Bots Don't Mind Waiting, Do They? Comparing the Interaction With Automatically and Manually Created Pull Requests

TL;DR

This study compares how open source project maintainers interact with human versus bot-created pull requests, revealing that bot PRs are less likely to be accepted and take longer to merge, indicating room for improvement.

Contribution

It provides empirical evidence on the differences in maintainer interactions with bot and human pull requests, highlighting challenges faced by bots in open source workflows.

Findings

Bot PRs are accepted less frequently than human PRs.

Bot PRs take longer to be merged despite fewer changes.

Approximately one-third of all PRs are from bots.

Abstract

As a maintainer of an open source software project, you are usually happy about contributions in the form of pull requests that bring the project a step forward. Past studies have shown that when reviewing a pull request, not only its content is taken into account, but also, for example, the social characteristics of the contributor. Whether a contribution is accepted and how long this takes therefore depends not only on the content of the contribution. What we only have indications for so far, however, is that pull requests from bots may be prioritized lower, even if the bots are explicitly deployed by the development team and are considered useful. One goal of the bot research and development community is to design helpful bots to effectively support software development in a variety of ways. To get closer to this goal, in this GitHub mining study, we examine the measurable differences in how maintainers interact with manually created pull requests from humans compared to those created automatically by bots. About one third of all pull requests on GitHub currently come from bots. While pull requests from humans are accepted and merged in 72.53% of all cases, this applies to only 37.38% of bot pull requests. Furthermore, it takes significantly longer for a bot pull request to be interacted with and for it to be merged, even though they contain fewer changes on average than human pull requests. These results suggest that bots have yet to realize their full potential.