Do Automated Legal Threats Reduce Freedom of Expression Online? Preliminary Results from a Natural Experiment

Automated law enforcement systems support privately-operated enforcement bots to take legal action in hundreds of millions of cases a year. In the area of copyright, legal scholars have hypothesized the existence of "chilling effects'' that harm public discourse by influencing people to self-censor protected speech. We test this hypothesis in a large-scale quasi-experiment with 9,818 accounts on Twitter that made 5,171,111 tweets. In a confirmatory interrupted time-series analysis, we find evidence that people reduce how much they post online after receiving a take-down notice from a copyright enforcement bot. On average, accounts sent fewer tweets after enforcement. Accounts also changed from a daily increase in public tweets to a decline on average. We also report on novel software that conducts third-party monitoring of the behavioral outcomes of automated law-enforcement systems. Since automated law enforcement can influence public discourse, third-party monitoring like this report will be essential to governing the power of enforcement algorithms in society.

We have published this preliminary analysis to invite feedback and suggestions for an upcoming follow-up study with a matched comparison group.
Hosted on the Open Science Framework