Research talk:Directed diabetes info-seeking behavior in Wikipedia

Questions

How are you going to ensure that a reader who gets the question is not a registered user? Is it possible to only show it to IP users?--Ymblanter (talk) 12:10, 2 August 2013 (UTC)

Readers will be recruited for the study ahead of time, and will be asked to read while on a specific account. Rather than have them install our user gadget (and potentially create an account if they're an unregistered reader), we'll provide an account that's pre-configured with the gadget. They just log in on the given account, go to any article, and the gadget pops up to guide them through the study. When they've gotten all the way through the study, the gadget will instruct them to log out. If they fail to log out for some reason, we'll both be disabling the collection server (for IRB/data privacy purposes) and changing passwords on the account, to make sure they can't keep using the gadget after the study. Does that help to clarify? Equalx (talk) 14:41, 2 August 2013 (UTC)

Yes, it was helpful. As far as I am concerned, you can go ahead with the study.--Ymblanter (talk) 07:33, 3 August 2013 (UTC)

Link types

Look forward to seeing what comes out of this project. Will you be classifying links in any way? For example, for any claim 'x', there might be a number of very closely related user behaviours that indicate different things. E.g., traversing to article 'y' which is in some way linked to 'x' (they co-occur in the same sentence, etc.) indicates a view of connectedness, looking at citations - and perhaps the tags associated with them, or notes on talk pages - indicates something else, as might looking for multiple citations for the same claim (which might indicate a desire to corroborate). I'm not sure how much data is questionnaire or logs, but I suppose another interesting area is where users want information but there is none (e.g., a missing citation), not sure how you capture that though! Vague ideas! Anyway, I'll be interested to see what you come out with Sjgknight (talk) 10:47, 20 September 2013 (UTC)

Thanks for the vote of confidence! Currently we have classified links into two categories: Searches and direct links. These are pretty simple, but since users can highlight ambiguous text, if we couldn't link it to a page through the API we wanted to let them search on the selected text. If we get a lot of searches, it means people may be using the tool differently or may be expecting pages to have different names. The same is true for redirected page links - If users frequently go to a page that ends up being a redirect to a different page, do they find what they were looking for on the redirect? As for corroboration of sources, I don't know if we've found any behavior like that yet (though we do have a case where that might be really visible, so I'll make a note to look into it). I know that very few if any users visited talk pages, and I don't think very many users left Wikipedia to pursue other sources. As (likely) non-editors, I suspect many of them may not even be aware of talk pages, or be interested enough to follow references off-site, but that's certainly something I'd want to investigate if we found. Our survey generally asked them to locate answers to questions while on Wikipedia, so we likely have inadvertently and unfortunately discouraged things like checking citations.

All of that said, we've got a few more detailed (but hard to process en masse) logs to help us figure out "where on the page" someone was viewing when they linked out. If we can find useful ways to analyze created links automatically, we're hoping to find really cool patterns around link placement - do people generally make links in the same locations as they already existed on the page? If not, can we figure out why not? Equalx (talk) 14:54, 20 September 2013 (UTC)