Shirky, Syllogism and the Semantic Web

A friend recently sent me the link to Clay Shirky’s piece on the Semantic Web  with a “I assume you’ve seen this, what do you think?”

I had seen it, but I hadn’t looked at it for years. So I went back for another look.

As usual, Shirky’s writing is intelligent, insightful and even funny. Recommended reading. I had hoped the ensuing years would prove “us” (Semantic Technologists) right, and that the argument would look amusing in retrospect.

Alas we still have a long way to go to staunch the critics. More on that in a future blog.

For today’s blog I have to point out the real irony of the article that I managed to miss the first time I read it.

At the risk of oversimplifying his article to the same degree he oversimplified the Semantic Web, the essence of the article went like this:

· The Semantic Web relies on syllogisms “The semantic web is a machine for creating syllogisms”

· Nobody uses syllogisms “it will improve all the areas of your life where you currently use syllogisms, Which is to say, almost nowhere”

· Therefore nobody will use the Semantic Web “it requires too much coordination and too much energy to effect in the real world”

The first two quotes from the opening the last from the closing

The irony being of course, that this entire article is a syllogism. To make one of the major premises of an argument that something will fail because nobody uses that style of argument, reminds me of the admonition Yogi Berra gave to some teammates who had suggested a restaurant for the evenings dinner “Nah, nobody goes there anymore. It’s too crowded.”

The article points out some areas we need to pay more attention to, including controlling the hype machine. Reading between the lines, it appears that one of his major points is: the web is complex and only humans can really understand the nuances that our complex utterances mean.

But traffic is complex, and we know that traffic lights will never be as good as police in managing an intersection, but we’ve decided that an automated solution that gets us consistently pretty good results is good enough.

Back to the article, he relies on Lewis Carroll’s syllogisms as a critique of the medium, and by extension, the Semantic Web. The knock out punch was meant to be a five line syllogism about soap-bubble poems. But even here there were two implications: one that humans could follow this logic, and two that formalized ontologies could not. I of course rose to the bait and tried to formalize this syllogism.

I was not successful. Not because of the poverty of expression in the Semantic Web, nor even my own understanding, but attempting to get formal about this doggerel shone a light on the fact that it doesn’t make any sense at all. Indeed if he makes a point at all it is that humans can often get fooled by things that sound like they make sense, but actually don’t. Seems to me, defending that level of confusion and ambiguity isn’t an argument against the Semantic Web.

Comments

[Mehalia]

Keep up the good work.

[Marge]

Thanks for writing this.

[Ivan Handler]

Another take on this is that Shirky is right but, depending on how things evolve, he comments are probably irrelevant.

The problem with AI (where the ideas behind the semantic web ultimately originate) is that attempts to find universal mechanical solutions to the problem of intelligence are probably doomed. On the other hand, this does not mean that specific problems can not be tackled. The real implication is that the ability to extend a solution in one problem domain to other domains is limited.

In that case, as specific problems are solved using semantic technology, the utility of the technology will grow. The fact that it may not be universal in the sense that the Star Trek computer was is a minor point at best. Businesses need better ways to solve problems faster. I believe that proper use of semantic technology has the potential to be very useful.

Once you leave the realm of complete generality the need to restrict one's deductive technique to syllogisms also evaporates (though I am far from clear that that restriction is justifiable in the first place). Based on a model of the domain, semantics (or ontologies) can be used in ways that take advantage of assumptions of structure that are true throughout the domain. If this is starting to sound like computer programming, it should since that is all we can ever do with computers to start with. Semantic technology allows us to apply techniques we understand to a new class of data, it can not allow us to do anything you can not do with a computer...