Discover new research on online content moderation in a talk by Francisco Tomás-Valiente Jordá from ETH Zurich. He will present findings from a randomized field experiment with an Austrian online newspaper comparing pre- and post-moderation methods for managing harmful comments. The study shows that algorithmic pre-moderation significantly reduces toxic content, without lowering user engagement.
Abstract from the speaker:
Many online platforms attempt to prevent the publication of harmful content using pre-moderation, in which online contributions are screened – by algorithms, humans or a combination – before they are published. The (implicit) alternative is post-moderation, in which moderators police comments after publication, sometimes relying on user reports of harmful comments. Compared to post-moderation, pre-moderation has disadvantages: it can be costly, depending on the cost of the human (and/or algorithmic) moderators used, and it can inhibit on-platform interaction by delaying or even preventing the posting of comments, including harmless ones. But does it actually work to reduce the publication of harmful comments? In this study, we partner with an Austrian online newspaper to run a randomized field experiment testing the efficacy of algorithmically supported pre-moderation as compared to relying on post-moderation. The newspaper’s moderation team randomizes articles into two groups - one in which comments are passed through their custom moderation algorithm before publication, and one in which comments are (almost) all immediately published. Both types of articles have user-led comment reporting and post-moderation by moderators. We find that pre-moderation reduces the share of published comments that are toxic — both when looking at those comments that are initially published, and those that remain published after post-moderation. While some of the initially-published toxic comments are subsequently deleted in post-moderation, most are not. We thus find that post-moderation, both by users and by moderators, does not replace pre-moderation at removing toxic content. At the same time, contrary to our expectations, pre-moderation does not affect user engagement; both types of articles get similar numbers of comments and have similar numbers of unique users.
About the speaker:
Francisco Tomás-Valiente Jordá is a PhD student at the Public Policy Group and the Immigration Policy Lab. His doctoral research, funded by a Doc.CH grant, employs quasi-experimental and experimental designs, alongside NLP and survey methods, to study how parties use moral rhetoric to electorally compete against exclusionary parties like the far right. His interests lie at the intersection of political behaviour, party politics and political communication, with a focus on morality in politics, election campaigns, and attitudes towards minorities.
He previously worked as a pre-doctoral researcher at the Immigration Policy Lab, studying online discourse, as well as antisemitism in 19th Century Switzerland. Before that, he completed a Master's in Comparative and International Studies at ETH Zürich and University of Zürich, and a Bachelor's in Philosophy, Politics and Economics at King's College London. His Master's Thesis explored whether joining Cabinet led politicians to receive more online incivility, and whether such penalty was larger for women in politics.
Bring your own lunch bag! Light pastries and drinks will be available in case you forget to bring it.
—
The Data Science Brown Bag Series is an informal and interactive gathering where participants bring their own brown bag lunch and engage in discussions on research and insights the field of data and computational social science (light pastries and drinks will be available if you forget your lunch bag!).
The series provides a platform for data enthusiasts, researchers, and practitioners to share their experiences, best practices, and emerging methodologies and research in using data science to analyze and understand social and political phenomena. The brown bag talk series is for anyone interested in data science and social science to network, learn, and share ideas in a casual and friendly setting.
