How GLAAD Is Training Google AI to Be Less Homophobic

At the South by Southwest Conference, GLAAD and Alphabet, Google's parent company, announced they will collaborate to make artificial intelligence more inclusive of LGBT users. This effort is one to end biases in AI. Content related to the queer community tends to generate more hateful comments on the web, which has caused algorithms to process LGBT phrases negatively. GLAAD has found that those unintentionally biased algorithms sometimes censor and delete "bad" LGBT content on social media.

The advocacy group will be working with Jigsaw, a division of the Google parent company that creates tools that deal with abusive comments. The unit, with research coveriing identity terms and a database of hundreds of thousands of real comments, is looking to improve its algorithms. Together they will train AI to recognize which phrases are slurs against LGBT people and which are just representational.

"Our mission is help communities have great conversations at scale. We can't be content to let computers adopt negative biases from the abuse and harassment targeted groups face online." Jigsaw product manager Cj Adams said at SXSW, according to a GLAAD blog post.

The hope is that Jigsaw will be able to better judge which content to spotlight, rather than supressing all LGBT content. This topic is particularly relevant, as many have complained that YouTube, which is owned by Google, has been demonetizing and restricting videos from LGBT creators. Putting unnecessary age restrictions on LGBT educational videos could keep queer youth from accessing the facts they need.

"AI has the potential for amazing benefits, but also has the potential to widen social divisions and further harm marginalized communities like LGBTQ people," James Heighington, chief digital officer at GLAAD, told an audience at SXSW. "That is why it is crucial that we are collaborating with important organizations like Google to build inclusive AI that accelerates acceptance for all people."