Social giant Facebook has begun to roll out a hate speech checker to a limited number of users, asking them on every post they view if it contains hate speech. This almost certainly means that Facebook is trying to enhance its AI-based detection of hate speech, and possibly other unsavory text on the service. The check manifests as a very simple bit of text at the bottom of a post asking you whether it contains hate speech, with a yes or no option. Thus far, Facebook has not said anything about the rollout, and there is no indication of whether the checker will roll out to more users or when that may happen.
One interesting element of this rollout is how Facebook plans to manage it to ensure that its algorithms aren’t mistaught, purposefully or otherwise. The service has a very exact definition of what constitutes hate speech, and people who aren’t familiar with those guidelines may well mark speech that they don’t like as such. When a crowd is put in charge of teaching an AI with any degree of known control, the internet at large tends to make things go in unintended directions, as demonstrated by the rise and fall of Microsoft’s TAY AI. A similar situation could happen here, to an extent, as people could very well decide to get together and teach the hate speech checker to go after specific content that doesn’t necessarily qualify as hate speech. Facebook will likely be reviewing or vetting hate speech declarations in some way, such as having its team check reports of hate speech before feeding content to the AI, or only rolling the checker out to parties it considers trustworthy.
This new hate speech checker comes to light just a couple of days after Facebook put out an announcement wherein the company said it was working on making its AI more capable of detecting hate speech, among other unsavory content. This could mean that similar checkers could pop up in the near future to enlist users’ help in detecting other kinds of unsavory content. Hate speech has been one of the biggest thorns in Facebook’s side in the last couple of years, and the company has taken steps in the past to curb its presence on the service with limited success.