All Models Are Useful, Some Are Invaluable
The great success of polling meta-analysis by prognosticators like Sam Wang, Nate Silver and others for the 2012 elections, really underscores the increasing role of models in our day to day lives. This does not come as a surprise to those of us who have been involved with modeling and simulation in their professional careers. On the contrary we are indeed amused by the real (or feigned) shock expressed by the political punditry over the accuracy of these models.
Two things were conclusively established by the success of these poll aggregators’ models:
- Reliable data is the basis for good models
- Good models trump gut feel and instincts any time
Every one of the forecasters mentioned above put out implicit disclaimers that their models were only as good as the underlying polling data; however, by aggregating or averaging the data from various polls, any partisan bias or data collection errors were minimized and their final predictions were really close to the outcome.
Data collection methods also have come under scrutiny for not keeping up with the times and demographics. Polling agencies which used online responses and cell phones to collect data were more accurate than those that only called landlines. The bottom line is that models need to be built on a foundation of good, representative data.
So why do models work?
If we were to critically examine the reason why (good) models work, we would find that models are doing nothing more than abstracting usable knowledge from facts. Let us unpack this statement with a practical example of how the human mind builds “models” that work all the time.
When you try to cross the street and see a car coming down at 25 mph, you mentally estimate if there is enough time for you to cross the street before your two paths are likely to intersect. Now how do you know if you have “enough” time? What is enough? This is where the mind relies on your past experiences built over the many years of successful street crossings! Although you don’t need to crunch numbers and calculate the exact trajectories, your mind is well trained to come up with good estimates. If you believe that you cannot walk away quickly enough, you wait. If you believe that car is reasonably far away, you walk.Continued on the next page