Batching of Tasks by Users of Pseudonymous Forums: Anonymity Compromise and Protection

Abstract

In a number of applications where anonymity is critical, users act under pseudonyms to preserve their privacy. For instance, in scientific peer review using forums like OpenReview.net, reviewers make comments on papers that are publicly viewable. Reviewers who have been assigned multiple papers operate under different pseudonyms across their papers to remain anonymous. Other examples of publicly visible tasks where users operate under pseudonyms include Wikipedia editing and cryptocurrency transactions. In these settings, it is common for users to engage in batching - the completion of several similar tasks at the same time. Batching occurs both due to natural bursts in activity (e.g., a person visits a website and makes many comments at once) or as a productivity strategy used to streamline work. In peer-review forums such as computer science conferences, reviewers and meta-reviewers are often assigned multiple papers. We find empirically that reviewers are highly likely to batch their comments and/or reviews across papers. In analysis of data from a top Computer Science conference with thousands of papers, reviewers, and discussion comments we find that when reviewers and meta-reviewers comment on multiple papers, they have a 30.10% chance of batching their comments within 5 minutes of one other. In comparison, any randomly chosen pair of reviewers and meta- reviewers had only a 0.66% chance of making comments on different papers within 5 minutes of each other.