Program/Using artificial intelligence to keep Wikipedia open
Video recording
- Notes
- Format
- Presentation
- Keywords
- Artificial intelligence
- Knowledge equity
- Machine learning
- ORES
- New page patrol
- ACTRIAL
- When and where
- Friday 20 July, 16:00
- Duration: 55 minutes
Hong Kong
-->
- Presenter(s)
- Topic area
- Relationship to theme
Community health
Abstract
Recently, new page patrollers (NPP) in English Wikipedians have decided to make a tradeoff between openness and the overbearing work of reviewing new article drafts. In an effort to reduce their workload, they've prevented new editors from being able to create new articles directly. While this is a challenge for the values of "the open encyclopedia that anyone can edit", it's hard to find fault in their decision. A backlog that only grows is frustrating and exhausting.
Backlogs aren't inevitable and there are many shapes that backlogs can take. In my presentation, I'll tell a story about where English Wikipedia editors defined a process and set of roles that formed a backlog around new page creations. I'll make the argument that this formalization of quality control practices has created a choke point and I'll show that alternatives exist. I'll draw a contrast between counter-vandalism and NPP that highlights some missing infrastructure that makes NPP so difficult. Finally I'll present a vision for an alternative patrolling strategy that is enabled by quality and topic-routing prediction models that we have developed for ORES, the open machine prediction service for our wikis.
These machine prediction models have great potential to (1) help patrollers delete the worst of the worst more quickly, (2) to reduce the difficulty of judgement calls with regards to the notability of topics, and (3) to route new editors toward experienced editors with similar subject interest -- potentially improving their newcomer experience in Wikipedia. This presentation targets patrollers, tool developers, newcomer socializers, community organizers, and researchers. The implications of this presentation are targeted at English Wikipedia but they are not limited to English Wikipedia. ORES is multi-lingual service and an outcome of this discussion will be spec'ing out development on similar models for other wiki-communities.
Outcomes
Attendees will gain a systems-level understanding of some of the most difficult to scale problems that openly edited wikis will face. Attendees will see how basic machine predictors (a subset of AIs) can help them with their own problems of scale. This presentation will help drive the conversation about what we can do with NPP forward.
This presentation targets patrollers, tool developers, newcomer socializers, WilkiProject organizers, and academic researchers. The implications of this presentation are targeted at English Wikipedia but they are not limited to English Wikipedia. ORES is multi-lingual service and an outcome of this discussion will be spec'ing out develop on similar models for other wiki-communities.
Signup
- John Andersson (WMSE) (talk) 03:40, 15 June 2018 (UTC)
- Rehman 09:56, 24 June 2018 (UTC)
- Rossouw van Rooyen (talk) 17:06, 3 July 2018 (UTC)
- May Hachem93 (talk) 13:19, 20 July 2018 (UTC)
- <add your username here if you are interested in attending>