It is common for digital platforms to issue consequences for behaviors that violate Community Standards policies. However, there is limited evidence about the relative effectiveness of consequences, particularly lengths of temporary suspensions. This paper analyzes two massive field experiments (N1 = 511,304; N2 = 262,745) on Roblox that measure the impact of suspension duration on safety- and engagement-related outcomes. The experiments show that longer suspensions are more effective than shorter ones at reducing reoffense rate, the number of consequences, and the number of user reports. Further, they suggest that the effect of longer suspensions on reoffense rate wanes over time, but persists for at least 3 weeks. Finally, they demonstrate that longer suspensions are more effective for first-time violating users. These results have significant implications for theory around digitally-enforced punishments, understanding recidivism online, and the practical implementation of product changes and policy development around consequences.
https://dl.acm.org/doi/10.1145/3706598.3713163
The ACM CHI Conference on Human Factors in Computing Systems (https://chi2025.acm.org/)