The only winning move is not to play, or to watch these first.
January 13, 2012 1:04 PM Subscribe
Game Theory 101 has a selection of video and text lectures covering such topics as How to Fly on an Airplane with an Empty Seat Next to You for Free, Why You Should NOT Maximize Your Score in Words With Friends and How to Catch a Ball at a Baseball Game. It's not all light and fluffy, though. Some other topics include Why You Should Support International Aid, How the United States Debt Crisis Will End and Why the Intervention in Libya Was a Bargain. If you're new to game theory, start with The Prisoner's Dilemma.
The empty seat video amounts to "Select the 2nd worst seat on the plane, in the rear row (typically with no recline), next to the bathroom. Since no one will want the very worst seat, assuming the flight doesn't sell out, you'll have an empty seat next to you. On the other hand, if it does sell out, you're still next to someone and you're in the 2nd worst seat on the plane."
Since the airlines try very hard to make sure that the planes are full, I think I'd rather optimize for the worst case and pick a better seat, even if it's more likely that someone will sit next to me.
posted by jedicus at 1:11 PM on January 13, 2012 [3 favorites]
Since the airlines try very hard to make sure that the planes are full, I think I'd rather optimize for the worst case and pick a better seat, even if it's more likely that someone will sit next to me.
posted by jedicus at 1:11 PM on January 13, 2012 [3 favorites]
Why You Should NOT Maximize Your Score in Words With Friends
I was so sure this would be: So people will want to play with you again. Sigh.
posted by yellowbinder at 1:16 PM on January 13, 2012 [4 favorites]
I was so sure this would be: So people will want to play with you again. Sigh.
posted by yellowbinder at 1:16 PM on January 13, 2012 [4 favorites]
How to fly on an airplane with an empty seat next to you: be brown. Works every time. On both sides even! It's awesome, you guys are missing out.
posted by Errant at 1:21 PM on January 13, 2012 [8 favorites]
posted by Errant at 1:21 PM on January 13, 2012 [8 favorites]
be brown
Most of the airlines I fly simply refuse to provide me this information about my fellow passengers when I'm booking my seat.
posted by found missing at 1:30 PM on January 13, 2012 [2 favorites]
Most of the airlines I fly simply refuse to provide me this information about my fellow passengers when I'm booking my seat.
posted by found missing at 1:30 PM on January 13, 2012 [2 favorites]
Place a single, clean square of toilet paper on the seat next to you. Nobody will sit there.
posted by Faint of Butt at 1:33 PM on January 13, 2012 [6 favorites]
posted by Faint of Butt at 1:33 PM on January 13, 2012 [6 favorites]
And then have the flight attendants, who usually use those seats when they want to relax, give you grief the entire flight.
posted by zippy at 1:34 PM on January 13, 2012
posted by zippy at 1:34 PM on January 13, 2012
Be too fat for one seat. The guy next to you will get frustrated and go find an empty seat somewhere else.
posted by pracowity at 1:48 PM on January 13, 2012
posted by pracowity at 1:48 PM on January 13, 2012
Have oozing sores and a deep troubling cough.
posted by found missing at 1:52 PM on January 13, 2012 [3 favorites]
posted by found missing at 1:52 PM on January 13, 2012 [3 favorites]
The plane seat one neglects the fact that premium coach seats are often blocked for people with higher status and shown as sold. If you log in to AA.com as Platinum, you get a different view of the available seats than the commoners. It also neglects that people with higher status will have a soft hold on the seat next to them. So the last row middle will often times get filled before the second row middle seat. Also, the people with status will request upgrades to first/business and those nice window/aisle seats will empty a few hours close to the departure date as the premium fliers are bumped up to the front of the plane.
If you want to have a good seat without the middle, you just need to fly a lot. For me the best perk of premium status was early access to better seats in coach. Requesting the last aisle seat means you have to deal with the asses of passengers queuing to the lav. And heaven help you if the toilet is broken and the last few rows smell like a porta-john. Oh, and on the MD80 there's a jet engine right outside your window. I'd rather volunteer for a middle seat in the front than a guaranteed seat in the back.
posted by birdherder at 1:54 PM on January 13, 2012 [2 favorites]
If you want to have a good seat without the middle, you just need to fly a lot. For me the best perk of premium status was early access to better seats in coach. Requesting the last aisle seat means you have to deal with the asses of passengers queuing to the lav. And heaven help you if the toilet is broken and the last few rows smell like a porta-john. Oh, and on the MD80 there's a jet engine right outside your window. I'd rather volunteer for a middle seat in the front than a guaranteed seat in the back.
posted by birdherder at 1:54 PM on January 13, 2012 [2 favorites]
Buy both seats.
posted by eustacescrubb at 1:54 PM on January 13, 2012 [1 favorite]
posted by eustacescrubb at 1:54 PM on January 13, 2012 [1 favorite]
It occurs to me that beanplated sitting arrangements etc may make living with a game theorist a bit like living with Sheldon of Big Bang Theory. :)
posted by -harlequin- at 1:56 PM on January 13, 2012 [1 favorite]
posted by -harlequin- at 1:56 PM on January 13, 2012 [1 favorite]
Select the 2nd worst seat safest on the plane, in the rear row (typically with no recline), next to the bathroom.
In the event of a crash, the entire plane is your airbag.
posted by CynicalKnight at 1:59 PM on January 13, 2012
In the event of a crash, the entire plane is your airbag.
posted by CynicalKnight at 1:59 PM on January 13, 2012
It occurs to me that beanplated sitting arrangements etc may make living with a game theorist a bit like living with Sheldon of Big Bang Theory. :)
Deeply contrived and unfunny? *canned laugh track*
posted by codacorolla at 2:23 PM on January 13, 2012 [4 favorites]
Deeply contrived and unfunny? *canned laugh track*
posted by codacorolla at 2:23 PM on January 13, 2012 [4 favorites]
Also, there are just far too many other factors to take into consideration: where the toilet traffic is, where the screaming newborns are likely to be accommodated, where there's an "entertainment box" blocking half your legroom. Hence SeatExpert.com.
posted by oliverburkeman at 2:29 PM on January 13, 2012
posted by oliverburkeman at 2:29 PM on January 13, 2012
Regarding the airline seats video, I have also noticed that the last few times I have purchased tickets, when I was given the option to select seats all of the available seats were only available for a premium. And, I was not shown the seat I was currently assigned. It seems that this is their new evil trick. I have no idea where I will be sitting, so I could already have a decent seat, and there are no available choices to even pick without paying extra. It seems like the airlines have moved on to Game Theory 102, and we are the prisoners.
posted by This_Will_Be_Good at 2:34 PM on January 13, 2012 [3 favorites]
posted by This_Will_Be_Good at 2:34 PM on January 13, 2012 [3 favorites]
How to fly on an airplane with an empty seat next to you: be brown
Haven't tried it. HOWEVER! I recently flew with my 10 mo. old son, who had an ear infection at the time... you can see where this is going. The guy behind me DEMANDED! to be reseated. On a full flight. Don't know how they accomodated that one, but they did somehow. And lo, I could recline to my heart's content.
Still had to have a baby with an ear infection trying to gouge my eyes out for eight hours. But I could recline!
posted by sonika at 2:37 PM on January 13, 2012
Haven't tried it. HOWEVER! I recently flew with my 10 mo. old son, who had an ear infection at the time... you can see where this is going. The guy behind me DEMANDED! to be reseated. On a full flight. Don't know how they accomodated that one, but they did somehow. And lo, I could recline to my heart's content.
Still had to have a baby with an ear infection trying to gouge my eyes out for eight hours. But I could recline!
posted by sonika at 2:37 PM on January 13, 2012
A 10 mo. old baby, with an ear infection, on a plane.
...you can see where this is going
posted by clearly at 2:58 PM on January 13, 2012
...you can see where this is going
posted by clearly at 2:58 PM on January 13, 2012
Most of the airlines I fly simply refuse to provide me this information about my fellow passengers when I'm booking my seat.
1: Most of them?
2: It's cool, people usually call an audible once they see any other seat available after takeoff. So you still have options. Unless the flight is full. Then there's just the uncomfortable squirming. Believe me, I don't like it either, I was really looking forward to having a drink on every tray table in this aisle. Now I have to fly someplace in pressurized silence, just like the rest of you. I hate not being special.
posted by Errant at 3:15 PM on January 13, 2012
1: Most of them?
2: It's cool, people usually call an audible once they see any other seat available after takeoff. So you still have options. Unless the flight is full. Then there's just the uncomfortable squirming. Believe me, I don't like it either, I was really looking forward to having a drink on every tray table in this aisle. Now I have to fly someplace in pressurized silence, just like the rest of you. I hate not being special.
posted by Errant at 3:15 PM on January 13, 2012
Wow, always exciting to see a friend's work on the blue. I've forwarded him this thread. :)
posted by raihan_ at 3:32 PM on January 13, 2012
posted by raihan_ at 3:32 PM on January 13, 2012
What a weirdo I am. That's my favorite seat on the plane. Nobody behind you kicking your back or digging through the seat pocket, easy access to the toilet, view of *everything*, chat with the flight attendants and best service. And the seat beside is usually empty!
posted by iamkimiam at 3:51 PM on January 13, 2012
posted by iamkimiam at 3:51 PM on January 13, 2012
These are pretty good. But I just watched the Foreign Aid one, and I agere with both his premises and conclusions, and I think I agree with how he got there, but for a video meant to explain the scenario to laymen, he kind of omits a big part of it.
Namely, he never explains what ∂ is or means.
I mean, we get to the point where ∂=2/3 for the U.S. and 1/2 for the E.C. (I think) but with no application of what that variable means or how to compare the two or anything.
Additionally, he proposes point values for the hypothetical game and then assumes those point-values are both constant and relate to something tangible which makes those point values have a real-world meaning.
Again, I agree with him and I think for the reasons he explains, but it feels like he's missing a few step, especially when starting from a highly abstract hypothetical and then ending with, "so, as you can see from the spherical cows version, not following this in the real world of diplomacy is monumentally stupid."
I hope he pops in here for a slightly more in-depth explanation of those issues.
posted by Navelgazer at 3:56 PM on January 13, 2012
Namely, he never explains what ∂ is or means.
I mean, we get to the point where ∂=2/3 for the U.S. and 1/2 for the E.C. (I think) but with no application of what that variable means or how to compare the two or anything.
Additionally, he proposes point values for the hypothetical game and then assumes those point-values are both constant and relate to something tangible which makes those point values have a real-world meaning.
Again, I agree with him and I think for the reasons he explains, but it feels like he's missing a few step, especially when starting from a highly abstract hypothetical and then ending with, "so, as you can see from the spherical cows version, not following this in the real world of diplomacy is monumentally stupid."
I hope he pops in here for a slightly more in-depth explanation of those issues.
posted by Navelgazer at 3:56 PM on January 13, 2012
Huh, the How will the United States Debt Crisis end one was completely wrong. Not wrong in the sense of "I disagree", but rather -- the video was actually about what would happen with the whole "will we increase the debt ceiling, or default". In the end republicans actually got everything they wanted, Getting big spending cuts and no new taxes (IIRC)
Apparently his 'matrix' for what Obama wanted was actually incorrect somehow, and taking the '14th amendment option' was actually ranked lower then rolling over for the republicans.
(He also assumed that everyone had 'perfect information' -- but that wasn't really true. Each side didn't know the others preferences, so the delay was an effort to discover what those preferences were)
posted by delmoi at 4:49 PM on January 13, 2012
Apparently his 'matrix' for what Obama wanted was actually incorrect somehow, and taking the '14th amendment option' was actually ranked lower then rolling over for the republicans.
(He also assumed that everyone had 'perfect information' -- but that wasn't really true. Each side didn't know the others preferences, so the delay was an effort to discover what those preferences were)
posted by delmoi at 4:49 PM on January 13, 2012
delmoi- I think the mistake was that assuming Obama would consider the 14th amendment choice, and if I remember correctly, he wasn't considering that.
posted by gjc at 7:49 PM on January 13, 2012
posted by gjc at 7:49 PM on January 13, 2012
I've always been irritated by claims that the Prisoner's Dilemma leads rational players to the defect-defect outcome, because the following line of reasoning has always struck me as completely sound:
The cops have no reason to offer us different deals, so I can assume that my partner in crime is facing the same choice as I am. Since each of us knows that the other is smart enough not to get involved in a land war in Asia and neither of us is Sicilian and death is not on the line, we know we're both going to make the same choice. Which means that the options on the table are limited to cooperate-cooperate and defect-defect, of which cooperate-cooperate is clearly the better.
In fact if one of us were Sicilian, getting this wrong might well put death on the line.
posted by flabdablet at 1:02 AM on January 14, 2012 [1 favorite]
The cops have no reason to offer us different deals, so I can assume that my partner in crime is facing the same choice as I am. Since each of us knows that the other is smart enough not to get involved in a land war in Asia and neither of us is Sicilian and death is not on the line, we know we're both going to make the same choice. Which means that the options on the table are limited to cooperate-cooperate and defect-defect, of which cooperate-cooperate is clearly the better.
In fact if one of us were Sicilian, getting this wrong might well put death on the line.
posted by flabdablet at 1:02 AM on January 14, 2012 [1 favorite]
What a weirdo I am. That's my favorite seat on the plane. Nobody behind you kicking your back or digging through the seat pocket, easy access to the toilet, view of *everything*, chat with the flight attendants and best service. And the seat beside is usually empty!
No no no no no! Those seats are terrible. None of you guys should try and book them, no sir!
posted by Infinite Jest at 1:49 AM on January 14, 2012 [1 favorite]
No no no no no! Those seats are terrible. None of you guys should try and book them, no sir!
posted by Infinite Jest at 1:49 AM on January 14, 2012 [1 favorite]
I long ago decided I would rather go to the back of the plane and have a space next to me. The downside to being at the back is really a delay in getting off the plane, but in my experience many airports will disembark from both front and back and there is no absolute corollary between boarding at the front and exiting from both front and back, ie it is sometimes possible to board only from the front and exit from both ends. And the reverse may be true on the return journey, something which should inform any decision on the gaming process.
posted by biffa at 6:48 AM on January 14, 2012
posted by biffa at 6:48 AM on January 14, 2012
@yellowbinder
Why You Should NOT Maximize Your Score in Words With Friends
I was so sure this would be: So people will want to play with you again.
Just so - especially in a 3 or 4 player game. Someone who constantly leaves high-score openings for the next player is deeply infuriating to the players 'downstream' of that interaction.
posted by raygirvan at 12:20 PM on January 14, 2012
Why You Should NOT Maximize Your Score in Words With Friends
I was so sure this would be: So people will want to play with you again.
Just so - especially in a 3 or 4 player game. Someone who constantly leaves high-score openings for the next player is deeply infuriating to the players 'downstream' of that interaction.
posted by raygirvan at 12:20 PM on January 14, 2012
Derren Brown says if you are sitting next to an empty seat on the train and you want it to stay that way, don't put your luggage on it or make it seem unaccessible. Instead, when anyone approaches to take the seat, establish eye contact, smile warmly and pat the seat.
posted by devious truculent and unreliable at 7:54 AM on January 16, 2012
posted by devious truculent and unreliable at 7:54 AM on January 16, 2012
Lol. It's true, look excited that you're going to get company on a train. Works like a charm. Unless of course it's a hot girl and you would like her to sit next to you. Then you're stuck... Maybe put your bag on the seat then?
posted by defcom1 at 11:00 AM on January 16, 2012
posted by defcom1 at 11:00 AM on January 16, 2012
I've always been irritated by claims that the Prisoner's Dilemma leads rational players to the defect-defect outcome,
It does and has to, given the payoffs.
because the following line of reasoning has always struck me as completely sound:
The cops have no reason to offer us different deals, so I can assume that my partner in crime is facing the same choice as I am.
This is explicitly part of the structure.
Since each of us knows that the other is smart enough not to get involved in a land war in Asia and neither of us is Sicilian and death is not on the line, we know we're both going to make the same choice.
This is wrong, in the sense of being incorrect, because...
Which means that the options on the table are limited to cooperate-cooperate and defect-defect, of which cooperate-cooperate is clearly the better.
If you believe that your partner is going to cooperate, you are better off defecting, so you should defect.
And, in fact, if you believe that your partner has defected, you are better off defecting too.
It's symmetrical, so flip it: no matter what you do, your partner is better off defecting against you. What do you think he's going to do given that the payoff structure by definition encompasses the totality of the consequences of his choice?
In fact if one of us were Sicilian, getting this wrong might well put death on the line.
If you're playing a game with different payoffs (ie your outcome for defecting is "you get whacked"), then you aren't playing a prisoners' dilemma any more. You're playing a game with a superficial resemblance in the storytelling, but the game itself (which in a strategic form game is just the set of strategies and payoffs) is just not the PD.
I think what you mean is that the PD game is not a full and rich description of the ways in which criminals and police might actually interact in a setting like that, set within some larger context. Well, yeah. The cop setting is just a story-telling standin for any of a vast group of situations where individual rationality leads to pareto-inferior outcomes. On the one hand, it was never really intended to be a real analysis of that situation. On the other hand, the whole point of modeling is to strip away most of the details until you're left with the strategic heart of whatever's going on. How do you know you did that right? Not by looking at the details you left our or the appeal of the assumptions you're making. You know you did it right (enough for now) if your model predicts well, or better than existing ones.
Or, sometimes, even models that predict spectacularly badly can be useful. Most commonly when you take a situation, lay out some commonsense and appealing assumptions, and show that those assumptions lead to conclusions that strongly diverge from observed reality. Now you know that those commonsense and appealing assumptions were probably wrong, and that the truth is something more complex. (see rational choice and voter turnout)
posted by ROU_Xenophobe at 11:50 AM on January 16, 2012
It does and has to, given the payoffs.
because the following line of reasoning has always struck me as completely sound:
The cops have no reason to offer us different deals, so I can assume that my partner in crime is facing the same choice as I am.
This is explicitly part of the structure.
Since each of us knows that the other is smart enough not to get involved in a land war in Asia and neither of us is Sicilian and death is not on the line, we know we're both going to make the same choice.
This is wrong, in the sense of being incorrect, because...
Which means that the options on the table are limited to cooperate-cooperate and defect-defect, of which cooperate-cooperate is clearly the better.
If you believe that your partner is going to cooperate, you are better off defecting, so you should defect.
And, in fact, if you believe that your partner has defected, you are better off defecting too.
It's symmetrical, so flip it: no matter what you do, your partner is better off defecting against you. What do you think he's going to do given that the payoff structure by definition encompasses the totality of the consequences of his choice?
In fact if one of us were Sicilian, getting this wrong might well put death on the line.
If you're playing a game with different payoffs (ie your outcome for defecting is "you get whacked"), then you aren't playing a prisoners' dilemma any more. You're playing a game with a superficial resemblance in the storytelling, but the game itself (which in a strategic form game is just the set of strategies and payoffs) is just not the PD.
I think what you mean is that the PD game is not a full and rich description of the ways in which criminals and police might actually interact in a setting like that, set within some larger context. Well, yeah. The cop setting is just a story-telling standin for any of a vast group of situations where individual rationality leads to pareto-inferior outcomes. On the one hand, it was never really intended to be a real analysis of that situation. On the other hand, the whole point of modeling is to strip away most of the details until you're left with the strategic heart of whatever's going on. How do you know you did that right? Not by looking at the details you left our or the appeal of the assumptions you're making. You know you did it right (enough for now) if your model predicts well, or better than existing ones.
Or, sometimes, even models that predict spectacularly badly can be useful. Most commonly when you take a situation, lay out some commonsense and appealing assumptions, and show that those assumptions lead to conclusions that strongly diverge from observed reality. Now you know that those commonsense and appealing assumptions were probably wrong, and that the truth is something more complex. (see rational choice and voter turnout)
posted by ROU_Xenophobe at 11:50 AM on January 16, 2012
If you believe that your partner is going to cooperate, you are better off defecting, so you should defect ... It's symmetrical, so flip it: no matter what you do, your partner is better off defecting against you.
Provided you're taking your partner to be an individual automaton with no information about you, that's true.
But if you know that your partner's reasoning ability is at least as good as yours, and you both know that the other has also been exposed to the game-theoretical PD, then you can safely conclude that if either of you allows co-de or de-co as a possibility it won't stand, because the only reasonable option to choose given that possibility is de-de. Which means that co-de or de-co are illusory outcomes; the only options actually available to you are co-co or de-de, of which co-co is the better.
posted by flabdablet at 3:55 PM on January 16, 2012
Provided you're taking your partner to be an individual automaton with no information about you, that's true.
But if you know that your partner's reasoning ability is at least as good as yours, and you both know that the other has also been exposed to the game-theoretical PD, then you can safely conclude that if either of you allows co-de or de-co as a possibility it won't stand, because the only reasonable option to choose given that possibility is de-de. Which means that co-de or de-co are illusory outcomes; the only options actually available to you are co-co or de-de, of which co-co is the better.
posted by flabdablet at 3:55 PM on January 16, 2012
If you believe that your partner is going to cooperate, you are better off defecting, so you should defect.
And, in fact, if you believe that your partner has defected, you are better off defecting too.
But if you believe that your partner is going to make the same choice as you do because they understand as well as you do that your circumstances are symmetric, you are better off cooperating.
The Cold War doctrine of Mutually Assured Destruction strikes me as a real-world example of this dynamic.
posted by flabdablet at 4:04 PM on January 16, 2012
And, in fact, if you believe that your partner has defected, you are better off defecting too.
But if you believe that your partner is going to make the same choice as you do because they understand as well as you do that your circumstances are symmetric, you are better off cooperating.
The Cold War doctrine of Mutually Assured Destruction strikes me as a real-world example of this dynamic.
posted by flabdablet at 4:04 PM on January 16, 2012
Provided you're taking your partner to be an individual automaton with no information about you, that's true.
Your partner has all the necessary information about you: your payoff structure.
But if you know that your partner's reasoning ability is at least as good as yours
Rationality assumes something like this.
you both know that the other has also been exposed to the game-theoretical PD
You know this under the usual assumption that everybody knows that everybody knows ... that everybody knows the strategies available and the payoff structure.
you can safely conclude that if either of you allows co-de or de-co as a possibility it won't stand
This doesn't make sense. It's not for either of the players to allow or deny, and there's no standing or not standing. It's right there in the structure of the game: if I defect and you cooperate, we arrive at the relevant payoffs. If you're cooperating, you can't prevent me from defecting -- if you could, that would be reflected in the strategies.
But if you believe that your partner is going to make the same choice as you do because they understand as well as you do that your circumstances are symmetric, you are better off cooperating.
No, you are always better off defecting, and so is your partner. In the jargon, it's a dominant strategy for both players. You don't need any insight into your partner's knowledge or state of mind. All you need to do is look at your partner's payoff structure.
The Cold War doctrine of Mutually Assured Destruction strikes me as a real-world example of this dynamic.
The cold war doctrine of MAD is not a PD game. It's easily modeled as a deterrence game. The neat thing about a deterrence game like the cold war is that the actual outcome requires threats that would be irrational to follow through with (jargon: is not subgame perfect).
posted by ROU_Xenophobe at 8:16 PM on January 16, 2012
Your partner has all the necessary information about you: your payoff structure.
But if you know that your partner's reasoning ability is at least as good as yours
Rationality assumes something like this.
you both know that the other has also been exposed to the game-theoretical PD
You know this under the usual assumption that everybody knows that everybody knows ... that everybody knows the strategies available and the payoff structure.
you can safely conclude that if either of you allows co-de or de-co as a possibility it won't stand
This doesn't make sense. It's not for either of the players to allow or deny, and there's no standing or not standing. It's right there in the structure of the game: if I defect and you cooperate, we arrive at the relevant payoffs. If you're cooperating, you can't prevent me from defecting -- if you could, that would be reflected in the strategies.
But if you believe that your partner is going to make the same choice as you do because they understand as well as you do that your circumstances are symmetric, you are better off cooperating.
No, you are always better off defecting, and so is your partner. In the jargon, it's a dominant strategy for both players. You don't need any insight into your partner's knowledge or state of mind. All you need to do is look at your partner's payoff structure.
The Cold War doctrine of Mutually Assured Destruction strikes me as a real-world example of this dynamic.
The cold war doctrine of MAD is not a PD game. It's easily modeled as a deterrence game. The neat thing about a deterrence game like the cold war is that the actual outcome requires threats that would be irrational to follow through with (jargon: is not subgame perfect).
posted by ROU_Xenophobe at 8:16 PM on January 16, 2012
No, you are always better off defecting, and so is your partner.
By what logic is a five year stretch better than a one year stretch?
Let me put it this way:
0. My partner is going to reason the same way I do, because my partner is facing the same circumstances that I am.
1. I can either allow myself the possibility of defecting, or not. My partner, likewise.
2. If I am about to reason my way into allowing that possibility, then so will my partner. And if defection is an option, then it is clearly also the right option (per the payoff rules). Which means that we will both defect, which means we'll both get five year stretches.
3. On the other hand, if I am about to reason my way out of allowing that possibility, then so will my partner. And if defection is not an option for either of us, then neither will do it and we will both get one year stretches.
4. A one year stretch is better than a five year stretch.
5. Therefore I will not allow myself the possibility of defecting, and neither will my partner.
posted by flabdablet at 8:18 AM on January 17, 2012
By what logic is a five year stretch better than a one year stretch?
Let me put it this way:
0. My partner is going to reason the same way I do, because my partner is facing the same circumstances that I am.
1. I can either allow myself the possibility of defecting, or not. My partner, likewise.
2. If I am about to reason my way into allowing that possibility, then so will my partner. And if defection is an option, then it is clearly also the right option (per the payoff rules). Which means that we will both defect, which means we'll both get five year stretches.
3. On the other hand, if I am about to reason my way out of allowing that possibility, then so will my partner. And if defection is not an option for either of us, then neither will do it and we will both get one year stretches.
4. A one year stretch is better than a five year stretch.
5. Therefore I will not allow myself the possibility of defecting, and neither will my partner.
posted by flabdablet at 8:18 AM on January 17, 2012
By what logic is a five year stretch better than a one year stretch?
It isn't. But you can't make that choice between a 1 and 5 year stretch. Your stretch is only between 1 and 0 years, or between 10 and 5 years.
0. My partner is going to reason the same way I do, because my partner is facing the same circumstances that I am.
Just like you, your partner is going to see that whatever choice you make, he is better off defecting. But that's as far as it goes. You can't communicate during the process.
1. I can either allow myself the possibility of defecting, or not. My partner, likewise.
No. The possibility is right there in the strategies and can't be wished away.
What you're describing is a different game than the PD, one from some sort of cooperative game theory in which somehow binding yourself from defecting is a strategy. But in the PD game, it isn't.
3. On the other hand, if I am about to reason my way out of allowing that possibility, then so will my partner.
You can't do that. If that were possible, it would be there in your strategy hyperspace. You'd have to choose from the set {cooperate, defect, deny self the possibility of defection}. Defection is an option for both of you, because it's right there in the set of available strategies. It can't be wished away.
The sort of "If I do this my partner will too" reasoning is, as it turns out, just incorrect in this setting. Your guide to what your partner will do is his strategy space and payoffs. "I want to do this, but if others did it too that would be bad, therefore I won't do it" is fine moral reasoning, but it's not strategic reasoning.
Another way of thinking about this: if you play a series of PD games against others, the people who reason strategically will clean your clock, and you will rot in prison forever. Then you will stop playing PD games, and only strategic reasoners will play them.
posted by ROU_Xenophobe at 10:46 AM on January 17, 2012
It isn't. But you can't make that choice between a 1 and 5 year stretch. Your stretch is only between 1 and 0 years, or between 10 and 5 years.
0. My partner is going to reason the same way I do, because my partner is facing the same circumstances that I am.
Just like you, your partner is going to see that whatever choice you make, he is better off defecting. But that's as far as it goes. You can't communicate during the process.
1. I can either allow myself the possibility of defecting, or not. My partner, likewise.
No. The possibility is right there in the strategies and can't be wished away.
What you're describing is a different game than the PD, one from some sort of cooperative game theory in which somehow binding yourself from defecting is a strategy. But in the PD game, it isn't.
3. On the other hand, if I am about to reason my way out of allowing that possibility, then so will my partner.
You can't do that. If that were possible, it would be there in your strategy hyperspace. You'd have to choose from the set {cooperate, defect, deny self the possibility of defection}. Defection is an option for both of you, because it's right there in the set of available strategies. It can't be wished away.
The sort of "If I do this my partner will too" reasoning is, as it turns out, just incorrect in this setting. Your guide to what your partner will do is his strategy space and payoffs. "I want to do this, but if others did it too that would be bad, therefore I won't do it" is fine moral reasoning, but it's not strategic reasoning.
Another way of thinking about this: if you play a series of PD games against others, the people who reason strategically will clean your clock, and you will rot in prison forever. Then you will stop playing PD games, and only strategic reasoners will play them.
posted by ROU_Xenophobe at 10:46 AM on January 17, 2012
What you're describing is a different game than the PD, one from some sort of cooperative game theory in which somehow binding yourself from defecting is a strategy. But in the PD game, it isn't.
Which is precisely why I've always found the PD game, and its widespread use as a primer to introduce game theory, so irritating. Game theory is all about what purely self-interested actors will do. It's not about what actual human beings will do. And yet it's frequently presented (just as you're doing here) as if it were a mathematical set of Iron Laws Of All Behavior rather than a framework for constructing simplified models of empathy-free behavior.
This wouldn't be such an irritating thing, were it not for the fact that educated people do tend to remain invested in the conceptual frameworks their education has given them; from which it follows that game theory modeling can operate as a self-fulfilling prediction of, and pseudo-justification for, the disintegration of interpersonal trust. And without trust as a basic starting position, there are many circumstances in which people find themselves serving a metaphorical five year stretch when they could have got off with one.
if you play a series of PD games against others, the people who reason strategically will clean your clock, and you will rot in prison forever. Then you will stop playing PD games, and only strategic reasoners will play them.
Facts say otherwise.
posted by flabdablet at 4:34 PM on January 17, 2012
Which is precisely why I've always found the PD game, and its widespread use as a primer to introduce game theory, so irritating. Game theory is all about what purely self-interested actors will do. It's not about what actual human beings will do. And yet it's frequently presented (just as you're doing here) as if it were a mathematical set of Iron Laws Of All Behavior rather than a framework for constructing simplified models of empathy-free behavior.
This wouldn't be such an irritating thing, were it not for the fact that educated people do tend to remain invested in the conceptual frameworks their education has given them; from which it follows that game theory modeling can operate as a self-fulfilling prediction of, and pseudo-justification for, the disintegration of interpersonal trust. And without trust as a basic starting position, there are many circumstances in which people find themselves serving a metaphorical five year stretch when they could have got off with one.
if you play a series of PD games against others, the people who reason strategically will clean your clock, and you will rot in prison forever. Then you will stop playing PD games, and only strategic reasoners will play them.
Facts say otherwise.
posted by flabdablet at 4:34 PM on January 17, 2012
And it turns out that my approach to this thing has a name, which I was unaware of before coming here. So thanks for that.
posted by flabdablet at 4:37 PM on January 17, 2012
posted by flabdablet at 4:37 PM on January 17, 2012
The sort of "If I do this my partner will too" reasoning
This is not the sort of reasoning I'm indulging in here. My partner's choice does not follow from mine; rather, both our choices follow from the assumption of a shared understanding that our reasoning processes are similar and our circumstances are identical.
Look, I'm not denying that game theory has useful real-world applications. But it's a matter of empirical fact that simplified theoretical frameworks often predict outcomes at odds with the real world to a greater or lesser extent. To assume that human beings can reasonably be modeled as game-theory-rational actors strikes me as akin to assuming that pi equals three. Sometimes, "close enough for government work" is actually not usefully close at all.
posted by flabdablet at 4:51 PM on January 17, 2012
This is not the sort of reasoning I'm indulging in here. My partner's choice does not follow from mine; rather, both our choices follow from the assumption of a shared understanding that our reasoning processes are similar and our circumstances are identical.
Look, I'm not denying that game theory has useful real-world applications. But it's a matter of empirical fact that simplified theoretical frameworks often predict outcomes at odds with the real world to a greater or lesser extent. To assume that human beings can reasonably be modeled as game-theory-rational actors strikes me as akin to assuming that pi equals three. Sometimes, "close enough for government work" is actually not usefully close at all.
posted by flabdablet at 4:51 PM on January 17, 2012
What you're describing is a different game than the PD
No, the game is the same; only the assumed characteristics of the actors is different.
posted by flabdablet at 4:53 PM on January 17, 2012
No, the game is the same; only the assumed characteristics of the actors is different.
posted by flabdablet at 4:53 PM on January 17, 2012
No, the game is the same; only the assumed characteristics of the actors is different.
The only characteristics the actors have are strategies (or, really, numbers of strategies; their names are irrelevant) and payoffs. You're describing adding a third row or column to the game, since "deny self the capacity to defect" has to be different from either cooperation or defection.
Game theory is all about what purely self-interested actors will do. It's not about what actual human beings will do. And yet it's frequently presented (just as you're doing here) as if it were a mathematical set of Iron Laws Of All Behavior rather than a framework for constructing simplified models of empathy-free behavior.
No, you can work empathy in easily enough if you want to by making player 1's payoffs depend on player 2's, so that player 1 receives utility from player 2's utility. Then player 1 responds to its own payoffs, that are correlated with player 2's.
Not done much, but easy enough. You could easily do it to the PD game by making the players both happy from each other's happiness... but then it wouldn't be the PD game any more, since what defines the PD game as opposed to some other 2X2 strategic form game are the payoffs.
Facts say otherwise.
Yeah, I should have been clearer that I meant a string a one-shot games where your probability of meeting the same player again is exactly zero. Without repeated interaction with the same player, you can't play trigger strategies.
it follows that game theory modeling can operate as a self-fulfilling prediction of, and pseudo-justification for, the disintegration of interpersonal trust
There's actually at least one paper on something related to this... I can't find the cite now. Anyway, it revolves around the centipede game, which you can hit on wikipedia when it returns. The central idea is that it's a game where the longer two players cooperate, the bigger their payoffs, but it's structured in a way so that the equilibrium result is for the first player to immediately defect and receive a small payoff.
Anyway, when researchers have people play this game in experimental settings, they rarely defect immediately, instead getting a little ways out on the tree before someone defects. The usual assumption is that the stakes in experiments, a few dollars to maybe, at absolute most, $100 or so, just aren't high enough to induce strategic behavior given some inherent utility in fucking around.
...except when economics grad students play it. They just solve the game and defect immediately.
On the other hand, there are also strongly game-theoretical explications of how to build trust as well. Often in a signaling or repeated-game setting rather than simple game trees or matrices, but still.
But it's a matter of empirical fact that simplified theoretical frameworks often predict outcomes at odds with the real world to a greater or lesser extent.
I'd agree that the PD as presented is sort of dumb, since hardly anyone has experience with being interrogated by police and the dramatic nature of the contrived situation invites all manner of thought experiments.
It's easier to present the same class of collective action problems in other ways, ranging from simple public goods to tragedies-of-commons to "Who's going to pick up the dorm common room?"
The real world actually matches up to the predictions of this class of model quite well. Just look at the oceans, or check the pollution count in your favorite air, or keep an eye on your thermometer for a 50-year span.
To assume that human beings can reasonably be modeled as game-theory-rational actors strikes me as akin to assuming that pi equals three.
Game theory, or rational choice theory more generally, works pretty well when the stakes are high and/or you can appeal to a pseudo-evolutionary force that drives out actors who behave non-strategically.
Absent institutions or structures that shift actors from one game to another, firms that don't maximize profit tend to be driven from the market, and criminals who don't defect in PD games tend to be driven from the criminal world into prison, and people who lightly use common pool resources tend to be outcompeted by people who abuse them (or, sometimes, just starve and die).
posted by ROU_Xenophobe at 12:08 AM on January 18, 2012
The only characteristics the actors have are strategies (or, really, numbers of strategies; their names are irrelevant) and payoffs. You're describing adding a third row or column to the game, since "deny self the capacity to defect" has to be different from either cooperation or defection.
Game theory is all about what purely self-interested actors will do. It's not about what actual human beings will do. And yet it's frequently presented (just as you're doing here) as if it were a mathematical set of Iron Laws Of All Behavior rather than a framework for constructing simplified models of empathy-free behavior.
No, you can work empathy in easily enough if you want to by making player 1's payoffs depend on player 2's, so that player 1 receives utility from player 2's utility. Then player 1 responds to its own payoffs, that are correlated with player 2's.
Not done much, but easy enough. You could easily do it to the PD game by making the players both happy from each other's happiness... but then it wouldn't be the PD game any more, since what defines the PD game as opposed to some other 2X2 strategic form game are the payoffs.
Facts say otherwise.
Yeah, I should have been clearer that I meant a string a one-shot games where your probability of meeting the same player again is exactly zero. Without repeated interaction with the same player, you can't play trigger strategies.
it follows that game theory modeling can operate as a self-fulfilling prediction of, and pseudo-justification for, the disintegration of interpersonal trust
There's actually at least one paper on something related to this... I can't find the cite now. Anyway, it revolves around the centipede game, which you can hit on wikipedia when it returns. The central idea is that it's a game where the longer two players cooperate, the bigger their payoffs, but it's structured in a way so that the equilibrium result is for the first player to immediately defect and receive a small payoff.
Anyway, when researchers have people play this game in experimental settings, they rarely defect immediately, instead getting a little ways out on the tree before someone defects. The usual assumption is that the stakes in experiments, a few dollars to maybe, at absolute most, $100 or so, just aren't high enough to induce strategic behavior given some inherent utility in fucking around.
...except when economics grad students play it. They just solve the game and defect immediately.
On the other hand, there are also strongly game-theoretical explications of how to build trust as well. Often in a signaling or repeated-game setting rather than simple game trees or matrices, but still.
But it's a matter of empirical fact that simplified theoretical frameworks often predict outcomes at odds with the real world to a greater or lesser extent.
I'd agree that the PD as presented is sort of dumb, since hardly anyone has experience with being interrogated by police and the dramatic nature of the contrived situation invites all manner of thought experiments.
It's easier to present the same class of collective action problems in other ways, ranging from simple public goods to tragedies-of-commons to "Who's going to pick up the dorm common room?"
The real world actually matches up to the predictions of this class of model quite well. Just look at the oceans, or check the pollution count in your favorite air, or keep an eye on your thermometer for a 50-year span.
To assume that human beings can reasonably be modeled as game-theory-rational actors strikes me as akin to assuming that pi equals three.
Game theory, or rational choice theory more generally, works pretty well when the stakes are high and/or you can appeal to a pseudo-evolutionary force that drives out actors who behave non-strategically.
Absent institutions or structures that shift actors from one game to another, firms that don't maximize profit tend to be driven from the market, and criminals who don't defect in PD games tend to be driven from the criminal world into prison, and people who lightly use common pool resources tend to be outcompeted by people who abuse them (or, sometimes, just starve and die).
posted by ROU_Xenophobe at 12:08 AM on January 18, 2012
The only characteristics the actors have are strategies (or, really, numbers of strategies; their names are irrelevant) and payoffs.
Not so. They also have some form of rationality.
Standard game-theory/economic rationality says that a player ought to consider only the payoff matrix, and make no assumptions about the reasoning processes of other players.
Superrationality says that a player ought to take into account the reasoning processes of other players, because doing so can make it clear that certain parts of the payoff matrix are in fact unreachable and ought not to be evaluated.
I come back to the point that what irritates me about game theory in general and the standard PD in particular is that the first kind of rationality is generally just assumed without examination. This kind of thinking and modelling frequently leads policy makers to accept stable Nash equilibria as the only "realistic" outccomes, even when suitable policy can make an unstable Nash equilibrium (featuring better outcomes for all than does the stable one) quite achievable. The present climate-change predicament is an excellent illustration of why I find this attitude irritating: unless Chinese, Indian and US policymakers start acting superrational PDQ, we're all fucked; but until the rest of us do likewise, there is no social pressure on any of the Big Three to move away from economic "rationality" AKA ecological insanity.
Your dorm common room example is another good example. Younger people do, by and large, tend to reason along economic-rationalist lines; adults tend more toward superrationality. Which is exactly why the common room in a student dorm never gets cleaned up and stinks of spilled bong water, while the shared spaces in a retirement village will be spick and span.
And I've lost count of the number of AskMe relationship threads I've weighed into with advice about the desirability of extending unconditional trust to a significant other until such time as one's nose is rubbed in a betrayal, which is an extension of the same basic superrational attitude. There are lots of real-world games where the best way to proceed is to play as if your partners are also superrational and switch partners as soon as they prove not to be.
posted by flabdablet at 2:46 AM on January 18, 2012
Not so. They also have some form of rationality.
Standard game-theory/economic rationality says that a player ought to consider only the payoff matrix, and make no assumptions about the reasoning processes of other players.
Superrationality says that a player ought to take into account the reasoning processes of other players, because doing so can make it clear that certain parts of the payoff matrix are in fact unreachable and ought not to be evaluated.
I come back to the point that what irritates me about game theory in general and the standard PD in particular is that the first kind of rationality is generally just assumed without examination. This kind of thinking and modelling frequently leads policy makers to accept stable Nash equilibria as the only "realistic" outccomes, even when suitable policy can make an unstable Nash equilibrium (featuring better outcomes for all than does the stable one) quite achievable. The present climate-change predicament is an excellent illustration of why I find this attitude irritating: unless Chinese, Indian and US policymakers start acting superrational PDQ, we're all fucked; but until the rest of us do likewise, there is no social pressure on any of the Big Three to move away from economic "rationality" AKA ecological insanity.
Your dorm common room example is another good example. Younger people do, by and large, tend to reason along economic-rationalist lines; adults tend more toward superrationality. Which is exactly why the common room in a student dorm never gets cleaned up and stinks of spilled bong water, while the shared spaces in a retirement village will be spick and span.
And I've lost count of the number of AskMe relationship threads I've weighed into with advice about the desirability of extending unconditional trust to a significant other until such time as one's nose is rubbed in a betrayal, which is an extension of the same basic superrational attitude. There are lots of real-world games where the best way to proceed is to play as if your partners are also superrational and switch partners as soon as they prove not to be.
posted by flabdablet at 2:46 AM on January 18, 2012
firms that don't maximize profit tend to be driven from the market
That certainly appears to be a common ideology among CEOs and economists and yes, that irritates me too (I turn 50 next month, so I'm entitled to be irritated by these things).
But I give you the counterexample of my present ISP, which has been in business for about twenty years and continues to show healthy growth in both revenues and customer numbers; their strategy is explicitly not profit maximization but business sustainability.
posted by flabdablet at 2:58 AM on January 18, 2012
That certainly appears to be a common ideology among CEOs and economists and yes, that irritates me too (I turn 50 next month, so I'm entitled to be irritated by these things).
But I give you the counterexample of my present ISP, which has been in business for about twenty years and continues to show healthy growth in both revenues and customer numbers; their strategy is explicitly not profit maximization but business sustainability.
posted by flabdablet at 2:58 AM on January 18, 2012
Your dorm common room example is another good example. Younger people do, by and large, tend to reason along economic-rationalist lines; adults tend more toward superrationality. Which is exactly why the common room in a student dorm never gets cleaned up and stinks of spilled bong water, while the shared spaces in a retirement village will be spick and span.
No, the shared spaces in a retirement village will be spick and span because (a) the usually partly incapacitated people in the village are less able to make a mess, but mostly (b) because they pay someone to clean it up.
The present climate-change predicament is an excellent illustration of why I find this attitude irritating:
Is the problem with game theory that it is too inaccurate and doesn't describe what people do, or that it too accurately describes what people do but that's bad?
In any case, the second criticism doesn't really make a whole lot of sense as it (or cognate forms of analysis) are very commonly used to try to solve those problems without magic wands that make everyone empathetic and wonderful. The answer is in the first bit, "Absent institutions or structures that shift actors from one game to another." Make the damn institutions and the problems go away to be replaced with other problems.
posted by ROU_Xenophobe at 6:37 AM on January 18, 2012
No, the shared spaces in a retirement village will be spick and span because (a) the usually partly incapacitated people in the village are less able to make a mess, but mostly (b) because they pay someone to clean it up.
The present climate-change predicament is an excellent illustration of why I find this attitude irritating:
Is the problem with game theory that it is too inaccurate and doesn't describe what people do, or that it too accurately describes what people do but that's bad?
In any case, the second criticism doesn't really make a whole lot of sense as it (or cognate forms of analysis) are very commonly used to try to solve those problems without magic wands that make everyone empathetic and wonderful. The answer is in the first bit, "Absent institutions or structures that shift actors from one game to another." Make the damn institutions and the problems go away to be replaced with other problems.
posted by ROU_Xenophobe at 6:37 AM on January 18, 2012
mostly (b) because they pay someone to clean it up
This option is also available to the young. They typically don't exercise it.
Is the problem with game theory that it is too inaccurate and doesn't describe what people do, or that it too accurately describes what people do but that's bad?
No, the problem with game theory is that the model of "rational behavior" that underlies it gets treated as gospel by people who create public policy, meaning that many of them actively resist taking positions of leadership in doing things that everybody understands are necessary because it's game theoretically more advantageous to let some other player do that. To a fairly large extent, we're fucked because game theory says we are.
posted by flabdablet at 7:06 AM on January 18, 2012
This option is also available to the young. They typically don't exercise it.
Is the problem with game theory that it is too inaccurate and doesn't describe what people do, or that it too accurately describes what people do but that's bad?
No, the problem with game theory is that the model of "rational behavior" that underlies it gets treated as gospel by people who create public policy, meaning that many of them actively resist taking positions of leadership in doing things that everybody understands are necessary because it's game theoretically more advantageous to let some other player do that. To a fairly large extent, we're fucked because game theory says we are.
posted by flabdablet at 7:06 AM on January 18, 2012
Make the damn institutions and the problems go away to be replaced with other problems
As economically rational agents, it's clearly in our best interests to let somebody else find out what those other problems would be. And making those damn institutions might cost somebody something. How about we delay that for another few years?
posted by flabdablet at 7:10 AM on January 18, 2012
As economically rational agents, it's clearly in our best interests to let somebody else find out what those other problems would be. And making those damn institutions might cost somebody something. How about we delay that for another few years?
posted by flabdablet at 7:10 AM on January 18, 2012
To a fairly large extent, we're fucked because game theory says we are.
This is becoming bizarre. Depending on how you define it, game theory dates from anywhere between 1928 and 1951. I think you will find the odd spot of environmental damage or common-pool-resource depletion before then. The idea that we are fucked because of game theory flies in the face of essentially the entire global history of what are now recognized as collective action problems.
posted by ROU_Xenophobe at 7:50 AM on January 18, 2012
This is becoming bizarre. Depending on how you define it, game theory dates from anywhere between 1928 and 1951. I think you will find the odd spot of environmental damage or common-pool-resource depletion before then. The idea that we are fucked because of game theory flies in the face of essentially the entire global history of what are now recognized as collective action problems.
posted by ROU_Xenophobe at 7:50 AM on January 18, 2012
The idea that we are fucked because of game theory flies in the face...
Didn't mean to give the impression that I thought game theory, or economics, or other disciplines based on the assumption that all agents are solely self-interested was the only cause of the policy deadlocks besetting us today. Clearly that's not the case. But I remain convinced that a lot of policy makers do start from the assumption that naked self-interest is the only realistic motivation for human action, that we fairly often see the suboptimal results that follow from this which is both sad and irritating, and that training in game theory or economics or both is indeed a strong contributing factor to this worldview.
I'm glad to live in a country where TINA's relentless march has not yet managed to destroy the socialist medicine system set up by the Whitlam Government in the 70s. We're apparently quite superrational about our health here.
posted by flabdablet at 5:06 AM on January 19, 2012
Didn't mean to give the impression that I thought game theory, or economics, or other disciplines based on the assumption that all agents are solely self-interested was the only cause of the policy deadlocks besetting us today. Clearly that's not the case. But I remain convinced that a lot of policy makers do start from the assumption that naked self-interest is the only realistic motivation for human action, that we fairly often see the suboptimal results that follow from this which is both sad and irritating, and that training in game theory or economics or both is indeed a strong contributing factor to this worldview.
I'm glad to live in a country where TINA's relentless march has not yet managed to destroy the socialist medicine system set up by the Whitlam Government in the 70s. We're apparently quite superrational about our health here.
posted by flabdablet at 5:06 AM on January 19, 2012
« Older Mapstalgia - video game maps drawn from memory | The Earth and its Peoples, over 100 years ago Newer »
This thread has been archived and is closed to new comments
posted by furtive at 1:09 PM on January 13, 2012 [5 favorites]