Crowdsourcing moderation without sacrificing quality

Abstract

Crowdsourcing moderation of online discussions can improve quality without sacrificing efficiency. A system of virtual moderation is proposed, where a trusted moderator provides a limited amount of input to train a machine learning model that predicts the moderator’s judgments about comments. Users can also provide judgments, which are used to improve the model’s predictions. This system allows the moderator to focus on high-level decisions while the model handles the majority of the work, potentially saving time and improving the overall quality of the discussion. In the long run, personalized views of the discussion can be generated, tailored to each user’s preferences, further enhancing the user experience. – AI-generated abstract.