World Views Are Not Data Inconsistencies

I’m at Wikimania 2013, listening in on the WikiData session. WikiData is the Wikimedia Foundation’s attempt to go beyond prose in Wikipedia pages and provide a reference data source. An obvious problem is that any such data source needs an underlying model of the world, and that sometimes it is not only hard to gain consensus on that model, sometimes it is impossible. Basically, different world-views are simply incompatible. When asked about this fundamental problem, the audience was told that such inconsistencies are handled using multi-valued properties. Ignoring for a second, that world-views cannot be reduced to individual properties, my major point here is that world-views are not inconsistencies in the data. Different world-views are real and justified, and there will never be only one world view. The moment we all agree on one world-view, we have become the borg.

Update: Daniel Kinzler corrected me that this must be a misunderstanding: WikiData can handle multiple world views well by way of multi-valued properties.

4 thoughts on “World Views Are Not Data Inconsistencies

    1. Dirk Riehle Post author

      Hi Danny: Daniel Kinzler already corrected me, see above. I think I responded to your comment (as I heard it) that the Taiwan conflict would be a data inconsistency. Daniel also pointed out that you’d be the first to acknowledge that you can’t unite the world under a single world view but that there’ll always be many (probably as many as there are humans on this planet).

      Reply
    1. Dirk Riehle Post author

      This is not to say all technical problems seem to have been solved. If you view one world view as its own graph, and WikiData wants to accommodate any number of such world views, then the overall structure becomes an overlay of graphs. Daniel assured me that some of this can be modeled, but it is not clear to me how to keep these graphs separate where necessary and join them on data where desired. This may be easy with base-level data, but gets much more tricky once you have disagreement on the model level. Anyway, kudos to an ambitious effort.

      Reply

Leave a Reply to Denny Vrandecic Cancel reply