The freedoms that exist on the Internet often pave the way to convert certain spaces into hostile environments. In forums and social networks it is normal to find trolls, haters and other figures that, unpleasantly and unfortunately, are part of the collective identity of the network.
Within these digital spaces, free software bases its existence on the collective support of its community. A study deepened how “toxic” the environment can be in these instances.
Study analyzed the levels of obnoxiousness in the free software community
Assuming the aforementioned hateful characters as part of the universe of network users, with variants that appear depending on the space in which they operate, a joint work developed by scientists from Carnegie Mellon University and Wesleyan University, advanced in the understanding of toxicity on platforms where the open source community interacts, such as GitHub.
To analyze what toxicity looks like in the open source community, the research team first collected content listed as toxic. To do so, they used a toxicity and politeness detector, developed for another platform, to scan nearly 28 million posts on GitHub that were made between March and May 2020. The team also searched for “codes of conduct,” posts that they are interpreted as a possible signal of reaction to toxic content; as well as blocked or removed issues, which can also be a sign of toxicity.
Following the information curation process, the team developed a final dataset of 100 toxic posts. They then used this data to analyze the nature of the toxicity, which can be categorized under the labels of insulting content, authoritarian, arrogant, trolling or unprofessional. Under the same criteria, it was analyzed whether the hateful content was directed at the code itself, at the people involved, or at some other destination.
“You have to know what that toxicity looks like to design tools to handle it”said Courtney Miller, one of the authors of the article documenting this research. “And managing that toxicity can lead to places that are healthier, more inclusive, more diverse, and just better overall.”he added.
Considering the particularities of these spaces, the research was able to define how hatefulness is manifested. “Toxicity is different in open source communities”Miller said. “It’s more contextual, titled, subtle and passive-aggressive,” pointed out regarding the characterization built from the study.
On the collected data, some interesting observations stand out. About half of the toxic posts the team identified contained obscenity. Others were from demanding users of the software. Some posts come from users who post a lot of issues on GitHub, but contribute little else. Also, there are cases in which comments that started on the code of a software, became personal. In conclusion, none of the archived posts within this sample helped improve open source software or the community.
The team noticed a unique trend in how people responded to toxicity on open source platforms. Often the project developer went to great lengths to accommodate the user or fix the issues raised in the toxic content. This usually resulted in frustration.
“They wanted to give the benefit of the doubt and create a solution,” Miller said. “But this turned out to be quite tiring”.
“We have been hearing from developers and community members for a long time about the unfortunate and almost ingrained toxicity of open source”Miller said, alluding to the good response to this research within the community. “Open source communities are a bit crude. They often have terrible diversity and retention, and it’s important that we start to address and deal with the toxicity there to make it a better, more inclusive place.”he added regarding this “social x-ray”.
This work is at an early stage. According to Miller, the team hopes to lay the groundwork through this research for future development of a toxicity detector for the open source community.
