As research for a potential presentation topic at EMWCon 2020 [0], I'm
polling the community.
What algorithms do you program into your wikis? I'm curious to learn about
technical methods, policies, and criteria for measuring "success". I'm
interested in this concept because of how algorithms are used in social
media tools and many have been called out for being "bad". So I wonder if
wikis are any better. And don't just limit the analogy to social media.
There's things like Amazon's suggested purchases or the stack exchange
voting system selecting a "correct" answer by election. For example, the
extension Watch Analytics recommends which pages need review based on how
many people have reviewed each changed page. I think there are
voting/rating systems for wikis. I wonder how those influence decisions. At
one time I was working on an extension that built on the idea of
contribution score, but I feared it would incentivize the "wrong"
behavior.
I'm looking for how algorithms influence content presented to users, but
also if algorithms are used to make decisions.
Here are some other examples of what I would consider algorithms in wikis:
- Human driven voting systems like stack overflow
- Category and Property driven - Using Wikibase, Cargo, or SMW to
determine a priority or to identify a missing value
- Machine-learning - Antivandalism/spam tools
- Suggestions for which wiki pages to watch based on links between wiki
pages
- Suggestions for which wiki pages to review based on watchlists and how
many people have seen the page since it was last edited
- Suggestions for which users to collaborate with based on pages each
person has edited
- Contribution scores drive users to make more edits and to edit more
pages. Is this what we really want?
- Other?
Of the algorithms you use in your wikis, how do you measure success? What
does it mean to say that your algorithm is "working" or "succeeding"?
How do you know your algorithm is not inducing bias in a bad way?
If you'd like to share any info on how you use algorithms and how you
measure success in using them, please reply directly to me (unless you want
to share with the list). I'd appreciate screenshots and other ways to help
me understand how your algorithm works and is used, so I can properly
represent it in my presentation. I hope to get enough responses to showcase
the flexibility of MediaWiki in how it can be used to drive decisions based
on knowledge.
Daren
[0]
https://www.mediawiki.org/wiki/EMWCon_Spring_2020
__________________
https://www.mediawiki.org/wiki/User:Darenwelsh
http://mixcloud.com/darenwelsh