Wikimedia Product/Perspectives/Augmentation/Content Curation

Content Curation

Summary

Content curation is the aspect of wiki activity related to editing, refining, and cleaning up content that has been generated. The Wikimedia movement's ambitious aspiration to make the sum of all knowledge available to everyone in the world means that the movement has a tremendous amount of work to do with respect to making judgments about what information belongs, and how to organize, phrase, and cite it. Most of the hundreds of languages in the world have Wikipedias with less than 10% the number of articles that English Wikipedia has, and even the largest Wikipedias have serious gaps in terms of the depth of their articles, and the subject matter covered by their articles. As all that content gets added, the curation workload will increase beyond what humans are capable of doing.

Augmentation is a potential pathway to curating the massive amount of information needed in the Wikimedia projects.  By applying algorithms and artificial intelligence in the right ways, human editors can be aided in making the most important judgments about the content in the wikis, allowing the content to be well-organized and reliable.  This kind of human-machine partnership is not new in the wikis. Tools like Twinkle and Huggle have been helping to automate the tasks of reviewing recent changes and patrolling for vandalism since 2007 and 2008.  ClueBot has been independently reverting vandalism since 2011.  And in more recent history, ORES machine learning models have begun to surface the edits and pages most in need of attention.

As humans and machines work together to curate content, we can think about that interaction on a spectrum of how much work the human editor does and how much work the machine does.  In some scenarios, the machine may just direct human attention to important curation needs. In other scenarios, the human may review tasks completed (e.g. edits reverted) by an algorithm. This paper explores some specific examples of content curation activities that can exist in the future, drawing from all along the spectrum of the human-machine partnership.

Because bias and unfairness already exist in the contents of the Wikimedia projects, algorithms have the potential to magnify and exacerbate those problems.  The Wikimedia movement should confront this with the same principles that have led to our success in the past: transparency and the ability for anyone to contribute.

White Paper

DRAFT