Thinking Clearly - Latest Comments in Why Reasoning Matters: Explanations (3)

Re: Why Reasoning Matters: Explanations (3)

Bijan Parsia — Thu, 26 Jun 2008 11:13:08 -0000

Valentin, first, one way I personally make new knowledge is by recombining, specializing, and applying.

Second, this is an old puzzle. Poke around for the paradox of analysis.

Re: Why Reasoning Matters: Explanations (3)

kendall — Tue, 24 Jun 2008 11:31:44 -0000

Valentin, you make a fair point. I probably wouldn't put it just that way in some other contexts, though I think's it a defensible claim. We might need to say "new knowledge relative to existing set of beliefs". It's certainly the case in bioinformatics that previously unknown knowledge was discovered using deductive reasoning. I think that counts as "new knowledge" or close enough.

Re: Why Reasoning Matters: Explanations (3)

Valentin — Tue, 24 Jun 2008 09:10:37 -0000

I very much dislike the contention that a deductive reasoner like Pellet can "create new knowledge", it can recombine, specialize and apply; it does not create knowledge in any meaningful sense of the word.

Re: Why Reasoning Matters: Explanations (3)

kendall — Mon, 23 Jun 2008 19:47:32 -0000

In the talk I point out that both the traditional logic view and the more “small ontologies near the data” view are producing ROI, and the key is to figure out the appropriate things to use for the applications you are building.

Indeed, this is a matter of engineering tradeoffs and requirements analysis -- boring, sure, but true nonetheless. I thought I'd said this clearly in the piece.

Re: the utility of small-o approach--see, I read yr slides right after SemTech!--we've built two production systems for NASA that do exactly that. Customers are happy with those systems, which we built vastly cheaper than off-the-shelf stuff was going to cost. Yay for ROI!

We also develop a Linked Data Browser, jSpace (which Linked Data people ignore since, I suppose, it's not *also* a web browser), so we really do understand that stuff.

As to the rest of yr objections, I take them to come down to style, where tastes and opinions diverge. Our mileages, as they say, obviously vary! :)

Re: Why Reasoning Matters: Explanations (3)

Jim Hendler — Mon, 23 Jun 2008 17:47:26 -0000

Kendall
I've been giving a talk lately called the "Two Towers" where I discuss the issue of the different views of ontology. I pretty strongly reject your contention that every would want sound and complete answers (or consistency) if they could only do it. The whole field of hueristic programming shows that to be untrue - rather there are important problems where it is needed, and important ones where it isn't. Further, the requirements needed to get consistency to be meaningful may be restrictive in many cases.
In the talk I point out that both the traditional logic view and the more "small ontologies near the data" view are producing ROI, and the key is to figure out the appropriate things to use for the applications you are building. Your contention that

"By way of comparison, Linked Data and RDF triple store vendors try to make virtue of their vice—they can’t do consistency checking, so they claim no one would ever want or need to do it. As to this tendency, I blame no one. I’d say the same thing, too!"

strikes me as just throwing more confusion onto the fire, and your last paragraph, which actually makes an important point, does it in a pretty condescending way.

Maybe if you'd like people to stop bad=mouthing logic, you could contribute by making the differences clear (as you're doing) without misrepresenting the alternate view or denigrating the alternate approach. Frankly, I think we all win if we make the different strengths clearer, than if we add further confusion to an already blurred messaging.
-JH
p.s. Feel free to check out the slides of my talk at http://www.cs.rpi.edu/~hendler/presentations/Se...