Hi Richard,
This is a great idea, and is something I would like us to push towards in Structured Commons: easy crowd-focused ways for us to refine understanding of what is involved in a multimedia file. However, all of the precedents we have use either machine prediction, or human curated prediction (categories/infoboxes,etc) of whether or not the content is in there. Also, the further we spread these games, the more we need to think about confirmation of the "new" data that we are generating: for example, most of the crowd-sourcing strategies on Zooniverse, require a certain number of folks to affirm a decision, before they are added to a likely positive queue (
https://www.zooniverse.org/). We may want to start with items that have "depicts" compatible categories on Commons that could be sorted into more specific categories -- or where the institutional descriptions have string matches.
As Magnus describes also: I don't think the structure or data model right now is very crowd-sourcing friendly at the moment: asking folks to identify what is "depicted" in an image, may solicit a lot of less than-accurate descriptions (for example, if you aren't familiar with farm equipment in 19th century france, you are likely to not know what the right language is for describing those pastoral settings, etc). The more we push this beyond folks who are wiki-savvy, the more we loose the self-filtering of folks interested in understanding our data and community model before contributing (I am also not familiar with any documentation of institution-friendly ways for monitoring batches of data that get worked on by these institutions per #2 at
http://blog.hatnote.com/post/161358369742/wikicite-2017-and-the-7-features-wikidata-needs).
Thats not to say we shouldn't work on this: I just think some on-wiki consultation and design (per Magnus) would be rather important.We should also take a look at the interface/strategy used by GeneWiki:
https://github.com/SuLab/genewiki . They have the best other-environment, open ecosystem for contributing to a particular domain on Wikidata.
Cheers,
Alex