Study: AI models fail to reproduce human judgements about rule violations
In an effort to improve fairness or reduce backlogs, machine-learning models are sometimes designed to mimic human decision making, such as deciding whether social media posts violate toxic content policies....
