Developers don't use the Semantic Web because they do

When I recently took a trip (to W3C TPAC!) my mobile phone guided me every step of the way, from advising me on ticket purchases based on my previously declared preferences and behavior, to reminding me of important deadlines, to offering to take care of incidental tasks (like renting a car or hotel room) or warning me about appointments I would miss. When I got to Lyon, maps and suggestions clearly based on inference from my position and history came at me furiously.

Essentially, much of what TBL promised in 2001 is real. Now. Today. Those parts that haven't come to pass yet often haven't because of legitimate non-technical issues involving privacy or agency or equity or any number of important and complicated social questions.

But in case anyone is unfamiliar with the kinds of services I mentioned, none of them were provided by open tech, none were substantially under my control, and none of them offer the potential for me to join with (or even examine) the data on which they are based, except in the most trivial ways. In my case, it's Google (I have an "Android" -type phone) that owns all that, and Google has no incentive at all to share it with me. (I don't own one of the new "home assistant" devices, but I imagine the situation is pretty similar there; astounding capabilities that fulfill the SemWeb vision, but which are outside the control of the user except in the bluntest ways, and are decidedly owned by the business entity that supplies them.)

I'd like to offer the following thought about why "middle-third" devs don't use semantic tech: they do. They just don't use it in anything like the form we imagine it taking.

Devs, just like the rest of us, already use semantic technologies _constantly_, when they rely on a Google API that itself relies on Google's Knowledge Graph, or issue a query to Facebook's Graph API, etc. They just don't do it with open tools or open data or in any way that might change the relations of ownership. Is that really surprising? Is that really a technical problem to be corrected by emending our specifications? Have we as a community done what we can to make clear how Facebook's API differs from a downloadable IMDb dataset differs from Wikidata differs from the Gene Ontology? These can all be called "linked data" but they are obviously all very different. The use of RDF or its absence isn't at all what distinguishes them or makes them variously attractive to various audiences.

Semantic facilities have taken off like wildfire, but because they turn out to require a great deal of capital to build and deploy at scale (=on the Web), they have been built and are owned by those entities that can gather and deploy capital at scale. That's not individual users, or small businesses, or academic institutions. It's exactly the large corporations that do, in fact, dominate the web now. Why wouldn't they dominate the Semantic Web as well?


---
Adam Soroka
Research Computing : Office of the CIO : the Smithsonian Institution

Received on Thursday, 22 November 2018 16:17:51 UTC