<a model might estimate prospect for peace based on applying military force, applying political leverage, and doing nothing>
Your post resonated with me and in an ideal world (where I'm the benevolent despot), I'd give it a shot.
As Disraeli and Twain both said, "There are lies, damned lies and statistics." As a person who uses statistics a lot and has played in Systems/AI/game theory for 25 years, I realize that there is always a bit of confirmation bias no matter who gathers the statistic. The linear regression may provide the best theoretical path for an unbiased data set where errors are random, but real data sets are probably never unbiased and frequently have systematic errors. Besides, I can't imagine that we could have a government that broadly used any implementation of nuanced game theory because of personality cults that form in organizations.
At best, we can hope that most agencies will use some practice like the one you propose and that in bulk, there will be little bias. I'm not sanguine about that prospect, though. IMO, there will always be an intuitive component to leadership. I'd conclude that as an engineer, in the presence of broken or unreliable sensors, the intuitive seat of the pants is far superior and more error resistant. |
