Envel Le HirAnother history of Denelezh
This post is only a short summary and is part of a series, yet to be written. See also A short history of Denelezh and A short history of the Wikibase Community User Group.
Events
In 2014, Maximilian Klein built Wikidata Human Gender Indicators (WHGI), a tool that provided statistics about the gender gap in the content of Wikipedia.
In 2017, after a few weeks of work on my free time, I released a proof of concept, the first version of Denelezh, a tool that provides statistics about the gender gap in the content of Wikidata and Wikimedia projects. One of the innovations of my tool was the ability to combine several dimensions to gather specific statistics (for instance country of citizenship + occupation = French politicians), which required a completely different architecture from WHGI. Wikimedia Deutschland was aware of Denelezh, as they cited it at Wikimania 2017 and WikidataCon 2017.
In 2018, I discovered that Wikimedia Deutschland was building another tool on the same topic, WDCM Biases Dashboard, with a full team involved. At that time, I paused the development of my tool, published an overview of my plans, and contacted Wikimedia Deutschland, looking for information about their project, worried by the overlaps in the features of our tools.
At first, Wikimedia Deutschland blamed their external consultant, Goran S. Milovanović from DataKolektiv. They then provided inconsistent answers, for instance alternatively stating that their project was finished and then that it was still under active development. I took the time to explain several times to Wikimedia Deutschland that I wanted to apply for a Wikimedia grant for Denelezh and that I did not want our tools to unnecessarily overlap, pointing out that it was already the case. Lydia Pintscher, Product Manager for Wikidata, and Léa Lacroix, Project Manager Community Communication for Wikidata, both promised to come back to me. Despite several reminders, they never did.
After five months of confusion, considering that it was impossible as a volunteer to maintain my tool (building a proof of concept and maintaining it are two different things), that Wikimedia Deutschland was actively developing an equivalent of it and was refusing to communicate their intentions, and that it was too risky to devote time and to apply for a Wikimedia grant as features of the tools were overlapping, I decided to discontinue mine.
The community started to ask questions. DataKolektiv quickly replied, recognizing the overlaps between the tools, providing useful explanations and asking Wikimedia Deutschland for guidance. They then removed their message. Wikimedia Deutschland contradicted DataKolektiv, denying that the features of the tools overlap. Then, while Wikimedia Deutschland has refused to communicate for several months, they denied their responsibility on the outcome, expressed fake compassion, and trolled me, stating that one can apply for a grant from the Wikimedia Foundation if needed and that they can offer their help in that matter. I did not feed the troll. Because of the situation, the opportunity of an application for a Wikimedia grant was missed in November 2018.
In the meantime, Wikimédia France contacted me to help with Denelezh. Thanks to them, Denelezh was back online a few months later.
In 2019, Maximilian Klein and I decided to merge our tools (WHGI and Denelezh), sharing our efforts and relying on our experience in the domain. This led to the project Humaniki, with a grant application in February 2020. DataKolektiv endorsed the application, Wikimedia Deutschland did not.
About the tools
The overlaps
Wikimedia Deutschland states that “the two projects [Denelezh and WDCM Biases Dashboard] are quite different”. It is true that the tools rely on slightly different methodologies. However:
- Anthere, an experienced volunteer working on the gender gap, noticed the overlaps.
- DataKolektiv, the data scientist working on WDCM, recognized the overlaps.
- In June 2020, Martin Gerlach, Research Scientist at the Wikimedia Foundation, produced a state of the art on the metrics for quantifying the gender content gap in Wikimedia projects. He was unable to notice the difference of methodology between the tools.
In practice, even with different methodologies, the tools provide roughly the same kinds of statistics.
The features
Except for a small part, the tool made by DataKolektiv and Wikimedia Deutschland is only a subset of the features provided by Denelezh. Particularly, you can’t combine dimensions with WDCM (for instance, the analysis country of citizenship + occupation = French politicians is not possible).
Wikimedia Deutschland stated that their tool was based on a request made by a volunteer. This volunteer wrote to me that WDCM Biases Dashboard did not cover their needs.
Issues
While starting to develop a tool:
- DataKolektiv did not study the state of the art on the subject, as they should have done in their role of data scientist.
- Wikimedia Deutschland did not involve the community:
- the people likely to use the tool,
- the people experienced to build such a tool.
As a consequence, they took one year with a lot of resources to develop a tool that provides only few original and useful features and that does not cover the needs of the community.
When the community reached them:
- DataKolektiv and Wikimedia Deutschland made inconsistent declarations.
- Wikimedia Deutschland made promises they did not keep.
- Wikimedia Deutschland did not take responsibility for their decisions and actions.
- Wikimedia Deutschland did not provide any practical help, nor tried to fix the “mess” they created.
As a consequence, they durably undermined community work about the study of the gender gap in the content of Wikimedia projects.
