Go Going Fart rther r Together: r: The The I Impact mpact o of So Social C cial Capit apital o al on n Sus Sustaine ned P d Partici cipa pation n in Open S in Open Sour urce ce Huilian Sophie Qiu , Alexander Nolte, Anita Brown, Alexander Serebrenik, and Bogdan Vasilescu @sophiehsqq @alexander_nolte @aserebrenik @b_vasilescu
Ma Maintain, or or su suffer https://w3techs.com/technologies/ history_overview/web_server (Greenstein and Nagel, 2016) 1
Op Open-so source exp xperience boosts s resu sumes Employers (and job seekers) use open-source experience to make inferences (or form (Marlow et al., 2013) impressions) about a candidate’s technical skills. https://codeformore.com/how-to-write-up-open-source-experience-when-you-dont-have-any/ 2
Kn Knowledge gap ? long-term developer outsider newcomer developer motivation retention onboarding This work 3
RESULT T HIGHLIGHTS: Social capital ex explains susta tained participation Being part of teams with more diverse Higher social capital information, especially for women 4
~40% 40% of of G GitHub c con ontri ributor ors d disengage 1.00 + + + + + + + + + + + + + + + + + + + + Survival probability 0.75 0.50 0.25 0.00 0 12 24 36 48 60 Time in months 5
Soci Social c capital t theor ory f for s or sustainable p part rtici cipation on Bonding social capital : Bridging social capital: benefiting from strongly connected network benefiting from network with diverse info Willingness to continue Opportunity to continue (Coleman, 1990) (Burt, 1998, 2001) 6
H1: more soci cial capital ~ ~ mo more e prolonged ed en engagemen ement Bonding social capital : Bridging social capital: benefiting from strongly connected network benefiting from network with diverse info Willingness to continue Opportunity to continue (Coleman, 1990) (Burt, 1998, 2001) 7
Cohesive networks might foster discr crimination / excl clusion 8
On On GitHub, , women dise sengage earlier than men After one year ca. 70% of men are still active but only ca. 60% of women 9
On On GitHub, , women dise sengage earlier than men After one year ca. 70% of men are still active but only ca. 60% of women Why do contributors disengage? • Developer survey in our paper: • Work-related (e.g., new job) • Personal* (e.g., different hobby) * women cite more often than men See also: • Miller et al, OSS 2019 • Iaffaldano et al, SoHeal 2019 10
H2: H2: Tea eams ms with mo more e diver erse e informa mation ~ mo more e pr prolong nged d eng ngagement, t, esp. p. for r women Information diversity should reduce the risk of demographic- based echo chambers. 11
Large-sc La scale mixed-me method ods s study Survey Small sample 88 replies 1,000 users Small sample Survey 1,000 users 88 replies 12
In Infer erring ing gender ender from nam names es Public name lists & celebrity names, including 3,000 East Asian names https://github.com/tue- gender mdse/genderComputer Computer https://www.namsor.com Naive Bayes Binary gender classifier prediction name features, e.g., the last two characters 13
Bon Bonding soc ocial capital – Te Team Familiarity TIME Project A Project C Project B Project X End of data (de Vaan et al., 2011, Lutter 2015) 14
Bon Bonding soc ocial capital – Re Recurring Cohesion TIME Project C Project D Project X End of data (de Vaan et al., 2011, Lutter 2015) 15
Bridging soc Bri ocial capital – La Language D Diversity TIME Java C C++ Project X C C++ End of data (de Vaan et al., 2011, Lutter 2015) 16
Bri Bridging soc ocial capital – Sh Share of of Ne Newcome omers TIME Project X Project X Project X End of data (de Vaan et al., 2011, Lutter 2015) 17
COX regression model CO Contributor Time Active Social capital Control variables 2008 Jan – Mar True Project size Team familiarity Recurring cohesion Project owner Language diversity …… Share of newcomers active, major contributor, proj owner, social capital 2008 Jan – Mar True Project size Team familiarity Recurring cohesion Project owner Language diversity active, major contributor, not owner, social capital …… Share of newcomers 2009 Apr – Jun False Project Size Team familiarity active, minor contributor, not owner, social capital Recurring cohesion Not project owner Language diversity …… left, minor contributor, not owner, social capital Share of newcomers 18
H1: more soci cial capital ~ ~ mo more e prolonged ed en engagemen ement + + 1.00 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + 0.75 Survival probability 0.50 Recurring cohesion + High 0.25 + Low 0.00 0 12 24 36 48 60 Time in months 19
H2: Language diversity interact cts with gender Survival difference between contributors with high and low language diversity 0.08 Difference in survival probability 0.06 0.04 Gender Female 0.02 Male 0.00 0 12 24 36 48 60 Time in months 20
Wha What ne next? Recommend projects that Offer mentorship Use badges to show a can help build social capital community’s culture 21
Soci Social c capital e explains p prol olon onged e engageme ment Acknowledgements: Contact: Sophie Qiu @sophiehsqq hsqq@cmu.edu @alexander_nolte alexander.nolte@udo.edu @aserebrenik a.serebrenik@tue.nl @b_vasilescu vasilescu@cmu.edu Code and data: 22
Recommend
More recommend
