« Reviving the Personal Touch: Netflix Bucks a Trend | Main | Bet2give: Can Markets Be Compassionate? »

31 August 2007

Comments

Fabius Maximus

1. I may have missed something here. The participants in this debate seem unaware of the large literature -- dating back to the 1920's, at least -- on the difference between risk and uncertainty.

I think (it's not my field) the definitive work on this was "Theory of Games and Economic Behavior" by John von Neumann and Oskar Morgenstern (1944) John von Neumann and Oskar Morgenstern. They are very different concepts, both relevant to this discussion.

2. I thought everyone knew that the major barrier to forecasting was that we cannot see beyond choices we do not understand. Here the definitive works are the Matrix movies.

Peak Oil, terrorism, how the US will (or will not) survive the next economic downturn ... all depend on our choices.

Many things might influence the odds of a major terror strike at the US.

* Bombing Iran.

* The decision to stay or leave Iraq.

* Who wins the domestic debate of “civil liberties” vs. “national security.”

* Do we continue to tolerate a Homeland Security establishment which in some respects operates like the Keystone Cops – or do the President and Congress force meaningful reforms?

Fabius Maximus

One last point.

Even more important, these forecasts cannot be made reliably without some form of model incorporating quantitative estimates for the major factors and assumptions about their interrelationships. Lacking that, these are just guesses.

You have yours. I have mine. Good to debate over drinks, but that’s about it. Even expert guesses are of little value (there is a large body of evidence for that, at least).

Art Hutchinson

All good points, FM. Thanks for commenting. The only thing I would add is this:

Models, while valuable, tend to restrict the range of people who can provide cogent, direct input. They also incorporate--in much more subtle, inaccessible fashion--the special assumptions of the model-builders, maintainers and interpreters. This priesthood gains special status in what is ultimately a human-based (as opposed to divinely inspired or scientifically grounded) forecasting effort.

This is not an indictment so much as a fact. Models can be useful. They are not a panacea. The antidote is transparency and education to ensure that a wide range of people can see, understand and offer their critiques of the models and their conclusions.

Quantitative (usually computer-aided) models, because they are even more powerful still, engender even more of the same dynamic.

They are just a way of processing all-too-fallible human assumptions much faster and with higher degrees of complexity. Unfortunately, that same complexity and speed (the things that give the models power) all too often create an unwarranted (or at least unprovable) impression of super-human prescience.

Sharp, powerful tools require especially strict precautions.

Fabius Maximus

Absolutely agree! This is the boundary problem. Where do we draw the limits of the domain that we model?

The larger the space modeled, the more likely we get all the key factors. But the complexity -- hence the number of required assumptions and the opportunity for error -- go up as fast or faster!

Models are useful tools, not magic wands. Unfortunately, like science and democracy, they are among the best tools we have. But still just tools!

Unfortunately senior decision-makers tend to dislike models.

1. Models limit their ability to select the decision (the process is often decision first, then justification).

2. Models require participation of mid-level experts (mere technicians), instead of limiting the discussion to the club of seniors. The Best and the Brightest describes the importance of this dynamic in policy-making for the Vietnam War, excluding from the room folks who actually knew anything about the relevant subject.

3. Perhaps worst fo all, models require all those numbers! Seniors prefer a greater role for the "experience" and "judgement" that seniors claim as their special assets.

Chris Albon

This post alone gave me reason enough to subscribe to this blog. Excellent!

The comments to this entry are closed.

My Photo

June 2009

Sun Mon Tue Wed Thu Fri Sat
  1 2 3 4 5 6
7 8 9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30        

Sites, People, Blogs