Human-Machine Collaboration for Content Regulation: The Case of Reddit Automoderator

By Shagun Jhaver, Iris Birman, Eric Gilbert, Amy S. Bruckman. ACM Trans. Comput. Hum. Interact., 2019

335 citations recorded by the source corpus.

What one may say on the internet is increasingly controlled by a mix of automated programs, and decisions made by paid and volunteer human moderators. On the popular social media site Reddit, moderators heavily rely on a configurable, automated program called “Automoderator” (or “Automod”). How do moderators use Automod? What advantages and challenges does the use of Automod present? We participated as Reddit moderators for over a year, and conducted interviews with 16 moderators to understand the use of Automod in the context of the sociotechnical system of Reddit. Our findings suggest a need for audit tools to help tune the performance of automated mechanisms, a repository for sharing tools, and improving the division of labor between human and machine decision making. We offer insights that are relevant to multiple stakeholders—creators of platforms, designers of automated regulation systems, scholars of platform governance, and content moderators.

Subfields: Artificial Intelligence, Communication, Gender Studies

View Paper: https://www.semanticscholar.org/paper/a3f0b0bcbcdc9364c55aad6cf8ebd133027c3a40

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Communication

In Gender Studies

Citations in the corpus

Consistent Estimators for Learning to Defer to an Expert (2020, methodology): “A current application of the latter setting is Facebook’s and other online platforms content moderation approach (Vincent, 2019; Jhaver et al., 2019): an algorithm is used to filter easily detectible inappropriate content and the rest of the examples are screened by a team of human moderators.”
Fast, Accurate, and Healthier: Interactive Blurring Helps Moderators Reduce Exposure to Harmful Content (2020, background): “In more scholarly work, this subject has been attracting increasing attention (Dang, Riedl, and Lease 2018; Dosono and Semaan 2019; Jhaver et al. 2019; Karunakaran and Ramakrishan 2019; Roberts 2019; Wohn 2019).”
On the Challenges of Using Black-Box APIs for Toxicity Evaluation in Research (2023, positive): “Automatic toxicity evaluation not only stabilizes processes but also adds consistency in decisions (Jhaver et al., 2019).”