WHY did I win my wager against @ProfessorPape?
I have great respect for him and my respect for him has grown recently. But what did he get wrong?
He believed, correctly, that US security principals faced a strong incentive to do whatever it takes to prevail. He expected them to get sucked into higher and higher levels of escalation to get their way, in a process he calls The Escalation Trap.
I told you from the start that, while this was a good model of an important dynamic at play, it suffered from a grave problem: there was nothing is the model that could ever account for deescalation—it could only explain higher and higher commitments.
So his model was theoretically incomplete in a very important way, and it was not pretty to watch him impose his theoretical priors on a dynamic reality that refused to obey the model.
I tried to think harder about the incentive structures for the two sides. I did eventually formally model the interaction as a bargaining game under asymmetric information. But the underlying logic was already clear intuitively beforehand.
The fundamental theoretical issue is that, with complete information, you get a deal immediately or not at all. There is no possibility of a costly process of armed bargaining. The thing that generates such a process is private information. It is bc one side is not certain about the other’s true preferences that you need costly signaling.
I intuitively knew and later showed formally how, under plausible assumptions about the distribution of Trump’s type, a deal was a certainty in any separating equilibrium.
But all this may be too technical for people with little training in game theory. The basic point is easy enough to grasp: never, ever forget the gap between your model of reality and the messy reality out their. It is not the job of reality to conform to your model. It is your job as an analyst to revise your model if you discover that you’re missing a key piece of the puzzle.
Pape got trapped by his own model. And he is hardly alone. Pape was the only one who got even within shouting distance of me throughout this whole affair.
“The basic point is easy enough to grasp: never, ever forget the gap between your model of reality and the messy reality out their. It is not the job of reality to conform to your model. It is your job as an analyst to revise your model if you discover that you’re missing a key piece of the puzzle.“
Gornji odstavek bi se morali študenti makroekonomine naučiti na pamet. Na izust, bi temu rekli nekoč.
Eden od razlogov, da nisem izbral akademske kariere (Hvala Bogu, čeprav je bila Mama sprva razočarana) , je ravna ta obsedenost z modeli, ki gre vse do točke, ko se meja med modeli in realnostjo popolnoma zabriše.
Stvar je še kako relevantna tudi v sedanji AI dobi. Kje bo vloga človeka v korekciji AI odločitev (ali analiz)?
Všeč mi jeLiked by 2 people
Se absolutno strinjam. In natanko to kažem študentom na predavanjih pri makroekonomiji: originalni model in njegove učbeniške interpretacije, nato pa empirične preverbe modela in realpolitično delovanje v praksi ekonomskih politik
Všeč mi jeLiked by 2 people