[Wittrs] Statistical Analysis v. Connoisseur Judgment
- From: Sean Wilson <whoooo26505@xxxxxxxxx>
- To: LAWCOURT-L@xxxxxxxxxx, wittrsamr@xxxxxxxxxxxxx
- Date: Wed, 2 Mar 2011 22:30:22 -0800 (PST)
(In reply to Tracy's mail [below] regarding the football coach who used
statistics).
Tracy, there are several points that your mail raises. This reply is only
concerned with the Paul Brown example.
Let's imagine two coaches: one who keeps stats on his or her players (and looks
for trends), and another other who doesn't. Let's call one Paul Brown and the
other Phil Jackson. For simplicity's sake, let's assume both are coaching the
same sport (football). Note that the issue isn't whether one of the coaches
sees
patterns or trends -- for surely they both do. The issue, rather, is what kind
of cognition (or personality type) one relies upon to make sense of the
activity. One doesn't need stats to see the dynamic nature of Jim Brown. Many
experts could see the trends in production that your stats purport to show
simply because they have developed "a trained eye." That is, they have
developed
"connoisseur sense." In fact, by the time one becomes a connoisseur, they have
ascended to the point where they can see through (or govern) stats. (See
Wittgenstein on aesthetics, "seeing as" and imponderable evidence).
Here's what the issue boils down to: what are you using the stats for? If the
answer is for memory, this says more about psychology than it does
epistemology.
If the answer is "analysis," this suggests that the logic of judgment is beyond
our control for some reason. This is fine when humans lack judgmental capacity
for some reason -- when information is poor, e.g., or when we ourselves are
simply not connoisseurs of the activity. As I said before, statistical analysis
is best when our experience and perception of the thing in question is poor. If
I knew nothing of football, I would indeed want to know what a rushing average
was for X before I decided to place him or her in the game. Just take a look at
fantasy football: that's what the entire activity is (who's better at making
those judgments). But if I'm a connoisseur, the rushing average is something
that looks to ME for correctness. I say what is significant (or not) about it.
I
can tell you its flaws and why it either does or does not correctly report the
happenstance.
And so, my sense is that the Paul Brown stuff is more relevant to personality
type than to the secret of coaching. In fact, I don't know if you realize this,
but your argument seems to be an endorsement for analytic formalism. What is
more important to knowing: analysis or meaning? For a period of time in
philosophy, people thought analysis was king. Wittgenstein showed us, however,
that meaning is more important. Wittgensteinians are anti-formalistic across
the
board -- we dislike it as much in language as we do numbers. Wittgenstein
didn't
change course: he simply transcended analyticity. It's like the moth turning
into the butterfly. Culture is what ultimately is king. This doesn't mean
analysis or formalism is bad; it only means that it can never be unto itself.
It's always there as a servant.
Politics is no different from football. Those who think that quantitative
statistical analysis of behavior reveals "the truth" are no different than
those
who understand football as a fantasy league. (Or those who say, e.g., "the
numbers tell you that you must run the ball more than 12 times to win in the
Super Bowl"). The real masters are those who have fostered a
acute connoisseur sense and who, therefore, have a keen eye for the aesthetic.
These people will tell you when it's time to stop running.
(I'll send one more mail tomorrow that address your other points)
Regards and thanks.
(P.S. -- sent to Wittrs)
Dr. Sean Wilson, Esq.
[spoiler]Assistant Professor
Wright State University
Personal Website: http://seanwilson.org
SSRN papers: http://ssrn.com/author=596860
New Discussion Groups! http://ludwig.squarespace.com/discussionfora/[/spoiler]
________________________________
From: Tracy Lightcap <tlightcap@xxxxxxxxxxxx>
To: Sean Wilson <whoooo26505@xxxxxxxxx>
Cc: LAWCOURT-L@xxxxxxxxxx
Sent: Wed, March 2, 2011 3:01:20 PM
Subject: Re: On Law Professors and Quantitative Methods
As to Sean's points, a mixture of football and social science.
1. Why did Paul Brown always see that Jim Brown carried the ball twice as much
as Leroy Kelly, despite that indisputable fact Kelly was a great running back?
Because Brown, probably the most data driven coach in football history (and,
coincidentally, one of the greatest) kept stats and knew that Brown was more
likely to a) carry the ball for a higher average and b) consistently get more
long runs. He also knew that Frank Ryan was consistently better throwing the
ball when Brown had the other side occupied. Was this something he could see in
games? Sure, but the deciding factor was the stats, just like it is for almost
every winning football coach in the business these days. I'm sorry, but it
simply isn't true that our perceptions can deliver the same kind of information
as a careful analysis of data can. Also, the experience of researchers is not
the final check on the analysis; the data is. The hallmark of good scientific
work is that it is counterintuitive.
2. Most "facts" in most sciences are, at bottom, stochastic. That's why our
perceptions are less use then we usually think. As to "constructing what you
show": all sciences do that. The stochastic nature of events forces us to
impose
an order on them in order to understand them. This does not compromise the
picture; it creates it and makes it useful. Ludwig was good on this point, but
I
doubt he would have gone as far as you have here.
3. Surely you aren't claiming that certain states of affairs are easier to
"scientize" then others? Now we are straying into an area where the actual
character of science is being mythologized. Different states of affairs require
different kinds of science. What you seem to be claiming here (if I'm not
mistaken) is that there is only one kind of science: what Freeman Dyson calls
"Athenian" science. That isn't the case. The scientific method must be applied
differently in - say - meteorology, quantum mechanics, and political science
since the subjects analyzed and the purposes of the analysis differ so greatly.
But it is still recognizable as science in each of these fields. As to the
methods used, I'm as leery of the maths as the next guy, but only because
parsing out of errors is important to know what the models tell us. As to the
rest, GBS put it best, "If experience was the only teacher, then the stones of
London would be wiser than the wisest man."
Well, enough.
Tracy
On Mar 1, 2011, at 11:02 PM, Sean Wilson wrote:
(In reply to Susan Lawrence, who writes, "Exactly what would a Wittgensteinian
>analysis of settlement rates; attorneys' fees in class actions; or the effects
>of defendants' criminal records on case outcomes look like?")
>
>First, let me concede your point. I never meant to say Wittgensteinian method
>could help for these types of issues. The method is only concerned with
>dispelling confusions. (One of the ways in which confusions dispel themselves,
>by the way, is that they turn purely informational: one merely has to look and
>see).
>
>But let me at least comment on what a philosophic-minded person might worry
>about with respect to studies you mention. Below are edited passages from a
>mail
>
>I just wrote to Frank Cross (privately) concerning the statistical analysis of
>football. They apply equally to the statistical analysis of any human activity
>(settlement rates, fees, etc).
>
>1. Methods are only helpful for things that humans cannot, themselves, see or
>experience. (E.g., what effect drug X has on certain kinds of lipids 24 hours
>after intake). The trouble happens when methods become superimposed on things
>we
>
>CAN experience (e.g., Who was a greater risk taker -- Bradshaw or Favre?).
>When
>we impose stats upon things we can, in theory, perceive by watching -- we lose
>information rather than gain it. The stats mislead. Example: as soon as you
>talk
>
>about a "quarterback rating," you mislead someone who neglected to see
>Roethlisberger in the AFC championship, and only heard about the low number.
>The
>
>person who watched it always has the advantage.
>
>In fact, what social scientists really do is bend their analysis to suit
>what they think they are seeing in the first place -- that's how they "check
>their work." It only gets released if it can act as a good piece of
>journalism.
>
>2. Stats work best with two companions: (a) things that are
>naturally commensurable (e.g., the economy and dollars); and (b) things
>naturally stochastic. Social studies tend to lack these things. You have to
>construct what you want to show. Same with applying stats to football
>and settlement analysis. This always compromises the picture.
>
>
>3. Very frequently, the thing social studies want to show is not something
>that
>can easily be scientized. Look at the idea of "inflation" (troubling). Now
>compare that to "ideology" (much worse). Now compare it to whether a team has
>to
>
>run to win the Super Bowl. Or whether something affects settlement. The
>settlement question is no different than the Super Bowl. Same exact thing.
>Your
>answer yesterday is not the same as tomorrow. You can only provide a kind of
>journalism on the subject.
>
>You have to ask yourself: what does a regression analysis of settlement rates
>tell you compared to what (say) a good NPR or 60 Minutes investigation might
>have? Or what 30 years of experience tells you? Or what a collection of people
>with major experience might say? What you must understand is that mathematics
>can only ever be a kind of sculpture when summoned into the service of social
>studies.
>
>The stats are only good if we think the journalism is accurate. If we think
>the
>portrayal fits.
>
>(Also, learning Wittgenstein can help students see this last point)
>
>[P.S. -- Sent to Wittrs]
>
>Regards and thanks.
>
>Dr. Sean Wilson, Esq.
>Assistant Professor
>Wright State University
>Personal Website: http://seanwilson.org
>SSRN papers: http://ssrn.com/author=596860
>New Discussion Groups! http://ludwig.squarespace.com/discussionfora/
>
>
>
>
Tracy Lightcap
Professor and Chair
Department of Political Science
LaGrange College
601 Broad St.
LaGrange GA 30240-2999
(O) tlightcap@xxxxxxxxxxxx (H) altlamp@xxxxxxxxxxxxxx
706.880.8226
www.lagrange.edu/academics/political-science/faculty/tlightcap.aspx
Other related posts:
- » [Wittrs] Statistical Analysis v. Connoisseur Judgment - Sean Wilson
