Testing the impact of filtering harmful posts: A randomized-controlled trial with 100.000 users on Nextdoor

We investigate the effectiveness of an intervention designed to filter offensive content on the local social media platform Nextdoor. The intervention, commonly used across large social platforms, was tested through a randomized controlled trial involving 100,000 people over nine months. We evaluated whether filtering content impacted fifteen different measures of platform behavior including views of offensive posts, visits to the platform, content consumption, and content production. This randomized-controlled experiment was an advanced followup-study of a Nextdoor-study (Katsaros et al., 2023) conducted a year before, showing mostly null effects of the intervention. In this study, the results showed a significant reduction in two target outcomes - views of offensive posts and comments - in the treatment group compared to the control group. However, as in the past Nextdoor-experiment, we observed no significant effects of the intervention on the thirteen measured outcomes between groups. This study provides empirical evidence on the impact of a prominent and commonly used intervention and highlights its limitation, demonstrating the complexity of effectively moderating online platforms.