Mind Your Ps and Qs: Supporting Positive Reinforcement in Moderation Through a Positive Queue
By: Charlotte Lambert , Agam Goyal , Eunice Mok and more
Potential Business Impact:
Rewards good posts to make online groups better.
Online communities are constantly growing, with dozens of platforms housing millions of users. Large and small communities alike rely on volunteer moderators to maintain order. Despite their key role, moderators are given a toolbox of punishments and asked to fend off barrages of harmful content. However, prior research shows that positive feedback may proactively encourage higher quality contributions and discourage norm violations. Moreover, moderators themselves have requested support for locating and rewarding content to encourage in their communities. These requests notwithstanding, there is a tangible lack of practical support through tools. Building off moderators' ideas, we build a novel moderation system, the Positive Queue, that augments Reddit's existing moderator interface with features to discover and reward desirable content. Through a user study of moderators, we find that the system has value to vastly different moderation settings. We present design directions and insights for incorporating positive moderation strategies into existing spaces.
Similar Papers
Scaling Reinforcement Learning for Content Moderation with Large Language Models
Artificial Intelligence
Teaches computers to remove bad online posts faster.
Active Query Selection for Crowd-Based Reinforcement Learning
Machine Learning (CS)
Teaches robots to learn faster from people.
Q-Ponder: A Unified Training Pipeline for Reasoning-based Visual Quality Assessment
CV and Pattern Recognition
Helps computers judge picture quality better.