TOKYO – The number of offensive comments posted on Yahoo News’ website declined by over 20% after the firm introduced an artificial intelligence feature that asks users to reconsider comments before posting, according to the site’s operator LY Corp.
The feature was introduced in September. The program detects vulgar words or offensive expressions in comments as they are about to be posted. It then notifies the commenter and suggests changing the expressions to less offensive wording.
LY compared the number of posts that the AI feature judged to be offensive in the four-week periods before and after its introduction. The firm found a decrease of around 24%, with the overall number of posts having not changed much.
It is still up to the commenter to decide whether to change the wording the AI caught, and they can choose to post without changes. Of the comments modified in response to the AI’s suggestions in the four weeks after the feature’s introduction, about 50% were rewritten with less offensive words.
“We understand that users are voluntarily reviewing their wording, and that has contributed to the integrity of our comment section.”
