One-Shot vs. Iterated Prisoner's Dilemma

This post by Aleatha Parker-Wood is very applicable to the things I wrote in Liars & Outliers:

A lot of fundamental social problems can be modeled as a disconnection between people who believe (correctly or incorrectly) that they are playing a non-iterated game (in the game theory sense of the word), and people who believe that (correctly or incorrectly) that they are playing an iterated game.

For instance, mechanisms such as reputation mechanisms, ostracism, shaming, etc., are all predicated on the idea that the person you’re shaming will reappear and have further interactions with the group. Legal punishment is only useful if you can catch the person, and if the cost of the punishment is more than the benefit of the crime.

If it is possible to act as if the game you are playing is a one-shot game (for instance, you have a very large population to hide in, you don’t need to ever interact with people again, or you can be anonymous), your optimal strategies are going to be different than if you will have to play the game many times, and live with the legal or social consequences of your actions. If you can make enough money as CEO to retire immediately, you may choose to do so, even if you’re so terrible at running the company that no one will ever hire you again.

Social cohesion can be thought of as a manifestation of how “iterated” people feel their interactions are, how likely they are to interact with the same people again and again and have to deal with long term consequences of locally optimal choices, or whether they feel they can “opt out” of consequences of interacting with some set of people in a poor way.

Tags: , , , , , ,

Posted on May 23, 2013 at 9:18 AM • 25 Comments

Comments

Jurgen • May 23, 2013 9:45 AM

… would anyone have a simple-to-use multi-period multi-play prisoner’s dilemma strategy tweaking app so we could demonstrate this (and/or in line with L&O) …?

Mike Petonic • May 23, 2013 10:41 AM

Interesting concept and sort of formalizes phenomena that I’ve experienced.

For example, consider the favorite topic of car driving behavior… Turn signals. Not everyone uses them. In fact, in California, I would suggest that only about 35% of the people have internalized the habit enough to make using them second nature. It takes more psychic energy to not use turn signals than it does for those folks, even in, ridiculously enough, a parking garage with no one else there.

Those of us who use signals have probably had enough feedback from our driving mentors or partners that we feel as if we are in an iterative game. Those that do not probably see the mass of humanity that they share the road with as presenting a one-shot game.

paul • May 23, 2013 11:07 AM

The CEO example is particularly apposite, because in a winner-take-all society it’s the very choice to defect that makes the dilemma non-iterated. If you decide to take a reasonable salary, you have to stay in the job and deal with all those people some more; it’s only if you trouser as much as possible that you get to opt out.

This takes Akerlof & Romer’s 1993 “Looting” paper one step further, because in the current system the act of acquiring huge piles of wealth essentially makes you immune to clawback.

Brian Krebs is the Chris Hansen of Cyber Security • May 23, 2013 12:35 PM

If encouraging the iterative game viewpoint is better for behavior and society, then what’s so great about fighting to maintain degrees of privacy online which have historically never existed in person?

Is a virtually untraceable identity a normal aspect of our physical culture, or is it for special cases like voting? If someone was running around under the cover of dark and doing even simple legal errands in a ski mask in the summer, they’d be denied service in a lot of cases and probably would be investigated on a regular basis–because in person, we all have the same expectation that we come to the table with our identities exposed. But online, some people don’t want to play by those same rules. I’m not talking about personal privacy like the right to control a book buying list at Amazon (the same as you may with cash) but the right to go out and do things to other people.

And I think a lot of the desire for anonymity online comes from a sort of arms race–people are operating on the knowledge that online communication is NOT an even playing field, which is unfair–an anonymous mob can hate a thing someone says, and then they will proceed to commit all sorts of fraud and mischief against their name, and they end up looking like a sucker for trying to act like a normal citizen in disclosing their identity.

vas pup • May 23, 2013 2:53 PM

When you are in electronic recording and saving all forever world, lack of privacy will force iterated model on you regradless of your intention after many years after your posting: HR or lawyers may hunt you with unpredictable (and even unreasonable) conclusions. The idea is that iterated model should be applied simmetrically to all parties involved, then transperency of one side (business, government, group, etc) could be balanced by openness of the other (person).

Geoff • May 23, 2013 6:09 PM

It’s important to consider when players perceive that the game is going to end. This can equate to short term vs long term strategies in the broader sense; which in turn can be mapped to Ethical Maturity imo.

Francois • May 23, 2013 6:28 PM

It’s interesting how one can be “shamed” into doing the right thing.

Ironically, it seems like the larger the group you participate in – and the more people you therefore interact with – the more anonymous you become, surveillance or no. The impact of each interaction is reduced as it is spread out among a greater number of people. As long as the spotlight isn’t turned on you, anonymity can be presumed further and further.

After all, increasing a number of security cameras doesn’t increase actual surveillance – only the observer can do that. But, increasing the number of cameras can contribute to a false sense of security, thus reducing the incentive to actually pay attention. We see this in security all the time.

So maybe privacy is not a thing of the past, after all. Maybe the nature of privacy is just changing. Instead of being defined by the possibility of being observed, it will be more defined by the actions – and focus – of the observers.

The easiest place to hide a book, after all, is in a library.

Brian • May 23, 2013 7:02 PM

I think that’s a very interesting way to look at social interaction, and it makes a certain amount of sense.

I also wonder about the impact of how people think OTHER people are playing the game. The CEO in the example might have a different approach than the long-term worker, but what’s the impact of the fact that the worker can SEE the difference in their required approaches? Does the worker resent the fact that he’s somewhat required to treat social interaction as an iterative process while the CEO can make a single optimal decision for himself and then leave the game? Even if the CEO doesn’t do that, the knowledge that he could have an impact on the worker.

Daniel • May 23, 2013 7:05 PM

IMO that is a poor way to conceptualize the issues because the frames of reference are inherently arbitrary. Every individual’s decision making structure in practice is the result of an infinite amount of prisoner’s dilemmas (both iterated and non-iterated) all happening at once. Take the CEO example. Even if we can imagine that the CEO might claw his way to the top and not care about the company when he leaves he still has to confront the fact he is playing an iterated game among his family, friends, and other business associates. In fact, the example as she outlines it is self-contradictory because it imagines a person who plays an iterated game to get to the top but then willingly abandons that game and plays an non-iterated game once at the top. But that contradicts everything we know about the way that human beings are conditioned and form psychological associations.

All she is really doing is switching frames of references to suit her a priori conclusions. It is neither effective nor persuasive.

Daniel • May 23, 2013 7:22 PM

Francois writes, “So maybe privacy is not a thing of the past, after all. Maybe the nature of privacy is just changing. Instead of being defined by the possibility of being observed, it will be more defined by the actions – and focus – of the observers.”

I agree. What needs to be understood is that this truth is not a genetically neutral event. For example, one of the primary ways that physical security could be maintained in the past was distance/space. This was one of the great drivers of human migration. If one part of a community did not get along with another part they simply packed up and left for unoccupied lands. This action is no longer possible in practice. No one can drop out or run away because no matter where you go another community is already there. The result is that the touchstone of privacy is no longer isolation but distraction. But some people are better at isolation than distraction. Their genes will tend to be disfavored.

AtomBoy • May 24, 2013 7:43 AM

Oooh, very deep stuff.

This concept can help analyze foreign based attacks and local.

It can analyze reader poster behavior.

For instance, posters here are effectively encouraged to give anonymous names. Some probably change their nicks frequently. Some clearly keep the same nick.

I find when I post on forums which interest me, if the consensus is against my opinion, I may choose to post and run, and so whatever “shaming” or flaming which may go on will give me little to no impact.

I think we all probably do this.

At other times, we may decide people are not that far gone, and we may want to stand our ground through all questions, flames, shames coming at us until we are left standing — despite how painful that can be.

This is all related to the real opinions we have, which we give to our family or close friends. Versus the ??? opinion we give in public.

Criminality wise – illegal surveillance wise – this can tie in. As that is one crime which is popular on this blog and in tech news lately….

… in consideration, I believe, though, cops, in general want to keep themselves segregated from society. Not unlike supervisors. Because they have to do things and say things people will not like.

That distance affords them the ability to do this. Otherwise, the emotional pain of being too close can creep in.

…

This concept can also speak of societal blinders we put on. Such as “why did they not notice that priest was pedophile”, or “why did they not notice Ames or Hanssen or Philby”. Because they were one of “you”. (Or, why, for that matter, have so many policing agencies been so opposed to the point of outrageous criminality when one of their own commits the crime — why did they not have, or do not have, internal affairs…? Was Serpico necessary to be outside of the group and so ostracisized already for his long hair and such… to become the man of change in American police corruption? etc?)

We look at external groups very differently then internal groups.

I think this relates to that concept.

And from there, goes into our often very serious psychological blinders when it comes to processing truth, in general.

The keyword I usually use is “investment”. Iterative or non-interative, if non-iterative, there is no investment.

Usually, serial killers are non-iterative. They originally called these “stranger killers”.

Gabe • May 24, 2013 8:21 AM

this is true in sports as well. If you’re 7 feet tall, you just need to be good enough to get drafted and then maybe for a few years in the pros, to get 1 big contract. Then you don’t need to care anymore, and can retire at 30 with $20MM+ in the bank.

vas pup • May 24, 2013 8:29 AM

Non-iterative v. iterative.
Who is making the choice: you or choice is made for you?
Example 1: you buying something and providing the company some information just to complete this particular transaction only. Your initilal intention is non-iterative towards the company. As soon as company is binding by your choice for future interactions (you specify that you do want future interaction – versus default – you don’t), that is okay, otherwise you forced for iterative model.
Example 2: man and woman (by consent – with hooker or not) have one-time sexual contact. By default both sides consider that non-iterative, and keep their autonomy for future, i.e. non-iterative. None is forced for iterative model if and only if that is their mutual choice, not society (police involvement with sting operation or shame by society).

Example 3: person is in jail, army, mental institution. No choice for person selecting the model. It is by default iterative.

Conclusion: more freedom/autonomy asumes more space to select non-iterative model by both sides of interaction (no extremes: – on positive side) without being forced to iterative.

Ambrose Bierce (1842-1914) • May 24, 2013 9:07 AM

“IMPUNITY, n. Wealth.”
(Ambrose Bierce, The Devil’s Dictionary)

Non-iterative systems (or interrupted iterative systems) wave the cost of sanctions.

The good news is that only a tiny minority can access cheating others.

The bad news is that this tiny minority makes its living by cheating all others.

Solution: make not cheating more rewarding than cheating.

Clive Robinson • May 24, 2013 4:56 PM

@ Ambrose Bierce (1842-1914),

    Solution: make not cheating more rewarding than cheating

Unfortunatly whilst Ambrose had a sense of humor, some people actually believe this “solution” is viable.

That is they don’t realise that some people have “a sense of entitlement” that unless they have more than anyone else they are not being correctly recognised (see failed bank executives as an example with their 20million salary plus bonus). These people inately believe that this recognition must be paid like “tribute” and unless given they will take it.

Paying these people more to stop them cheating only feeds the sense of entitlement so perversly only makes them want to cheat more and thus take more.

The correct solution is once their sense of entitlement has become known is to not pay them at all preferably by giving them free (for them) board and lodgings at a “state facility” untill they are cured of their sense of entitlement.

I’m told however that such kindness tends to also feed their sense of entitlement so whilst the solution works for the rest of us it might not work for them. Ahh well no solution is perfect 😉

Clive Robinson • May 25, 2013 12:06 AM

@ oh no,

    clearly there’s some final solution

I hate the expression “final solution” simply because there is only one that we know of and that is when entropy is at it’s maximum within the universe (unless the “big crunch” is back in favour).

Untill then or the human race nolonger exists as we currently know it I expect all social solutions to evolve with society…

However there is a certain base appeal to the cry of,

    Come the revolution, you lot…. 😉

Dirk Praet • May 25, 2013 4:35 PM

@ Daniel

Even if we can imagine that the CEO might claw his way to the top and not care about the company when he leaves he still has to confront the fact he is playing an iterated game among his family, friends, and other business associates.

The harsh reality however – and especially in the financial sector – is that whatever havoc they wreak in most cases will never backfire at them. Instead of doing jailtime or being burned forever, they receive golden parachutes and then get hired by other companies for their “unique and indepth knowledge of the business”. In the Wild West, cheating at a game of poker could get you killed. On Wall Street, it earns you the respect and admiration of everybody sitting at the table.

@ Atomboy

For instance, posters here are effectively encouraged to give anonymous names. Some probably change their nicks frequently.

No, we aren’t. We are given a choice. Sockpuppeting, bringing imaginary friends to the forum can and will get you banned.

@ Clive

I hate the expression “final solution” simply because there is only one that we know …

There is another one that comes to mind and which has tainted these words in a horrible way ever since a bunch of mass murderers at the Wannsee conference in 1942 cornered the word “Endlösung”.

Clive Robinson • May 25, 2013 11:27 PM

@ Dirk Praet,

    There is another one…

Yes and I was trying to avoid mentioning it because of it’s mid 20th Century meaning and it’s connections / connotations.

That said history shows us that bad as it was it was by no means the worst that goes in many peoples eyes to one of the three priniples at the Yalta Conferance [1].

However bad as these “solutions” were they did not reach the objective of “final” in that there were survivors who could subsiquently speak of what happened.

Sadly this has not always been the case, the history of America and Australia show that such ideas were still being acted upon against indigenous people as late as 1928 with “a blind eye being turned” by responsable officials and governments. Which is why when there were a few surviours they could not in most cases come forward to speak out. And even as I write this we know that around the globe the idea behind such “solutions” are still being acted upon on a daily basis in one way or another [2].

Man is perhaps the most distructive creature to have ever existed, but due to the fact we mainly don’t learn the lesons of history we almost endlessly repeat the mistakes usually more destructivly as our technology improves.

[1] Another of which is believed to have been the first to officialy sanction the use of chemical WMD against an indigenous people. His thoughts of “I am strongly in favour of using poisoned gas against uncivilised tribes.” Were recorded in May 1919 in a government memo and it is known that chemical weapons were certainly sent to Iraq which was then a mandate after WWI.

[2] Arguably it is actually getting worse, for instance one side effect “slavery” is believed to be at an all time high with something like 27million people kept as slaves currently. Other indicators such as “displaced persons” and civilian deaths by government sanction/action are likewise at very high levels and the world wide military death rates are currently greater than those of WWII.

AtomBoy • May 26, 2013 12:28 PM

@ Atomboy
For instance, posters here are effectively encouraged to give anonymous names. Some probably change their nicks frequently.
No, we aren’t. We are given a choice. Sockpuppeting, bringing imaginary friends to the forum can and will get you banned.
**

I do not mean “sock puppeting”. :/ That is dishonest, when you play up multiple parties often at odds with each other as a singular party.

I mean, it is easy and made friendly to be anonymous.

I think that is quite obvious from the context, and I am sure a number of posters do not use their real names.

Though, we can all pretend to shock here. Or, intentionally misunderstand in order to try and launch an attack against entities we believe are opposed to our own master entity.

Figureitout • May 26, 2013 8:27 PM

@AtomBoy
–So don’t reply to yourself and at the very least preview your comment and fix obvious errors. Never underestimate.

Figureitout • May 26, 2013 8:46 PM

@AtomBoy
–TLA’s have by no means monopolized methods to get in you.

braff • June 2, 2013 1:39 AM

My first thought about non-iterated vs iterated was that the main reasons for viewing social life as non-iterated was privilieges of diferent kinds, i.e. having the power and choice to evade the social consequences of your own actions.
Due to the recent rioting here in Stockholm, Sweden, I came to think about another aspect of it. Desperation and feeling of not beeing included in society might also lead to a sense of non-iteration, in that you might not even see that you have a future or that your future is so predictably bleak that your choices in the present moment cannot make it worse. Otoh, several rioters interviewed showed a clear sense of rioting as a conscious last-resort way of calling attention to their social situation due to mass medias ADD-like spectacle-focus.
Isn’t iterativeness an integrated part of all kinds of social interaction that have some value? Or actually that iterativeness gives value to social interaction? If every social interaction that I have with you is treated as a first encounter and nothing we say or do will have any lasting effect or repercussions, are we not all reliving the 50 First Dates?

Leave a comment

Login

Allowed HTML <a href="URL"> • <em> <cite> <i> • <strong> <b> • <sub> <sup> • <ul> <ol> <li> • <blockquote> <pre> Markdown Extra syntax via https://michelf.ca/projects/php-markdown/extra/

Sidebar photo of Bruce Schneier by Joe MacInnis.