropriate messages based on a list of potentially problematic keywords, for example, ultimately proved ineffective, with the program failing to accurately account for context.
However, because, as Tinder’s head of trust and safety products Rory Kozoll told Wired, “one person’s flirtation can very easily become another person’s offense,” the app’s current message flagging system still hopes to err on the side of caution when it comes to asking users if a match is bothering them. Meanwhile, the system will learn over time which oft-flagged messages are inherently harmless if users repeatedly respond that they weren’t bothered. The end goal, Kozoll told Wired, is for each user to eventually have a completely customized algorithm suited to their specific troll tolerance levels.
The new feature is currently available in 11 countries and nine languages, with Tinder planning to eventually expand it worldwide.
Subscribe here for our free daily newsletter.
Thanks for reading InsideHook. Sign up for our daily newsletter and be in the know.
