Don’t you trust me?

From Jill Walker's blog I came across a really interesting project which is trying to visualise "trust" in wikipedia. Basically, they scraped some of Wikipedia and run a program which coloured in authors' contributions based on how long their edits remained untouched. Contributors whose edits lasted a long time were uncoloured while those whose edits tended to last the least length of time had an orange background. They refer to this as a "content driven" trust system rather than trust systems such as ebay's where the rating is explicitly produced by users. The idea is to build a system that can automatically help a reader discern how reliable the article in question might be. For example, "untrusted" authors on "A Million Penguins" might have their contributions rendered in a banana colour: the brighter the banana, the less you can trust them…
It's fascinating to look at. Jill Walker makes the point that there are times when she might be more interested in looking at the "dissent" which brings up the notion of dimensionality and trust within a context. For example, someone might have had a lot of changes made to their work in one particularly contentious article while other edits might rarely have been touched. The visualization doesn't seem, at the moment to deal with that context.
The various attempts to help readers interpret the provenance of articles on wikipedia (and by extension similar texts) could possibly do with thinking about the transliterate skills of readers. Our notions of communicational trust are remarkably complex. Although the colouring project is interesting it really is just a dabble of the toe in some fascinating waters.


One thought on “Don’t you trust me?

  1. This sounds a bit like what Alan Liu shared with us during his presentation at the IoCT in July. He spoke about colouring the pages (I think with shades of red) to illustrate the “quality” of each page.
    I’ll be interested to see how they derive content-driven reputation. The project summary explains that content that gets built upon = reputable author…I’m not sure if it can work like that. What if all the comments are negative or spam? (but they’ll probably have some crafty algorithms!)

Comments are closed.