Game Theory – Prison Breakthrough | Hacker News Reader

Game Theory – Prison Breakthrough

economist.com

130 pointsjohnwheeler27 comments

This article presents the reflections of Game Theorist, Ariel Rubenstien, about the practical uses of his discipline.

http://www.faz.net/aktuell/feuilleton/debatten/game-theory-h...

It's worth reading the whole article, but he cautious about the predictive power of many of Game-theory's most elegant models:

"In my view, game theory is a collection of fables and proverbs. Implementing a model from game theory is just as likely as implementing a fable. A good fable enables us to see a situation in life from a new angle and perhaps influence our action or judgment one day. But it would be absurd to say that “The Emperor’s New Clothes” predicts the path of Berlusconi"

To give some context, he is one the 100 most cited economists in the world, coauthor of the most popular graduate Game Theory textbook and has a reasonable chance of winning a Nobel over the next couple of decades for his work on game-theoretic bargaining models.

Though now that he points it out, it seems like the path of Berlusconi is rather well explained by the "Emperor's New Clothes".

The article describes interpersonal Prisoner's Dilemmas, where if every person optimizes for themselves, the final result will be worse for each and every person (than if they had all shown restraint). As the article points out, this may explain pollution and overfishing.

There are also intertemporal Prisoner's Dilemmas, where each if person-at-time-T optimizes for their immediate goals, the final result will be worse for each and every moment in that person's life. This may explain why we overeat: any given moment is made better with a Big Mac, but if you ate a Big Mac all the time every moment your life would be made much worse than if you led a restrained, healthy life.

If only there was some way for versions of yourself at different times to coordinate. Oh, right...

It seems the act of coordination itself can be modeled similarly to not eating a big mac.

I came here thinking this was a recent breakthrough, but this appears to just be an overview of the 1948 Prisoners Dillemma and an update from 1994...

For another view on how the Prisoner's Dilemma should affect your worldview beyond academia, see this recent article by Tim Gowers on the Brexit: https://gowers.wordpress.com/2016/06/02/6172/

I skimmed the article and I have to say, I am not convinced at all. Decentralization can be achieved without causing situations like prisoner's dilemma. Look at Torrent for example, it works without a central authority enforcing things. While total decentralization is ineffective, centralization is authoritarian and destructive. We have to construct a system where it pays off to be honest and fair, such that force and subjugation is not necessary. This is the only way forward. And frankly, I do not see the EU play a positive role in any of this.

centralization also results in severely less optimal results because the information and experience available to central planners is substantially less then information available to individual actors.

Not only is the information of poorer quality prior to making the decisions the quality of feedback is poor. It becomes very difficult to understand the consequences of decisions when you have nothing to compare it to.

In a decentralized system the individual actors can see the results of decisions made by other actors and change accordingly. In a centralized system there is no other 'center' to compare it against. They could be making good decisions or they could possibly be not.

In the real-world the amount of unknown variables is so high that even decisions made with the best intentions, most rational approach, and most perfect knowledge possible can fail miserably. It wouldn't be the fault of the central planners, per say, but it will still be a negative outcome.

Also a decentralized system the consequences to negative decisions is dampened. Even if the bad decision was done out of perfect rationality there is enough irrational actors that somebody else will likely stumble across a superior decision and thus guide a more positive outcome. A bad decision by a individual actor will potentially be devastating to that actor, but the effects of that decision will be limited to that actor and those around him that may be negatively impacted.. it won't impact the entire group and his example will lead to better decision making in the near future.

With centralized planning the bad decisions affect the entire group and it can take much much longer for the bad decision to be recognized as bad and even longer to change. By that time you can fix the problem have the group adjusted to work around the bad decisions and undoing the bad will just create new issues.

> Also a decentralized system the consequences to negative decisions is dampened. Even if the bad decision was done out of perfect rationality there is enough irrational actors that somebody else will likely stumble across a superior decision and thus guide a more positive outcome.

Or, alternatively, it will be a race to the bottom until the housing market crashes.

> A bad decision by a individual actor will potentially be devastating to that actor, but the effects of that decision will be limited to that actor and those around him that may be negatively impacted

Unless that actor has nuclear weapons, poisons a water supply, destroys the source of some resource the rest of the world depends on...

> Or, alternatively, it will be a race to the bottom until the housing market crashes.

The housing market crash was due to central planning. Hardly a counter point.

> Unless that actor has nuclear weapons, poisons a water supply, destroys the source of some resource the rest of the world depends on...

Central planners are humans.

When comparing human systems of central vs decentralized management you have to understand that the central planners are just as prone to irrationality as individual actors.

Pointing out that individual choices can lead to bad outcomes does not mean that central planning choices will not lead to bad outcomes.

The problem is one of dealing with complex mulch-facteded world were historical information will very often lead to poor decision making in the face of change.

The lack of good information, lack of good feedback mechanisms, and inability to gain enough experience are all severely limiting to the chances of quality outcomes as a result of central planning.

>The housing market crash was due to central planning. Hardly a counter point.

Citation needed on how central planning led to MBS and CDS exploding.

Fannie Mae and Freddie Mac -> https://en.wikipedia.org/wiki/Federal_takeover_of_Fannie_Mae...

Admittedly not the only factor but a huge contributor.

> centralization also results in severely less optimal results because the information and experience available to central planners is substantially less then information available to individual actors.

This is a throwaway generalisation. A simple counter-example is pollution, where significant improvements come from centralised actions (watchdog agencies and regulation) because individuals either don't know or don't care about their contributions.

There isn't one thing you can say about centralisation that is going to usefully apply across the diverse cooperative situations people face.

flaws I described in central planning are pretty universal.

Nobody can say any one thing that will apply to all things all the time.

Your argument doesn't seem to apply to something like climate change. Of course it already pays off to be honest and reduce emissions, but there is a conflict between short-term profits and long-term goals. Unlike Torrent, in which a small number of good actors are sufficient for the health of the system, with climate change you need _most_ actors to be honest. Same goes for nuclear weapons.

I have yet to read or hear a convincing argument that decentralization is an effective solution to the climate change problem (and I work at a bitcoin company!). I'm open ears, though :)

It does NOT pay off to reduce emissions, in the short term or the long term.

A single factory will not suffer the consequences of global warming, and has an incentive to produce as many bad emmisions as possible, and make OTHER people pay for the consequences.

That's why it is a prisoners dilemma.

It pays off in a less tangible sense, that the people who run the factory will have a successful line of children.

> That's why it is a prisoners dilemma.

I disagree. In the prisoner's dilemma the prisoner chooses to defect despite it ultimately and predictably having a negative effect on himself personally.

Their graphic and dilemma are nonstandard and do not match, for anyone confused by the graphic. The normal formulation of the problem is:

- If A and B each betray ('snitch' in the article) the other, each of them serves 10 years in prison.

- If A betrays B but B remains silent, A will be set free and B will serve life in prison (and vice versa).

- If A and B both remain silent, both of them will only serve 1 year in prison (on the lesser charge).

In the article, the prisoners are given three choices: Silence, Betrayal, and Confession, which confuses the issue, and the graphic provides only Silence and Confession, which further produces confusion.

Perhaps the modern formulation is too indoctrinated with plea bargaining?

There is no third choice. They're using both 'confess' and 'snitch' as meaning 'telling the truth about what happened'. The description in the article may be confusing, but it does match the diagram and it's fundamentally the same as yours.

Thank you. I assumed my instinct was correct, but fell-back on the assumption that I am simply misunderstanding something.

For those of you thinking this is at all related to the recent de-privitization movenent in the US prison system, it is not. This is more of a breif history and overview.

Yeah, the "breakthrough" in the title was very misleading. It should be renamed to something like "the [early] history of the Prisoner's Dilemma".

Agreed. After reading the article and related comment threads, my perception is one that aligns with this view.

The prisoners dillema fails to take into account that both prisoners know each other, and one of the prisoners may not make a rational decision. Say there is prisoner a and b. Prisoner a is a drooling moron who will act again his self interest. Prisoner b knows this. Prisoner b will do the thing that will make him get out of jail for good. Essentially I'm not playing along with this "riddle" because its flawed in the first place because it fails to understand elementary consciousness.

A friend working with inmates says that they laugh at the idea of the prisoner dilemma. They say there is no dilemma: everybody knows that one must never confess or there will be consequences.

These are not occasional criminals but people for which crime is a career path, exactly the mobsters of the example in the article. They have rules to follow which affect all their professional life in the same way we have contracts with our customers.

Reply on news.ycombinator.com