Supporting Online Conversations at Scale with AI

Forums, comments sections and social media sites are vital parts of the web – places where knowledge is not just shared but created. For the managers of these platforms, however, maintaining an open, productive and welcoming environment can be a monumental challenge.

Inspired by the potential of early large language models, we created Perspective API, a free service that used machine learning to help platforms and publishers predict how a comment is likely to be perceived by others in a conversation. To allow others to benefit from our learnings with these models, we publicly shared our evaluations, training data and benchmarks.

Unlike traditional tools that focused on helping platform and community managers enforce specific policies, Perspective API showed how AI could be used for a new type of analysis. Moderators could customize their use of the API to identify comments that discourage or foster participation, tailoring the tool to their community's specific goals.

Since launching in 2017, the API has been used by over 1,000 partners in 18 languages almost two billion times a day. For publishers like the New York Times and the Wall Street Journal, the impact has been enormous. The Times was able to triple the number of articles that are available for commenting, expanding spaces for deep engagement among journalists and readers.

To learn how Perspective API works, visit our developer library.