Ontologies are overfeared
Clay's recent polemic misses a key point I think.
Yes, classification is political. Yes, classification is imperfect in a changing world. Yes, we need to be wary of all that. But what are "tags", "folksonomies", "ontologies", and "tag clouds" when you put them together? A classification system, albeit a glossary.
Ontologists of the Semantic Web ilk are not trying to model the world, as I understand it. It is a non-religious undertaking. What they are (generally) trying to do is to improve Computer-to-Computer communication so as better to support computer-mediated communication between us.
Where possible, ontologies add some logic into the mix to allow a level of inference. Zip code is equivalent to post code, on some level. But where they can't add the logic, then they don't.
Tags aren't and can't be a proof that ontologies don't work. What tags could be, is a means identifying which areas have enough (political) consensus to be worth developing an ontology for, for understanding how groups talk about certain ideas/events/things and so helping modellers mitigate their observer bias, and the areas where old ontologies are breaking down and an indicator for how to fix them.
Comments
I just read Clay Shirky's article, starting out sceptical. Actually, it's a bit of a tour-de-force, isn't it?
But I think we need more classification. We need lots of models for organising the information that's out there into useful sets. It doesn't matter how wrong or limited they are. They only have to be useful while they last.
Posted by: Mike Harper | June 5, 2005 9:59 PM
Hi Mike
Clay certainly has a take-no-prisoners point of view ;) But I think the general point - that classification is political, skewed by the interests and worldview of the classifier etc is right.
I like your "more classification" approach - really made me think - and I suppose the easiest way to do that is to get more people classifying things the way they want to.
The key thing for me though is the cost of classification. Ontologies are time-consuming things :( So while yes, I think I'd broadly agree that the models only have to be useful while they last, I think you need either need to aim as high as possible (make the models last a long time) or find an easier, cheaper way to make them.
With more classification (via more people) I think we're beginning to reduce the cost of classification, and are beginning to get a handle on where you get most "bang for your buck".
Cheers
Piers
Posted by: Piers | June 6, 2005 9:24 AM
Yes, classification is political. The answer then is to have explicitly political entities do the classification, and then compare the outcomes of several such results, mixing and matching as one prefers. For instance the list of policy terms at the Green Party of Canada living platform will differ from that of other parties, but, differences can be noted and resolved.
It's time to stop pretending this can be done without real world politics being involved.
Posted by: Craig Hubley | June 19, 2005 9:26 PM
Hi Craig, is that politics capital P or politics lower case p (or both)? To be honest I'd only really been thinking about lower case p politics, but I like your example (And on a geeky note, Tikiwiki's great eh?;))
Anyway, your comment has definitely got me thinking - thanks!
Posted by: Piers | June 20, 2005 6:35 PM