A new Roblox study shows how longer suspensions help curb bad behavior on the platform
Misbehavior on digital platforms can be tricky to manage. Issue warnings, and you risk not deterring bad behavior. Block too readily, and you might drive away your user base and open yourself to accusations of censorship. But a new study, presented at the Conference on Human Factors in Computing Systems, suggests a more effective path forward.
Researchers at Northeastern University and Roblox conducted two large-scale field experiments involving more than 770,000 Roblox users to study how suspension duration affects user behavior. The first experiment compared one-hour versus one-day suspensions for users with a single recent policy violation. The second experiment compared one-day to three-day suspensions for users with a second recent violation. They tracked outcomes such as likelihood of reoffending, number of subsequent violations, user reports against offenders, and engagement metrics like days active and total time spent on the platform.
“It’s very common for platforms to issue consequences for violations of the community standards, but there’s actually not a ton of causal evidence in the research about how effective different kinds of consequences are,” says Jeffrey Gleason, one of the study’s authors and a researcher at Northeastern University and Roblox.
The study found that longer suspensions significantly reduced reoffense rates, the number of consequences, and user reports. Longer suspensions also appeared to make users think twice before misbehaving again, increasing the time it took for them to reoffend.
The longer the suspension, the greater its impact on user behavior. A one-day suspension reduced reoffense rates by 6.7% compared to a one-hour suspension, while a three-day suspension reduced it by 8.1% compared to a one-day suspension. The deterrent effect lasted for at least three weeks, though its impact diminished over time, suggesting that some users eventually reverted to old behaviors.
One promising approach is to address bad behavior early. Harsh penalties for first-time violations were more effective at preventing repeat offenses. A one-day suspension lowered reoffense rates by 12.6% for first-time offenders, compared to just 4.4% for frequent violators. And despite concerns that suspensions might drive users away, the study found that users generally remained active on the platform after being temporarily banned.
“There’s always been this long debate around what is the trade-off between safety and engagement, and the fact that we’re seeing that you can do really important safety-related work while not sacrificing engagement is maybe not surprising for us, but may be surprising for other platforms,” says Alex Leavitt, principal researcher for trust and safety at Roblox and coauthor of the paper.
The experiments only tested suspensions of up to three days, and the effects of longer suspensions remain unknown. It’s also uncertain how these findings would translate to other platforms. Still, the research offers valuable insights into curbing bad behavior online.
“We now have more public evidence around that fact,” says Leavitt, “so we can lean into that.”