Critical issues presentations/Considering anons as a protected class: The hardships and value of IP editors
- Submission no. 86
- Title of the submission
Considering anons as a protected class: The hardships and value of IP editors
- Author of the submission
- Aaron Halfaker
- Country of origin
United States of America
Governance, Policy, Research, Technical
- Anonymous editors
- Social patterns
- Machine learning
- Article creation
- Quantitative methods
Your target – for whom is your presentation meant?
Technologists, Wikipedians, Researchers or anyone building Theory about how our projects operate. This presentation will involve discussion of research results that have implications for us all.
Your topic – which issue would you like to bring up?
See abstract. In a nutshell, I want to address the question "How much value does anonymous editing bring to our projects?" with hard data and propose a change in how we think of anons.
Your purpose – what is the result that you think should bring your presentation/proposal?
I want our projects to work better. Even if my proposals fall flat, I think it is important that people know what's up with anons so that they can think better about other proposals.
Your approach – in what ways is your presentation tackling the theme of Wikipedia as a driver for change?
Wikipedia is already a very unusual case of open knowledge production in that people who lack of a social identity (anons) can participate in many important ways. However, these contributors have been relegated to secondary consideration. I think we can do better. Awareness and buy in is the first step to making changes. That's the goal of this presentation. The next steps are experimentation and integration. As an example, I'll be talking about what experimentation I'll be working on next to minimize negative impact on good-faith anons.
Historically, IP editors (anons) have been approached with suspicion, in the name of quality and prevented from contributing at the same level as registered editors in many Wikimedia projects. It turns out that the intuitions that lead us to be suspicious of anons are often quantifiably wrong. I'll summarize research that describes how and where anonymous editors contribute to our projects with a focus on the top Wikipedia's by article count (English, German, French, Spanish, Italian, Japanese, etc.). Using this research (most of which is my own work), I'll make a case for considering anons as a protected class drawing from the distinction made in US law. Specifically, this research shows that:
- anons create high quality new articles when they are able
- anons add ~20% of the productive new content in Wikipedia
- when we make it more difficult for anons to edit in an experimental context, we see substantial productivity drops vs control
- anons are unfairly profiled by the machine learning models used to make quality control efficient
Finally, I'll conclude with proposals for what we can do now to open our projects more effectively. E.g. open up article creation on English Wikipedia to anons, allow anons to use gadgets/beta features, find other sources of signal for vandalism detecting models, and start asking the question, "How should good-faith contributors who would rather not maintain a social identity be able to participate?".