Research:Gender asymmetry models in CSCW
On June 2nd 2013 at the 5th Wikipedia Academy conference led by WMF chapter member Bishakha Datta a local chapter member announced that the Hebrew Wikipedia has recently reduced its gender gap from to 25%-75%. Yet on August the 11th at the Wikimania 2013 conference members of the audience, shocked by thier departing CEO's thundering omission regarding "the gender gap", requested clarifications. Gardner retorted bitterly that no progress has been made and accused the community of inaction.
This disharmony is not a typical of the handling of the Gender Gap. From the start the research on Wikipedia's gender gap has been riddled with problematic assumptions. The initial UN survey had miscalculated the gender gap, and subsequent research perpetuated these mistakes as well as testing a number of colloquial statements made by Gardner in lieu of more solid results. As such this research has not been actionable. Recently the UN research has been challenged opening the door to a new age of investigation - one of challenging established assumptions prevalent in this domain. Outreach by community coordinators in GLAM type outreach aimed primarily at increasing diversity have failed to focus on goals of retention and sustainability that would allow organic growth through crowd sourcing.
The goal of the project is to better understand gender gaps in CSCW settings. The gender gap is attributed to a number of factors. However it is unclear to what degree each contributes to the over all problem in differnt settings. By modeling gender gap development in differnt CSCW project, we consider how to chek and reverse the this check how differnt hope to find an over time we hope to check how creation and their growth in different CSCW stetting. By considering More specifically to investigate some of the underling causes for the gender gap on Wikipedia and other Collaborative on-line projects. To this end a family of behavioural models fomenting gender asymmetry have been formulated.
The research will aim to test:
Benefits of diversity
H1. Is diversity contribute to reduction in aggression?
H2. Does greater editor diversity correlate with reduced article bias embodied by sentiment and gendered language?
H3. Are the most active 1500 users aware of the advantages diversity, and the risks embodied in systemic bias?
Signalling and aggressive cascades
H4. Is aggressive behaviour by users in gate keepers roles a signalling device used for managing users in the newbie role?
H5. Is gendered language used by newbies as a triggering behaviour in newbie roles to users in gate keeping or curation roles?
H6. Does incidence of aggressive behaviour due to editing cascades correlate with cohort size?
Assume Good Faith
H7. Do users in gate keeper roles Assume good faith or Act Cynically' ?
H8. Does an asymmetric psychological make-up contribute a psychometric cause for the gender gap?
Benefits of feminism
H9. Does editor diversity correlate with more neutral content creation.
H10. Does an asymmetric hyperbolic preference a econometric component for the gender gap?
H11. Does an asymmetric social utility contribute a cognitive cause for the gender gap?
The different hypothesis require different research methods.
- Work on aggression has been studied since the beginning of CSCW research a prefer technique involves cataloguing power moves and deriving a geometric mode based on editing patterns.
- Work on bias in articles has originally used independent peer review more recently it has increasingly relied on sentiment analysis. Discriminatory language is a well studied domain in Linguistics. It's estimation requires the development of a novel technique .
- This is a well known group of editors. It will be split 4 ways and monitored for 3 months after work:
- an adaptive survey may be administered to a sample.
- a second group will be administered a modified Voight-Kampff type test instead
- a third group will be getting both
- group four will be a control
- This can be tested by combining features of 1 and 3. However to determine social roles SNA will be required as well.
- Same as last.
- Editing cascades can be extracted by longitudinal analysis of an ego graph of newbie editors. Cohorts size will depend on the number of stake holders in an article and on their social influence.
- AGF has also been studied by CSCW researchers as well as cynical behaviour. This information is only partly available in the logs and requires researcher type access. If this is not available it will be simulated based on available parameters.
- Psychologist have developed a 20 question mechanism for estimating this.
- Sentiment analysis and user edit count.
- This typically requires a 480 question computer administered test. I plan to create a 20-40 question adaptive survey.
- This requires a brief survey on social networks sizes. Snowball sampling will be used to collect information on non Wikipedian
The surveys, particularly the adaptive ones will require a small group for calibration.
Editors will also be asked to participate in surveys and be able to view the aggregate results on-line. As mentioned the project includes SNA methodology, by using snowball sampling we plan to reach a sufficiently large group of users and educate them about the gender gap.
Wikimedia Policies, Ethics, and Human Subjects Protection
Benefits for the Wikimedia community
Send me mail using my user page