Sunday, November 16, 2014

PSR: A Gem

Sometimes, there is a thread on PSR that amuses me.  Here's one:

Theory Presenter: [Snipping 75 minutes of reading without eye contact.] "...so as you can see, I have reconceptualized and reconsidered and -icized and -atized until this problem I talk about is clearly both like and unlike what Hobbes, Locke, Rousseau, Plato, and Arendt implied by choosing one word instead of a universe of other words in these few sentences no one else has really talked much about." 

Theory Search Committee Member: "Well, certainly, but since we have clear answers about this philosophical problem deriving from Augustine's flirtation with manichaeism [snipping 15 minutes of bibliographic citations] ... what could we turn to in order to understand why what you have presented improves our understanding of the problem at hand?" 

Audience Member In the Back: "Data."* 

*This totally happened.

1 comment:

Simon Spero said...

The key thing to remember about most philosophical issues is that they are not subject to empirical validation or invalidation.

Indeed it is possible for there to be no possible data that could distinguish between two consistent theories (undetached rabbit parts, or momentary rabbit stage?) .

They instead succumb to Death by Inconsistency, or collapse under the weight of all the incremental ad hoc changes needed to address the twisted counter-examples thrown at them.

The committee members question is the right one- what makes your approach more useful in understanding the problem- is the right one.

As a practical matter, one set of philosophical assumptions may make it easier to formulate a theory that is subject to invalidation. What matters is that the assumptions chosen are consistent (oh yeah, and that the theory adequately explains/predicts the data).

The main reason for understanding enough philosophy is to avoid accidentally doing it.

Choosing between two models based on AIC/BIC is using philosophical assumptions (principle of parsimony).

So is coming up with a predictive model that works ok, without a theoretical underpinning (pragmatism).