Linguistic Similarity Within Centralized FLOSS Development

A new study from researchers Matthew Gaughan, Aaron Shaw, and Darren Gergle challenges conventional wisdom about how centralized development affects open source communities. Published as "Linguistic Similarity Within Centralized FLOSS Development" and accepted to CHI Extended Abstracts 2026, the research examined three Wikimedia platform features built in MediaWiki using repository mining, linguistic style analysis, and principal component analysis. The team tracked feature development and issue discussions across multiple cases to understand how communication patterns evolve when project stewards centralize development.

Contrary to both intuition and prior research, the study found no identifiable differences in linguistic style between Wikimedia Foundation (WMF) affiliates and external contributors, even when feature development was guided by WMF contributions. This surprising result led to two key provocations: first, that stewards dominate development based on their own use of specific functionality rather than through communication hierarchies, and second, that centralized project development doesn't necessarily create hierarchical language within project discussions. The findings suggest that concerns about centralized control creating communication barriers in open source projects may be overstated, at least in the Wikimedia ecosystem where the research was conducted.