Thursday, April 20, 2006
CAA 2006 - Day 3: Archaeological predictive Models, XML, Tagging
Another enthralling day in Fargo, ND. There was an abundance of great papers today and plenty of thoughtful discussion to match.
With my chances for a free happy hour drink slipping away, I will just quickly cover some of the highlights.
Inductive vs. Deductive Predictive Models: Battle-Royal!!!
Will inductive models really send earth spiraling to hell in a handbasket? Or are deductive models the way to true enlightenment? Well, this ever present topic was brought up today in a great symposium on Archaeological Predictive Models.
After David Ebert, not to be confused with James Ebert, discussed the "7 deadly sins of inductive modeling", two authors followed with papers, based on opposed theories, that produced very compatible results. Scott Madry of the University of North Carolina, presented a 7 county wide predictive model based on an inductive correlative model. Following this, Thomas Whitley, of Brockington and Associates, presented a deductive based behavioral model for a 2600 square mile (did I record the right?) study area in South Carolina. For my money, this battle of the Carolina's was the pinnacle of the CAA so far. Okay, so no one was throwing chairs, but I sure did scribble some frantic notes.
Basically, the inductive model is criticized by distilling the vast diversity of the environment and archaeology to a series of correlations. Though, the end result is a quantitative, testable, and field verifiable model which fits very well with the requests of the Department of Transportation. Alternatively, the Deductive South Carolina model establishes both environmental and behavioral adaptations which are cross-correlated into a matrix of cost-benefit surfaces. These surfaces are composed into any number of formulas of settlement/subsistence & behavioral adaptations to make testable hypothesis to be modeled. The end result of this approach is a hypothesis testing framework of formulas. This is great for explanatory research, but not great of DOT review. In this case, Whitley combined all 46 testable formulas into a single surface which was encoded with a 1-10 rating of site "possibility"; I'm not really sure what the correct term for that metric would be.
All in all, it was a great demonstration of the application of both methodologies. I truly enjoyed both sides of the theoretical coin. Certainly in the future I will ramble on a bit more about the war waged between inductive and deductive models.
Okay... I would like to include a few words here about the papers on XML and DBs I saw today, but I happened to be out a little too late with a couple of good friends, so it will have to wait. Tomorrow (actually today!) will be the last full day of papers, so it should be exciting!