Prisoner's Dilemma Experiment Illustrates Four Basic Phenotypes

If you've read my book Liars and Outliers, you know I like the prisoner's dilemma as a way to think about trust and security. There is an enormous amount of research -- both theoretical and experimental -- about the dilemma, which is why I found this new research so interesting. Here's a decent summary:

The question is not just how people play these games­ -- there are hundreds of research papers on that­ -- but instead whether people fall into behavioral types that explain their behavior across different games. Using standard statistical methods, the researchers identified four such player types: optimists (20 percent), who always go for the highest payoff, hoping the other player will coordinate to achieve that goal; pessimists (30 percent), who act according to the opposite assumption; the envious (21 percent), who try to score more points than their partners; and the trustful (17 percent), who always cooperate. The remaining 12 percent appeared to make their choices completely at random.

Posted on August 18, 2016 at 5:36 AM • 36 Comments

Comments

WmAugust 18, 2016 7:50 AM

This was very interesting and beings back a memory of a person I knew. He was the youngest brother of a brother in law. This guy always had to win. We were the same age and around 12 years younger than his bother and my sister. Whenever he rarely came over to my house, we would go outside and play cowboys and Indians. When it was his time to hide, he would hide himself so well, that I could not possibly find him and he would always make a gun shot sound immediately upon seeing me that would pronounce me as having been shot. When I would hide, it wouldn't be that effective, thereby giving him a chance to shoot me also. After he completely hid himself the second time and shot me, I started towards the door to go inside. I remember him asking if we weren't going to play anymore and I said no, because you never give anyone a chance, followed by a sheepish laugh from him. So people who don't play to always win instinctively know that to do so is not conducive to building or keeping a relationship with others.

Myron FlapdoodleAugust 18, 2016 7:54 AM

I've played a bunch of board games in my life and it's been a rare instance seeing an "envious" player. Thank goodness.

RichardAugust 18, 2016 9:30 AM

Instead of the 12 percent who chose at random being treated as noise, perhaps they represent a fifth group who have a completely different approach to things than the optimists, pessimists, envious, and trusting types.

Richard HarperAugust 18, 2016 9:59 AM

@Richard (Twelve percent of respondents were categorized as noise.) One interpretation is what can be said of much of personality testing, that the tests function well to discover if an individual is stuck in a persistent pattern of inflexible behavior and not responsive to specific circumstances. Most personality testing in this sense is testing for weaknesses in social functioning. Another way of phrasing that from the opposite direction in terms of strengths is to say that some twelve percent of the subjects were experimenting on the experiment, trying out different patterns to see what worked. In either way of framing it, maybe the twelve percent are the most interesting group. (There's probably an appropriate quote from Sun Tzu that could be made here about the use of formlessness in warfare.)

DanielAugust 18, 2016 10:29 AM

@AlanS

Count me among them. I don't think the PD (or game theory in general) offers any genuine insight into human behavior. Game theory is a set of circular arguments, vaguely defined terms, and wild ass philosophical presumptions that no one outside of game theory finds likely let alone persuasive.

An example of this is "self interest". Well, what is the self? There is an entire body of psychological literature that argues there is no such thing as the self and thus by logical extension there can't be such a thing as self-interest. See by way of example the book entitled "The Self Illusion". This is especially galling because game theory people claim they are being pragmatic and avoiding "aprioristic assumptions" when what they mean is that they assume all sorts of things and then proudly declare them extraneous to the model. They are not in fact avoiding assumptions, they are willfully ignoring the ones they make.

All game theory reveals is that there are some people who like to play games. There is nothing wrong with playing games per se; what is wrong is the arrogance that claims that games playing reveals anything more about human behavior than the fact some people like to play games.

ArchonAugust 18, 2016 10:32 AM

@Myron

Yeah, we had one of those in our group. He didn't last terribly long, especially when he got mad during a "friendly" game of M:TG because the newbie who borrowed a deck from a more experienced player was getting some advice from the deck's owner.

These days I play mostly co-op games like Legendary, Elder Sign, the Pathfinder card game, Ghost Stories (when I *want* to lose), and Descent 2nd edition with the new app that takes over the Overlord's role. The competitive stuff I play has changed too - it tends to be faster or doesn't have a knock-out element. Stuff like Small World, Danmaku!, or Zombie Dice.

Gaming's a social event, not a competition.

Matthew NewhallAugust 18, 2016 11:55 AM

Game theory only produces reliable results for psychopaths. The conscience is a risk management engine. In it's absence all you can do is take a guess and double down if it turns against you. Logic is inherently selfish. Altruism, the human condition, is irrational to some degree, yet has better odds than models.

I'd be more interested in who changed their strategy if a game is run annually.

As an aside, jealousy is the psychopath's/animal conscience. Available to all but the only non-rational mechanism available to discern risk to psychopaths.

I did read your book and it was great BTW.

JeremyAugust 18, 2016 1:59 PM

@Daniel:

> "no one outside of game theory finds [it] likely let alone persuasive."

Not true. I've never gone out of my way to study game theory and do not routinely apply it in work or in life, but on the occasions I've encountered it, it's struck me as being useful and insightful within a narrow scope. If you think that means I'm "inside" game theory, then you are making a "no true scotsman" argument.

Your tangent about the non-existence of the self seems...esoteric, to say the least. I think I have a self. Even if I'm wrong, it's clearly still possible for me to act in perceived self-interest. Furthermore, from a practical standpoint, whatever I actually have is likely to be considered my "self" by most non-specialists, whatever the ultimate nature of that thing turns out to be. If you want an example of a line of reasoning that sounds completely unpersuasive to those outside its field? This is what one looks like.

> "All game theory reveals is that there are some people who like to play games."

Whatever the merits or flaws of game theory, this is most emphatically NOT what it demonstrates.

"Game theory" isn't named after its subject, it's named after its tool. Gamers (and game designers) don't generally care about game theory, like most physicists don't study architecture. Game theory doesn't study games, it studies decisions.

AlanSAugust 18, 2016 2:15 PM

@ Jeremy

Game theory only studies decisions in a very limited way as it makes ill-founded assumptions about the nature of the decision-makers and the nature of their reasoning.

k15August 18, 2016 2:32 PM

I do not like that cooperation gets smeared by this naming, overall it is a social good and yet in the PD it's not.

JeremyAugust 18, 2016 2:48 PM

@ Alan S

Regardless of how successful it is, game theory attempts to study decisions.

Someone who thinks that game theory "reveals that some people like to play games" isn't just disagreeing about how good of a job it does, they have failed to understand what is being attempted in the first place.

On a side note, I would be more interested in your comments if you articulated specific flaws rather than saying "it's all garbage; if you want to know more, read a huge book". Show me enough substance to make be believe that a book would be worth my time. (Also, it would help if you were recommending a book that was primarily on the topic at hand, rather what looks like a political book that only touches on the topic tangentially.) You are currently giving me the impression that you have strong emotions but weak reasons--which doesn't mean you're wrong, but isn't doing much to convince me that you're right.

WinterAugust 18, 2016 3:23 PM

Game theory studies rational decisions. It studies what moves lead to the "best" outcome for a player in a move based game.

In a sense it is logic for moves in a game. This means that it is actually a branche of mathematics. As such it is not related to real decision-making by humans or animals.

AlanSAugust 18, 2016 3:57 PM

@ Jeremy

The words "it's all garbage" are yours; not mine. I wrote: "they have very limited applicability".

I also provided two links. True, one to a large, expensive book but the other was a free paper that's a quick read. I'd suggest you click the link and read the latter.

Here is an additional comment on this topic by me on another thread. The Sahlins paper referenced is also worth a read. It's also free.

AlanSAugust 18, 2016 4:11 PM

@ Jeremy

It would help if you were recommending a book that was primarily on the topic at hand, rather what looks like a political book that only touches on the topic tangentially

The book contains a lot of heavy analysis of game theory. It's technical, references a lot of philosophy, and is not light reading.

And, yes it does make a political point but are you really claiming game theory, rational choice, neoclassical economics, etc. aren't political? These are ideas that have been enormously influential in the shaping of modern political economy. Lots of people in the social sciences and humanities beg to differ on their claims to neutral, scientific status.

GweihirAugust 18, 2016 5:24 PM

If Game-Theory can sort about 88% of people reliably into classes here, it works rather obviously very well and the deniers are either blind to the facts or have another agenda. They certainly have no scientific point-of-view.

Now, if you look for independent thinkers, you need to go look at the remaining 12%. Incidentally, from teaching, it is not only my observation that typically 10-20% of the students take much more form a lecture and can do much more with the things learned than the others. Face it: Independent thinkers are rare. The rest is far less determined by free will and insight into things than they like to think they are and very often just follow a pattern without actually making any real decisions.

Yes, that is not a nice thing to say. But it is also accurate.

DanielAugust 18, 2016 5:55 PM

@Winter

What is a rational decision? How is that defined without resorting to game theory itself? If it defined by resorting to game theory then it is circular. If it is defined by something outside game theory then it is an "aprioristic assumption". Again, there is nothing wrong with having assumptions because everyone has assumptions. What is annoying is people who do have value-laden assumption claiming that they have no such assumptions, which is what game theory does. The idea that game theory is neutral is preposterous and the idea that it is any way logical or scientific is ridiculous.

@Jeremy

Something isn't a tool simply because you call it one. A tool is something that is used to make something else. What does game theory produce besides more game theory? When game theory shows its ability to predict real world results that are verifiable, replicated, and generalized I would accept game theory as a tool. But all game theory does is predict how people play game-theory games. That isn't science, that is intellectual masturbation.

JeremyAugust 18, 2016 7:04 PM

@Daniel:

I wasn't claiming that game theory is a tool. I was claiming that the "games" in game theory are tools, not the subject of study. Again: game theory does not study games. It uses "games" to study decisions. Whatever conclusions it can draw are not going to be conclusions about games.

There is a whole industry that actually creates game for people to play. That industry reveals that there are people who like to play games. Game theory does not. The "games" in game theory are not typically played for fun (not even by game theorists).

I realize you were only trying to deliver an insult, but your insult makes it look like you don't understand the field you are trying to criticize, so it kind of backfired.

The fact that you responded to both my argument and Winter's argument by arguing semantics is also strongly suggestive that you have nothing substantive to say.

JeremyAugust 18, 2016 7:48 PM

@ AlanS

This is pretty clearly a personal cause of yours. I am (or should be) your target audience: someone who knows enough about game theory to care, without being so deep as to have an established opinion.

But you apparently want me to read an entire book, or failing that an academic paper, in order to figure out what your complaint about it actually is.

And when I remark that this doesn't do a good job of capturing my interest, you respond by nit-picking my explanation of why not.

Perhaps you had a convincing argument available, but you have exhausted my patience without ever bothering to make it, so I no longer care. Bye!

WaelAugust 18, 2016 8:11 PM

@Gweihir,

The rest is far less determined by free will and insight into things than they like to think they are and very often just follow a pattern without actually making any real decisions.

Can you elaborate more on that? Sounds interesting.

AlanSAugust 18, 2016 8:38 PM

@ Jeremy

A personal cause? Hardly. This sort of thinking has been quite imperialistic and has attempted to colonize vast areas of social sciences but, as I alluded to in my original posts, there are many social scientists and philosophers who have resisted and with good reason.

The problem with game theory, rational choice, neoclassical economics, etc. is that they explain everything and nothing. As Sahlins and Hodgson note, it's like pre-Copernican astronomy. Just add a few more cycles, and voila! it fits! If you strip human action of all cultural context you can have wonderful mathematical models but the models have very little bearing on the complex messiness of actual lived social relationships in particular places and times. They are parlor games.

This isn't to say that such thinking isn't without interest. As an anthropologist might observe, the notion of self-interested, utility maximizing, rational individuals is itself cultural and historically contingent.

Homo LudencrousAugust 18, 2016 9:19 PM

This is a scientific breakthrough of enormous explanatory power for all us typical average humans who pursue a univalent objective. Mine is sex. Oh, and weed. Oh, and money, unless obtaining it interferes with sex or weed or beer. And hanging out with my buds when there's no sex or money to be had but there is weed or beer or sometimes luls. Oh, and freedom and equity and social justice, especially if it helps with the sex unless, wait, uh,.

chris lAugust 19, 2016 1:31 AM

Games aren't just used by people who like to play games and game theorists. They're commonly used in engineering to make decisions with too many free parameters look as if there's an objectively best choice. They're typically set up as weighted evaluations where performance against parameter X gets weight A, performance against parameter Y gets weight B, etc., but sometimes also as a large set of "is this better than that" (analytic hierarchy process). The real value is that they make you at least consciously think about your biases, because if you understand the game you can get any result you want (sometimes by modifying the weighting). Unfortunately there are a lot of people who think that they actually get objectively correct answers from them, sometimes trying to hide the game aspect through a bunch of obfuscated internal computations.

Green SquirrelAugust 19, 2016 3:35 AM

While I am certainly not an opponent of Game Theory, I do think that overall it has only limited applicability to modelling human behaviour.

The PD may be a very valid way to model a clearly defined encounter, few if any real world interactions between people are so genuinely limited in scope.

CassandraAugust 19, 2016 3:57 AM

Dividing people into arbitrary types and drawing conclusions on that basis is a pretty old trait - think of the Ancient Greek concept of the four humours (Phlegmatic, Sanguine, Melancholic, Choleric). Type indicators (such as the well known Myers-Briggs) are not universally regarded as useful, and don't really work well for people who lie on the boundaries - i.e. demonstrating that, in general, there is a spectrum of behaviour of which the types are an attempt at describing an hoped-for bimodal distribution with 'typical' manifestations.

You also have to be careful with Social Science research, as it is riven with inappropriate use of statistics - I haven't the time to carefully review this paper, but if the conclusions depend on p-values exceeding 0.05 with very little content on why certain things are significant, my spidey sense will start tingling. It is worth reading the American Statistical Association's statement on use of p-values.

https://www.amstat.org/newsroom/pressreleases/P-ValueStatement.pdf

Finding correlations is easy. Finding meaningful correlations is less so.

If you are interested in such things, read about Nick Brown's take-down of the highly-cited Fredrickson-Losada "Positivity Ratio" - I won't give URLs, it is easy enough to find, and it makes for entertaining reading.

RealLifeAugust 19, 2016 9:09 AM

Real life prisoner's dilemma -- two gamblers who found a software flaw in video poker:

"The old gambling buddies had one more game to play together. It was the Prisoner's Dilemma. Without speaking, they both arrived at the optimal strategy: They refused the offer. A few months later, the Justice Department dropped the last of the charges, and they were free."
https://www.wired.com/2014/10/cheating-video-poker/

jerAugust 19, 2016 10:51 AM

Well, that was a waste of my time. Going back to work on my Schroedinger's cat inside Einstein's lift experiment. I'm so close to figuring out its gender.

rAugust 19, 2016 3:01 PM

@All,

http://www.economist.com/news/economics-brief/21705308-fifth-our-series-seminal-economic-ideas-looks-nash-equilibrium-prison (sorry about repeating but including HN commentary links this time around)
https://news.ycombinator.com/item?id=12320683

In the above link economist article there's a line:

"Employees competing to impress their boss by staying longest in the office will encourage workforce exhaustion."


Is lightly covered here:

https://codewithoutrules.com/2016/08/18/productive-programmer/
https://news.ycombinator.com/item?id=12320188

Both I picked up from HN

Leave a comment

Allowed HTML: <a href="URL"> • <em> <cite> <i> • <strong> <b> • <sub> <sup> • <ul> <ol> <li> • <blockquote> <pre>

Photo of Bruce Schneier by Per Ervland.

Schneier on Security is a personal website. Opinions expressed are not necessarily those of IBM Resilient.