Tuesday, July 11, 2006

The User-Generated Web (web 2.0?)

The user-generated web (blogs, wikis, Amazon customer reviews, etc. etc.) is all about trust and credibility and good stuff like that. And often I think of trust and credibility as a ranking system: X trusts(K%) Y, something like that, X gives Y a rank K between 0 and 1. Okay, if X trusts(P) Y and Y trusts(Q) Z, then we might have a preliminary X trusts(P*Q) Z, just as a loose way of pulling in trust over a network. (I don't have any actually plausible formulas for this, but it's not terribly difficult to code such things.) Right now, unfortunately, I'm thinking that this is not only a very inadequate model but it's actually going in a bad direction. Let's consider.

Tim Bray is an XML god whose words when he blogs are Truth. For me, that is. Or are they? When he described instapundit (a supporter of gay marriage and abortion rights) as "right-wing", I smiled and passed on but mentally downchecked his credibility...or did I? It certainly has no bearing on his credibility about XML. None. Well, almost none. My assumption is that instapundit's frequently-ill-thought statements (e.g., the Palestinians don't "deserve" a state, which to me seems not just ill-thought but ill-formed; what could it possibly mean?) annoyed Bray in a way which reminded him of genuine right-wingers, or more likely of his mental model of "the right wing"...it is possible that the thought process which led to the "right-wing" label could lead to other oddities.

Still, let me follow up the notion of "credibility about XML." Maybe credibility is a total ordering, but only one topic T at a time: X trusts(K%) Y about T. hmm. We could imagine a del.icio.us in which each topic-tag can be given a color, red-orange-yellow-chartreuse-green, (note wikipedia's color list) where the color indicates level of trust, yellow being caution, green is trusted, and so on. Brightness would indicate confidence in the trust-level being expressed.

This might be really useful: you could not only ask what articles on topic T are trusted, but you could ask who trusts them and what else they trust, and whose trust on topic T resembles yours, and so on. I'd really like to see that.

Instapundit is mostly libertarianish -- gee, so am I, some of the time, should I trust him? Well, some of the things that annoy Tim Bray annoy me too, though I subscribe because I trust Insty to point to interesting stuff, often. That's not readily expressed via a color scheme. Usually he even points to those who disagree or who think they refute what they think he said, but not always. When he seemed to approve one Bob Carter saying that global warming peaked in 1998, he failed to cite a fairly obvious and quite convincing (to me) Deltoid refutation. I think that's typical of what I see as a common failure on his part; he thinks about technology, not so much about science, and he's vulnerable to hearing what he (as a libertarian, not a "right-winger") wants to believe about global warming and the lack of a need for government intervention. And his blog is not a Thinker blog anyway. Can I express this in a useful way? I'd like to say that this Bob Carter's credibility comes way down -- careless or gullible or sneaky, whatever, at least this time -- while Tim (Deltoid) Lambert's credibility goes up significantly -- careful and accurate, at least this time -- and Glenn Reynolds' credibility goes down very slightly -- he didn't express a strong approbation, and he certainly didn't claim to have done any analysis.

Maybe that's something I can express. Somehow.

Or then again, maybe not.


Post a Comment

<< Home