Loading...
Quantifying Non-Code Contributions
Key questions that will point the way for us
[WHY] Why quantifying non-code contributions? What are some real-world use cases or applications?
[WHO] Who will be the users of our algorithm? Can we involve some lead users in this project
(so
that we can easily collect feedback)?
[WHAT] What does the algorithm do? What’s its input and output? Is the output qualitative or quantitative?
[HOW] Given the WHAT of the algorithm, how do we implement it?
Goals:
Look at how
contributions
are graphed
Find textual description of the algorithm
Look at code
Node Types
figure out relationships
How do they add these nodes to the graphs
Does it make sense to generate generic version of this graph
Look deeply in github contributions:
https://github.com/sourcecred
Look at how this will be taken in different angles
Suggestion
think about ideal algorithm
what characteristics an ideal algorithm would have
transparency
fairness
look at top down and top up
look from perspective of ideal authors and look into details in the graph contributions
What would you like from the algorithm
Look at fundamental issues with algorithm
SourceCred Algorithm
Nothing written up yet about algorithm specifics:
https://github.com/sourcecred/research
Videos
https://www.youtube.com/watch?v=qEIHmKPMpYE
Open-source
So many notificatoins for open-source maintainers
Goal for Source cred
empower more people to help with project maintenance
how to incentivize
Cred
project-specific reputation
triaging issues
maintainng the build
writing doc
fixing bugs
adding features
more…
How they fairly give cred to those who make valuable contributions
leave out features that are easily gamible
# of commits
# of followers
They use PageRank
before: website is more important if other pages mentionit on it
Please turn on JavaScript to use Paper in all of its awesomeness. ^_^
Key questions that will point the way for us
SourceCred Algorithm