CONCLUSIONS
Threats to validity As any empirical study our work is
subject to a series of threats to validity, pertaining e.g., the
choice of gender identication approaches and datasets. We
also made a simplifying assumption of the gender binary.
In this paper we evaluate 16 approaches to automatic gen-
der identication. We apply them to information about
Stack Overflow contributors obtained from earlier surveys.
We conclude that while individual elements of the gen-
der identication technology such as image recognition and
name-based heuristics are readily available, the technology
still needs to mature. Existing tools can generate con
icting
results and dierent approaches perform dierently on the
datasets. A promising direction would involve deeper inte-
gration of data from multiple sources including social media
sites targeting software developers (e.g., GitHub) or general
audience (e.g., Google+, Facebook, and Twitter) [4].