April 26, 2003
Getting metadata from your users
I've always found it's best not to confront your users with dialog-boxes-from-hell when asking them to apply metadata. Better to have the data collected from them over time, during slightly tangental situations.
I've struck on the idea of this being "Drive by metadata". The idea being you get them hit-and-run style instead of chaining them to the desk and beating them into submission. The latter is certainly tempting but usability studies don't favor it.
Sort of like how Quickbooks handles creating a transaction. If it doesn't know the rest of the details needed it only asks you enough to get to the next step. An invoice can be entered with just a name. The address info won't be requested until later. When you need to mail something it hits you up to apply the rest. This is the sort of thing that more software should try doing. This is also why the additive/fragmentary nature of RDF is a wicked cool thing.
I've used the HTML screensaver for this sort of thing. The pages come up when my machine's idle. I can click on the checkboxes and submit forms without interrupting the screensaver. Works reasonably well. This makes the application of metadata a pretty random event. If a user sees something they understand as needing to be marked up they can do it without disrupting their current work. The machine's idle, otherwise the screensaver wouldn't have come up, so it's obvious they're not 'doing something' on it.
Anyway, if you want better results, give better data.
+1 from me on this -- no questionaire should ask anymore than it absolutely must, and this is especially true on HTML forms; in the 'metadata' context of personal data records in member profiles, this is why teledyn.com doesn't require all the personal detail most others think they need for "marketing purposes" -- who cares about the demographics if I have your attention? We adhere to the "You First" privacy policy, and I think others should too.
there's another problem with asking for metadata: People, intentionally or otherwise, very often give wrong responses. Every psychologist knows this (why they persist in using questionaires is a paradox) -- if you stop to observe the subjects, you will invariably find their behaviours and their self-perceptions do not match.
Same is true of their ability to classify their own information. The ENT information they give will still be useful but it should only be taken as further input to the empirical observations on how the data actually behaves.
My guess is the more adept among them will also be observing themselves, watching how their choices affects the goals they have for their information strategy and thus will anneal themselves to TopicExchange maps that make more and more sense over time, but to be effective in mapping metadata, it's naive to think we can do it once and be done with it, or that a questionaire answer on Tuesday is the same as the answer for Thursday.
Posted by: mrG on April 26, 2003 10:03 AM
