Chasing Josh, Reading Laudan, Watching Fights

Looks like Josh is out-posting me pretty thoroughly so far. Well, he's more opinionated than me, and he's got less nuclear family to not neglect (one cat and some beer vs. a 6-month pregnant wife, 4.5 year old daughter, and some beer and wine), so, ipso facto, he has more blog posts than me.

I've also helped Josh by commenting on two of his posts, so not only does his blog have more original content, it has more readers, too.

Unfortunately, unlike Josh, I've been busy with non-class stuff that would take more time than I'm willing to spend blogging about. Either that or it would end up not doing to topic justice, as I would have to skimp on the details. Animal language, innumeracy, gaiety, bilingualism, alleged Mexican vote fraud,... Josh's got it all. Me? I've got an initial explanatory post, this meandering, self-absorbed silliness, and a mediocre political piece that has already been corrected and improved upon by none other than Josh, the bastard. He did say I made a good point (the bit near the end of my post wherein I spake unkindly of public school), though, so give me some credit.

Now I'm improving his google cachet by linking to all but one of his posts. Well, I don't like to do things half-assed, so: the only other post.

Josh, you're welcome.


Some non-horribly derivative (though still entirely derivative) content: I'm currently reading Truth, Error, and Criminal Law, by Larry Laudan (the "famous" philosopher and historian of science), and, strangely enough, its subject matter overlaps somewhat with my own professional interests, in this case, signal detection theory (link goes to a site of unknown quality that includes a java applet SDT tutorial that looks nifty). The book concerns the extent to which, and manner in which, criminal law functions as a truth-finding system. In laying out the basis for his investigation into criminal law's epistemic bona fides, Laudan distinguishes between material and probatory innocence and guilt. Material guilt and innocence are matters of fact, whereas probatory guilt and innocence (findings of guilt and innocence) are matters of inference from evidence.

Laudan argues, correctly it seems to me, that, insofar as criminal law functions to find the truth with regard to a criminal case, error reduction and and error distribution are central to the issue. Errors here are cases in which material guilt is paired with probatory innocence and cases in which material innocence is paired with probatory guilt.

On reading the early section describing this general schema, I immediately thought of signal detection theory. If you followed the SDT link above (or if you already know about it), you can see where I'm going with this. In the simplest signal detection task, each trial in a perception experiment either has a physical signal present or not, and the subject has two response options - signal present or signal absent. Errors here are cases in which the physical presence of a signal is paired with the signal-absent response and cases in which the physical absence of a signal is paired with the signal-present response. In SDT lingo, these are called misses and false alarms, respectively (non-errors are called hits and correct rejections).

In a simple perception task using, say, a pure tone signal, all of the relevant probabilities are either determined by the experimenter

Pr(signal) = 1 - Pr(no signal)

or readily calculated conditional probabilities

Pr("signal-present" | signal) = Pr(hit)

Pr("signal-absent" | signal) = Pr(miss)

Pr("signal-present" | no signal) = Pr(false alarm)

Pr("signal-absent" | no signal) = Pr(correct rejection)


Pr(A | B) = Pr(A & B) / Pr(B)

Now, the utility of a signal detection analysis depends, in no small part, on the fact that Pr(signal) and Pr(no signal) are determined a priori. The associated probabilities in Laudan's analysis of criminal law are Pr(material guilt) and Pr(material innocence), which are clearly not known, much less conciously determined, a priori (at least let's hope they're not).

The utility of signal detection theory is that these probabilities can tell us about sensitivity independent of decision bias and decision bias independent of sensitivity. Without the baseline stimulus frequencies, though, SDT is a much more limited tool.

It will be interesting to see what Laudan does with his SDT-like schema given the limitations the subject matter places on his data. Based on the Abbreviations and Acronyms Used page, he will be employing two ratios:

= true acquittals / false convictions


n = false acquittals / false convictions

It seems reasonably likely that whatever numbers Laudan uses as estimates of the number of true/false acquittals/convictions will be biased, though, since we don't know the truth of a case - we only know what a jury (or judge, depending on the legal tradition) decides about a case. Okay, we don't only know what a judge or jury says. It's possible that we may be confident of someone's guil but, for a variety of reasons, the case doesn't make it to trial. In any case, who knows how good the numbers are?

As I said, it will be interesting to see what he does. More later, when I find out.


I started this post before UFC Unleashed started on the TV (that's TEE-vee) this evening. Progress on these last couple paragraphs has been halting and irregular. The tally tonight, as of 9:39 Eastern Standard Time (actually, Indiana Stupid Time), is two guillotine chokes, one rear naked choke, and one bloody face (no victor yet, though it's looking pretty lopsided [update: round two, though fairly close, likely went for bloody-faced Georges St. Pierre, so it's not as lopsided as I thought it would be {update 2: bloody-face had a good third round, and BJ Penn looked pretty tired at the end, so... the decision goes to St. Pierre, which surprised me, as the next pay-per-view has, as its main draw, BJ Penn vs. Matt Hughes, and the "prize" for winning the Penn-St. Pierre fight was a (rematch) fight with Matt Hughes. Sounds like a mystery. More later...]).

Among the reasons I love the UFC is the eloquence of those involved. Just now, one of the announcers (I think it was Joe Rogan) described the bloody-face fighter as "wincing his eye." Some months ago, during the third season of The Ultimate Fighter, Tito Ortiz said, in describing the mental aspect of the fight game, that "it's like chest."

They pays 'em to fight, not to talk. The fighters anyway. I don't know what Rogan's problem is.

No comments: