4 ms·

Thanks for your honesty; I get really annoyed sometimes when economists are held up as the gold-standard of societal prediction. Some of the 'newer' ideas that

Thanks for your honesty; I get really annoyed sometimes when economists are held up as the gold-standard of societal prediction. Some of the 'newer' ideas that have made their way to the core of economics study, like game theory, are still painfully simplistic models that fail to take into account more complicated dynamics that result in outsize error signals. Usually, in the physical sciences it's possible to justify throwing out some higher order terms, but when you start talking about "externalities", things can get unhealthily misrepresented quickly. Frankly, it's no wonder that our species hasn't achieved a higher level of self-reflection when human behavior itself is so complicated.

qubex 7y ago
I came to macroeconomics by way of game theory. I'm going to step in because it's actually a far more sophisticated discipline than one would at first imagine given the (admittedly extremely pedagogical, verging on the puerile) examples one is likely to encounter when not studying it in depth (Prisoner's Dilemma, I'm looking at you). Combinatorial Game Theory has been fundamental in addressing “artificial intelligence challenges” such as defeating human Chess grandmasters and more recently, despite the hype about neural network self-training, Go. Bayesian Game Theory deals with how a rational agent starts off in a situation whereby it knows little of what it is competing against and gradually refines that knowledge whilst also getting the best results given the imperfect information available. Differential Game Theory de-discretises time and is useful, for example, in programming both the manoeuvres to be executed by a surface-to-air missile as the target attempts evasion and in modelling a financial market. Algorithmic Game Theory is concerned with measuring the computational complexity of various strategies, and ensuring that they are in fact computable. Taken overall, it is a very complete and compelling field of knowledge. The problem is that we game theorists tend to assume that because we found an optimal solution by some advanced analytical technique, it should/will be observed to be the norm “in the wild”; sometimes this turns out to be the case, perhaps because there are cumulative or incremental adaptive paths to arriving at it, oftentimes it does not (typically because it's of the “bolt from the blue” variety). Most game theory research publicised is often of the “so simplified it doesn’t obviously apply” (even the most banal thought-experiment, that of the one-shot Prisoner's Dilemma, involves somebody in police custody essentially making a decision whether to spend their life in jail or being fried on the electric chair, a pretty extreme decision to be making on the spur of the moment and all the assumptions made regarding this scenario, including the iron-clad case against them by the investigators and knowledge that the co-conspirator is also in police custody, are fairly unrealistic). There's a lot of things game theory is very good at and has contributed to. It isn’t represented very well in publicly-accessible material and what is expressed comes across as markedly simplified to the point of irrelevance (which the examples are). But it’s a very powerful tool, actually.

qubex 7y ago

I came to macroeconomics by way of game theory. I'm going to step in because it's actually a far more sophisticated discipline than one would at first imagine given the (admittedly extremely pedagogical, verging on the puerile) examples one is likely to encounter when not studying it in depth (Prisoner's Dilemma, I'm looking at you). Combinatorial Game Theory has been fundamental in addressing “artificial intelligence challenges” such as defeating human Chess grandmasters and more recently, despite the hype about neural network self-training, Go. Bayesian Game Theory deals with how a rational agent starts off in a situation whereby it knows little of what it is competing against and gradually refines that knowledge whilst also getting the best results given the imperfect information available. Differential Game Theory de-discretises time and is useful, for example, in programming both the manoeuvres to be executed by a surface-to-air missile as the target attempts evasion and in modelling a financial market. Algorithmic Game Theory is concerned with measuring the computational complexity of various strategies, and ensuring that they are in fact computable. Taken overall, it is a very complete and compelling field of knowledge. The problem is that we game theorists tend to assume that because we found an optimal solution by some advanced analytical technique, it should/will be observed to be the norm “in the wild”; sometimes this turns out to be the case, perhaps because there are cumulative or incremental adaptive paths to arriving at it, oftentimes it does not (typically because it's of the “bolt from the blue” variety). Most game theory research publicised is often of the “so simplified it doesn’t obviously apply” (even the most banal thought-experiment, that of the one-shot Prisoner's Dilemma, involves somebody in police custody essentially making a decision whether to spend their life in jail or being fried on the electric chair, a pretty extreme decision to be making on the spur of the moment and all the assumptions made regarding this scenario, including the iron-clad case against them by the investigators and knowledge that the co-conspirator is also in police custody, are fairly unrealistic). There's a lot of things game theory is very good at and has contributed to. It isn’t represented very well in publicly-accessible material and what is expressed comes across as markedly simplified to the point of irrelevance (which the examples are). But it’s a very powerful tool, actually.