Game theory can be fun and interesting. It’s central to current mainstream research in industrial organization and corporate strategy. It’s been taught to a generation of MBA students. Unfortunately, according to FastCompany, nobody in business actually uses it. (Hat tip: Nicolai Foss.)
Source link: http://archive.mises.org/3965/game-theory-the-business-tool-that-isnt/
Game Theory: The Business Tool That Isn’t
Previous post: Coffee for Capitalists
Next post: What they won’t tell you about capitalism
{ 19 comments }
Wow, I ought to read Fast Company more often…my opinion of them increased a hundredfold after reading this article on game theory, aka the lamest thing ever.
I believe it is only a matter of time until econometrists and “game theorists” are relegated to circus acts. It will be hilarious.
Speaking of game theory… http://mises.org/daily/1404
A poorly written attack on a very powerful theory that has broad business use … If one wishes to attack game theory, let’s alos attack economics and mathematics as tools that many business people find impractical for the “real” world … as a consultant and business executive I have found game theory to be very worthwhile understanding … This article is a better example of poor journalism
The article linked to by one of the above commentors is a pretty good critique of game-theory. Not only is it impractical in the real world, but it is wrong even on a theoretical level. The only people who act like game theorists think people act are people who are familiar with game-theory.
Well, the game theory is valid mathematically, but its applicability in most practical situations is limited by the inability to compare (or even learn) utilities of different outcomes for the parties concerned. In cases when (dis)utility is predictable (like as in the case of Mutual Assured Destruction) it works nicely.
I actually know an area of commerce where the theory of games is used often and fruitfully… in design of “AI” adversaries in computer games
The following excerpt I just stumbled across (on politicaltheory.info, for the record, which is a great site that often links to mises.org), from the new book The Economic Sociology of Capitalism (Princeton UP) seems to be a pretty good summary of game theory’s conclusions:
“Actors sometimes do not know their own interests, and if they do know them, they may still not know how to realize them. More generally, economic reality is often such that whatever the actors do, they will fail or only partially succeed in realizing their interests.”
Of course, if it’s true that actors don’t know their own interests (because they can only be revealed through game theory’s mathematical models), and can’t fulfill them in any event…well, then, someone else will just have to step in!
Game Theory applies more or less sucessfully to market microstructure. Pretty much any analysis on how change in trading rules will affect market participants carried out using GT apparatus.
IMHO the root cause has to do a lot with the way financial markets regulated: strictly defined product (equity share or commodity future, etc), forced inter-market links (US BBO), strict regulation environment in general (remember Single Stock Futures?).
So they don’t use it, huh? That’s like saying fish don’t use fluid mechanics because they never made a study of it or consciously decided to apply it. There’s a lot of it about, only people who (say) try to externalise their costs do it because they want to, not because of its game-theoretic rationale.
I note here another example of posters talking at cross-purposes.
There can hardly be any doubt that game theory has specific application in the arena of competitive business enterprise. It may also (though tenuously) have value in aiding the efforts of investors to achieve particular goals. It is critical to understand that these activities are, to a large degree (though not entirely), “zero-sum” in nature–that the gains of winners are, by and large, commensurate with the losses of losers. And the fellow who noted that the fish avail themselves of principles of fluidics without intellectual comprehension is dead right too: everyone in the “game” behaves according to his best comprehension of game theory-=-even if he’s never heard of such a thing (and, further, that having more knowledge of such theory may provide advantages).
But those deriding game theory as inapplicable to economics are equally correct: such theory provides no insight of use in that study and those who insist on believing that it does simply delude themselves; worse yet, they put roadblocks in the way of their ever comprehending such study. “Game theory” is a relatively new field: its concepts only date from the ’40s or so and are entirely concerned with the maximization of results in a zero-sum struggle. And here I must point out (and emphasize) that the key core of game theory lies not in quantitative methods per se but in the comprehension of the criticality of “randomized” (indeterminate by and unknown to the protagonist) behavior–a behavior entirely inconsistent with that bearing the description “economic.”
Simon,
If game theory is “a powerful theory that has broad business use” did you supply the author all of the many specific examples of how it’s used in business since he was unable to find any?
“If one wishes to attack game theory, let’s also attack economics and mathmatics as tools that many business people find impractical for the “real” world”
Having worked for more than 35 years in the “real” business world, I can’t imagine a businessman going a single day without using math and economics in his daily routine, unless of course, you’re definition of these terms is quite different from everyone else.
I think one needs to distinguish between game theorists and those who use game theory to justify state intervention (not necessarily overlapping sets). Game theory itself seems to be on firm mathematical and rational ground, however the statists inappropriately use it to justify many governmental interventions on the assumption that participants cannot cooperate to reach optimal outcomes. This assumption seems to be highly irrational to me.
If it had any real application to business, wouldn’t it be called “business theory” and not “game theory”?
I’m only familiar with the famous “prisoner’s dilemma” game, so if someone can point me to another kind of game that makes more sense with what I have to say here, please do.
The “prisoner’s dilemma” game does not map to any economic situation I can think of whatsoever. Any time I act with the intention of realizing an economic gain, I do so through some kind of co-operation. As a consumer, I must co-operate with a vendor to agree on an exchange, and as an employee I have to co-operate with co-workers and clients at least to some extent to remain employed. If I choose to “defect” from one interaction, it is almost always in favor of “co-operating” with someone else (a different vendor or client). I just can’t think of an economic situation that has the slightest resemblance to this game theory.
The other problem I see is that the notion of game theory suggests I play the game once, or some finite amount of times, whereas in real life I have no idea how many times I might play the game. In real life I try to find as many co-operative situations as possible to allow me the greatest number of future options for economic gain, and I very infrequently think of a decision as simply a “one-time” event.
Conclusion: games are lame.
Plowman,
In Google, type in Prisoner’s Dilemma and go to the Stanford link (for me it was the very first hit) which includes a good description of some of the variations on the game, including a discussion of “tragedy of the commons”, “free riders”, and “foul dealers”.
In response to Robert Murphy’s article at http://mises.org/daily/1404
The unexpected execution paradox arises out of a failure to accurately define what is meant by don’t know and an implicit assumption that it means that the execution is equally likely to occur on any day of the week, and that if it has not occurred on previous days, that it is equally likely to occur on any of the remaining days. Suppose that we define the day of execution as being unknown as being equivalent to not being able to assign the day of execution to a particular day with exact certainty (probability = 1). Then, framing the problem in this way, on Friday afternoon, the probability of execution on Saturday is 1.
However, what has been omitted is the probability of arriving in this situation. There is only a 1 in 7 probability of this situation arising. So, assuming that this is how every execution day is determined, 6 out of 7 times the prisoner will not know the day until the morning of execution. Or equivalently, if 7 prisoners are executed and the warden assigns each one a day, only the last one to be executed will actually know before.
What has been overlooked is that there is no requirement that the days be assigned with equal probability. Suppose that each day, the prisoner is executed with probability 99 out of 100. Clearly, each day, the prisoner knows with 99% probability that he is going to be executed. However, he does not know for sure. So while it is true that if he is still alive on Friday afternoon, he will know for sure that he will be executed on Saturday, the chances of him arriving in this situation is 1 in 10^12.
To make the situation a little less contrived, let us assume that no executions are carried out on Saturday. If we imagine that the days are assigned by drawing one of seven cards from a hat, in the event of Saturday being drawn, the card is put back in the hat, a new card is drawn, and the execution is carried out the following week, unless Saturday is drawn again, in which case the execution is postponed for another week. Now suppose that the judge has indicated that the execution must be carried out within 12 weeks, the chances of arriving at the last Thursday and knowing for sure is less than 1 in a billion. This is very much different from the 1 in 72 chance of knowing for sure if all 72 days in the 12 week period were equally probable. From a strict mathematical point of view the judges instruction can not be carried out. However, we can make the probability of the judge’s order being violated arbitrarily small.
Incidentally, the notion that if the probability of X is unknown is equivalent to assuming that it is uniform is not mathematically consistent. This is known as Boole’s Paradox.
For this reason, the use of Bayes Theorem to calculate conditional probabilities in the case of an unknown prior distribution was, for a long time, considered to be invalid. An easy counterexample is as follows: If we do not know the distribution of X, then we don’t know the distribution of X-squared. However if X is uniformly distributed then X-squared can not be uniformly distributed and vice versa. This situation is not resolved by assuming a normal distribution.
Years ago I taught a brief course in game theory. One of the things I mentioned was the uselessness of protectionism, using the “prisoner’s dilemma” argument.
See for example
http://internationalecon.com/v1.0/ch110/110c010.html
Wow! I can’t believe how ignorant you all are! The fellow that made the comment about the fish was nearly correct. I am a game theorist. And yes I have worked for a consulting company using this theory. Would you like examples. Ok. I have colleaugues that work for ebay. yes game theorists. I also have colleagues that work large energy companies…also game theorists. Microsoft and Intel both hire game theorists as well.
The point is Game Theory is just a representation of strategic interaction. A better point would be to ask the businesses if they have had any strategic interactions. Of course they have!
The advantage goes to those businesses that understand the exact nature of their strategin interaction. A great example of this is contracting. Ask these businesses if they have had to haggle over their contracts. What about optimal production. Do firms just produce whatever the feel is right. No. They hire consultants (like me) to estimate their market demand and (given their degree of market concentration – competition) to tell them how much to produce.
If you want to nail down game theory as an empirically irrelevant science than you might as well throw out the baby with the bathwater — i.e. forget about strategic interaction.
Jon,
If I may jump in here, in an effort to clear things up simply, than please consider the following:
Honest people interact openly, transparently, based upon a commonly shared code of conduct. They simply reject all obvious efforts to enforce falsehood. The idea is to throw out the baby (dishonest people) with the bath water (falsehood).
Dupes are the most obvious babies operating their game of robotic ignorance. ‘Stupid is what stupid does’. Their game is very obvious.
Hired intellectuals are more difficult to recognize. The company human resources ‘expert’ is a particularly illustrative example.
If said ‘expert’ is a dupe, then, the language used by the idiot is transparent. It goes like this: “The reality is:” followed by “You must tell the OSHA, MSHA, AQMD, EPA inspector everything because…” followed by obvious scenarios that place the employee at the mercy of open ended liabilities. An honest person may use the strategy of interaction by raising their hand and asking a question like: “Who has my back if they take me to court?” Depending upon the answer the honest persons in the group can then gain more insight in the actual “reality”.
If said ‘expert’ is not a dupe himself, then, the language used will be much less open and transparent. Free exchange will be thrown out as an obvious liability; an empirically irrelevant science, or, some such other fancy term.
I like your articles. I also write on business theory, news, and ideas. Good Blog Thank You.
http://www.thenewbusinessworld.blogspot.com
Comments on this entry are closed.