community and code: lessons from nescent hackathons
TRANSCRIPT
Community & CodeLessons from NESCent hackathons
Arlin Stoltzfus, Michael S. Rosenberg, Hilmar Lapp, Aidan Budd, Karen
Cranston, Enrico Pontelli, Shann Oliver, and Rutger A. Vos
https://nescent.github.io/community-and-code/
A lot of questionsWhat practices make hackathons effective or ineffective?• What makes a good scope or theme? • How best to advertise and recruit? • How to engage before the event? • What logistics are preferable? • What supportive technology to use?• What modes of team formation? • What methods of target selection?
What do people expect hackathons to deliver? Why do• sponsors underwrite hackathons?• organizers plan and execute hackathons?• participants participate?
What do hackathons do?
Why are hackathons a thing?
Approach
• Retrospective analysis of a series of 9 events• “Open Science” practices left a record of – 9 events with 17 event outcomes– 54 projects with 133 project outcomes– 207 participants
• Guiding questions– What were the outcomes and impacts? – What practices or conditions favored positive outcomes?
NESCent hackathons are distinctive• Open and community-oriented– Sponsors have broad community-oriented goals– Everything is open-source– Diversity is a priority
• Domain: evolutionary informatics– a dispersed community of thousands worldwide
• Participants (25 to 30)– post-docs, faculty, grads (mostly PhDs)– offered full travel support– most apply in response to open call
ScheduleDay 1 Days 2 to 4 Day 5
Work
LunchWrap-up
Work
Lunch
WorkStand-ups
Why the events happened
Why did NESCent sponsor hackathons?• to address interop issues involving collaboration or
building community resources• to foster and energize a community of practice
– Spread awareness of best practices– Grow a professional network– Catalyze awareness of domain challenges & opportunities
Outcomes for
• 9 events – 17 event-level products
• 54 projects (teams) – 133 project products (mostly repos and team reports)
The most common products• new code repositories – most remain inactive
• additions to existing codebases– E.g., open-source toolboxes (BioJava, BioPerl, etc)
atypical: group keeps working, writes paper
spike of 238 commits peaking at hackathon
Less common products• documentation
– CRAN task view– Report on file format and parser compliance
• designs, standards and schemata– Phylorefs– Skelesim
• data products – Machine-readable tree annotations
• community infrastructure– r-sig-phylo mailing list
Details: see speaker notes
Follow-on products
• demonstrations and production code– DendroPy
• Blogs, conference presentations• manuscripts for publication
– 2 event publications – rotl– skelesim
• proposals for funding– GSOC proposals– phyloreferences– Phylotastic
Link is not via actual code, but schemes, community buy-in and working relationships
Project builds on code
Note
• End of stuff we counted objectively• Everything from here is subjective
Intangible outcomes
• Technology learning• Exposure to best practices• Awareness of challenges and
opportunities• Team programming experience
Credit: Randall Munroe https://xkcd.com/1425/
Lessons: scope a theme "must have the capacity to inspire participation by being specific enough to indicate the direction, while possessing sufficient openness to allow for the imagination of the group to take over” (OpenSpace philosophy)
Lessons: assistive technology
• At least– source code repository– event-wide communication channel
• provide training before or on day 1
Lessons: diversity
• open call response rate: ~ 1/100• personal appeal response rate: ~ 1/2
earlier events
later
Details: see speaker notes
2- to 3-fold increased diversity
Lessons: being welcoming
• reach out to non-networked participants– “Thanks again for applying to the hackathon. <sentence that shows I’m a human who read your application>. I look forward to meeting you.”
• do pre-event engagement • model positive communication– e.g., not "Isn’t that idea out of scope?" but
"What are some ways that could align with our goals?"
Lesson: be wary of remote participation
Works better when• Remote participant – is in similar time zone– Is already networked, experienced– commits 100 % of attention to event
• On-site participants– stick to schedule– Establish 2-way channel (buddy system)
Lesson: manage team formation
• Things work better when– Facilitators model asking questions to ensure plan is
• Relevant (in scope)• Feasible given skills of participants• Aimed at getting tangible outcomes by the end of the event
– A team is 3 to 7 people – A project requires collaboration for success
Explore & sift
Inform
pitch Team up
WorkCarefully managed, facilitated steps
Lessons: pre-event engagement
• introduction in online space or telecon• chance to ask questions and discuss ideas• chance for organizers to gauge– how well theme stimulates ideas– where participants need training or info
• don’t expect full participation
preparation for "invested participation" (Briscoe & Mulligan)
Concluding thoughts
Why participate? • learning new skills• networking• awareness• experience valued for its own sake
Details: see speaker notes
Resourceshttps://nescent.github.io/community-and-code/
Acknowledgements
One of 2 working groups
Some of the 207 NESCent hackathon participants