What Computer Science Can Teach Economics

eldavojohn writes "A new award-winning thesis from an MIT computer science assistant professor showed that the Nash equilibrium of complex games (like the economy or poker) belong to problems with non-deterministic polynomial (NP) complexity (more specifically PPAD complexity, a subset of TFNP problems which is a subset of FNP problems which is a subset of NP problems). More importantly there should be a single solution for one problem that can be adapted to fit all the other problems. Meaning if you can generalize the solution to poker, you have the ability to discover the Nash equilibrium of the economy. Some computer scientists are calling this the biggest development in game theory in a decade."

Comments

Hayek

[homer_s]

By showing that some common game-theoretical problems are so hard that they'd take the lifetime of the universe to solve, Daskalakis is suggesting that they can't accurately represent what happens in the real world.

Hayek showed that about 50 years ago:
"The curious task of economics is to demonstrate to men how little they really know about what they imagine they can design." (The Fatal Conceit, p. 76)

Unfortunately, there is a lot of designing going on right now.

Re:Hayek

[zach_the_lizard]

And his teacher, Mises, before him in his work Human Action devoted an entire section of that massive tome to just this very topic: that humans are not equations.

From said tome:

No laboratory experiments can be performed with regard to human action. We are never in a position to observe the change in one element only, all other conditions of the event remaining unchanged. Historical experience as an experience of complex phenomena does not provide us with facts in the sense in which the natural sciences employ this term to signify isolated events tested in experiments. The information conveyed by historical experience cannot be used as building material for the construction of theories and the prediction of future events. Every historical experience is open to various interpretations, and is in fact interpreted in different ways.

The postulates of positivism and kindred schools of metaphysics are therefore illusory. It is impossible to reform the sciences of human action according to the pattern of physics and the other natural sciences. There is no means to establish an a posteriori theory of human conduct and social events. History can neither prove nor disprove any general statement in the manner in which the natural sciences accept or reject a hypothesis on the ground of laboratory experiments. Neither experimental verification nor experimental falsification of a general proposition is possible in its field.

Re:Hayek

[khallow]

No laboratory experiments can be performed with regard to human action.

One of the most profoundly stupid statements ever uttered by an economist. Sure you can't stick the global economy in a beaker and have controls and the other paraphernalia of controlled lab tests, the highest standard of science. But you can experiment with human action at the individual or small group in a controlled lab. It's routinely done these days. There is such a thing as experimental verification and falsification.

another intersection of CS and econ

[WhiteDragon]

Here's a proof that detecting "toxic assets" is impossible (or at least NP)

Its easy!

[Monkeedude1212]

Meaning if you can generalize the solution to poker, you have the ability to discover the Nash equilibrium of the economy

The general solution to poker is to end the game with everyone elses money to make yourself richer. Some people have already applied this strategy to the economy.

Obligatory

[Yvan256]

Economics involves people. So...

"To summarize the summary of the summary: people are a problem." - Douglas Adams

Re:The problem is not an efficient algorithm

[zach_the_lizard]

There are entire schools of economics that criticize the mainstream schools using this very line of reasoning. IIRC, the Austrian school economists (Mises, Menger, et. al) never use any sort of math at all, except in trying to determine things such as the rate of inflation. There are others, too, but their names escape me at the moment.

Re:The problem is not an efficient algorithm

[NeutronCowboy]

I was about to say the same thing. Unlike poker, the rules of the games are altered based on the current knowledge about the state of the game. This means that as soon as someone proclaims "We know the rules of Economics!", someone else is going to look at those rules and either game them to their benefit, or rewrite them to better suit their own purpose.

Computer Scientists - and Economists - have a habit of assuming that they just need to find the proper model for human behavior, and all the problems will be solved. That's because that's how it works in a science: you assume the rules don't change in an arbitrary fashion. Humans, however, do. This makes any prediction of human behavior a statistical undertaking at best. Your success will be measured by how much better you compared to a random decision making process. At worst, the statistical anomaly completely wrecks your model - see the Black Swan Theory in Economics.

Re:The problem is not an efficient algorithm

[gerddie]

Well, I'm not surprised there is such school. My impression is, that economists in general don't have a good grasp of math, specifically, they don't seem to understand the exponential function, otherwise they would not speak of "growth" all the time.
I'm not saying one should not take human behavior into account, but at least they should get the boundary conditions right, and one of those is that our resources are limited.

Only works with real money

[Chemisor]

Once you factor debt and fractional reserves into the picture, the game changes quite a bit. The current crisis is that the players bet WAY more than they had, and they are all afraid to call, since they secretly know that EVERYBODY is bluffing. So the game (and the stock market) keeps going up as the players trying to outbluff each other with "I'll see your billion and raise you three more". And it will keep going up until somebody has to actually put something of value in the pot.

Nice setup

[istartedi]

What can CS teach ECON?

How to crash routinely and have people shrug it off as normal.

Re:The problem is not an efficient algorithm

[Red Flayer]

I'm not saying one should not take human behavior into account, but at least they should get the boundary conditions right, and one of those is that our resources are limited.

That does not mean that additional wealth cannot be created without infusion of additional resources.

I know it's counterintuitive for most people with a "hard science" background... I struggled with it as an undergrad. But economics is not a zero-sum game. I give you $150 and you give me an hour of labor. We've both benefited by the trade. If we are really acting freely, we've both benefited (or we wouldn't have engaged in the trade), so we are both wealthier than we were before. This is the fundamental basis of perpetual economic growth... given a free market* in which to pursue trades, wealth increases as trades are made.

* Free as in some-kind-of-approximation-of-an-ideal-free-market, not free as in no-legal-restrictions-on-activity.

Re:The problem is not an efficient algorithm

[Marcika]

Well, I'm not surprised there is such school. My impression is, that economists in general don't have a good grasp of math, specifically, they don't seem to understand the exponential function, otherwise they would not speak of "growth" all the time. I'm not saying one should not take human behavior into account, but at least they should get the boundary conditions right, and one of those is that our resources are limited.

Your impression is wrong. Every economist knows about Thomas Robert Malthus and Malthusian economics -- for the pre-industrial era his model best explains demographics and the limits of growth. It only so happened that just after he published his thoughts, the industrial revolution happened and technological progress pushed the boundaries of growth further and further - in an exponential manner.

Would you dare to make an exact forecast where the limits of growth lie? Limited by fossil fuels? Or a single planet's worth of solar energy? Maybe a Dyson sphere's worth of solar energy? Technological progress moves the goalposts rapidly enough that you have to assume exponential growth punctuated by occasional catastrophes - at least for the next 50 years.

Bullshit!

[Anonymous Coward]

I give you $150 and you give me an hour of labor. We've both benefited by the trade.

This is the fundamental basis of perpetual economic growth... given a free market* in which to pursue trades, wealth increases as trades are made.

This argument echoes the exact same stupidity of the "perpetual growth" nuts that got us into this economic mess in the first place. You believe that infinite trades are possible, and that resource limitations don't apply. But even in your own example, you're talking about trading one limited resource (labor) for another (money). And yes, money is a limited resource - you can print all the money you want, but since doing so doesn't increase the amount of actual value that that money represents, all you're really doing is devaluing the existing money supply in order to redistribute the underlying value (i.e., stealing a little bit of value from everybody who's currently holding any of the existing bills, and giving the loot to someone else - usually a central bank).

Perpetual growth is nothing more than an illusion shared amongst fools. Value doesn't magically spring into existence by the mere act of trading something back and forth. Value can only be created by consuming resources. Whether that resource is energy, or some natural resource such as coal or iron, or human labor, etc, there is only a finite amount of that resource. Furthermore, many of these resource limits are things we are either already bumping into, or things that we will bump into in the foreseeable future, such as in the case of the various natural resources we've come to rely on.