Help teach AI the difference between hate and pride

When LGBTQ+ people use our own words — reclaiming language that was once used against us — AI moderation systems often treat this as hate speech. The same word that a community member uses to claim their identity, crack a joke, or push back against homophobia gets flagged identically to a genuine attack. We are building a dataset to teach the difference, and only community members can provide the nuance that this requires.

The task involves reading social media posts that contain LGBTQ+ slurs in a range of contexts — including hostile uses — and labelling the pragmatic intent. The cases are often subtle: not just obvious pride declarations, but the full range of how community members actually talk.

⚠ You will be exposed to explicitly homophobic and offensive language, including slurs used in hostile contexts. You can skip or stop at any time without penalty.

Estimated time per session: 20–40 minutes. You can annotate across multiple sessions.

We do not collect your name, email, or any identifying information. You will receive an anonymous ID that is yours to keep.