Topic: The kill/trade game - page 7. (Read 9332 times)

In all honesty, if he doesn't reply soon, we may have to move on without him, with me taking his place.

I should have specified a time limit for replies. 

Nah, in game 1 you cray cray so I don't trust you either.

You no kill me too?

Cameltoemcgee can't play any more. I killed him, remember? 

It would be hard to track rationality in a mathematical sense, on the individual level.  On societal or ecosystem level the amount of balance required for some things is quite amazing.

I'm not clear on what you mean.

People (the market) tend to make the best choices for themselves.  Rational self-interest.   Not always true of course.

I mean "rational" in a mathematical sense.

Riiiiiggghhttt. It is still a game, just like R/P/S, so... Sure, kinda.

Why do you say this?  I see plenty of instances of irrationality, no doubt, but I'm not sure it's right to say that people almost never act rationally in the real world.   If that was the case, the world would be almost completely chaotic.   

You can, that's the point There is pretty good proof that behaving rationally, and particularly superrationally, especially in a perfect information game, is most societally beneficial, and typically pretty individually beneficial as well. But, ironically, people in real life indeed almost never act according to what we could describe as rational, and certainly not superrational.

I'm a criminologist and a sociologist, and I agree

p.s. count me in the next game

But what if some of us try to play with irrational tactics? Or give image of such thing...

On other hand disinformation and appearing random is valid tactic maybe... Who knows what I'm up to

Let's see how this works out...

Nothing you have said is incorrect. I do not disagree or feel the need to argue.

I guess I should state that, while perhaps debatable since posts can be "strategic lies", in my view at least, once a player states their rules of play with a level of specificity that describes all iterations of the turns at hand, it becomes not a matter of rationality; rather, those rules can be integrated as part of the game, and treated as merely information.

In the interests of returning to the game at hand, let us all assume that anyone on board has already (or should have already) read at *least* the Wikipedia entry on game theory, so we can avoid further soliloquies on who is or isn't actually actively involved in the field we lovingly call game theory.

Erm, no we don't. Perfect and imperfect information refer to the knowledge (or lack thereof) of the state of the game, not the state of the other player's mind. Iterated prisoner's dilemma is a game of perfect information, as every player has full knowledge of all previous moves and what moves are available for the opponent to make (as opposed to, say, Battleship, where you don't know where your opponent has placed his pieces).

What you're talking about is superrationality, which is where a player assumes that, since he and his opponent are both completely rational and have access to the same information and have the same decisions to make, they will both make the same move. The reason the prisoner's dilemma is a dilemma is that a "rational" player will not make the best decision, since no matter what the opponent does, it is always better to defect; this leads to the conclusion that both players will always defect, which is the worst possible outcome. That's the traditional theory, anyway. A superrational player, on the other hand, will rule out outcomes in which the opponent makes a different move, so the only possibilities to consider are the cases where both players defect or both players cooperate. Here, cooperation is the best move, and the dilemma is resolved. Yay.

Hmm. The behavior that would model is punishing someone who attempts to extract excessive (in my opinion) retribution in return for a previous crime. Now, the Tit for Tat strategy models, roughly, the NAP. The NAP, itself, is silent on retribution. However, it is my personal interpretation that retribution is not particularly just, nor is it necessary. Therefore, I would kill a grudger if put up against one. Again, I might sometimes forgive, especially if asked.

A clarification: communication between players is allowed. Deals never to kill each other model contracts, promising not to kill a player in the next turn and then doing so models fraud, and so on. Just as in the real world, public contracts are more enforceable than secret ones, so keep that in mind if someone PMs you offering to make a secret alliance.

Excellent. In accordance with the rules, I will step out until another player comes in, making it odd again. The updated Round 2 roster:

Ekaros <--> Rassah
Cameltoemcgee <--> FenixRD
FCTaiChi <--> Elwar
Foxpup <-->wdmw

We're just waiting on you and Cameltoemcgee.

Can't literally speak for Myrkul, but *my* rules might work something like: since I state I will kill only in a tit for tat situation, I will kill anyone who "deserves" it, in the sense that they have previously killed and not been brought to "justice".


However, these are not my rules. Because game theory societally works best under perfect information, if not always individually, rules are as follows:

I will kill anyone who deserves it, in accordance with the principles of NAP as I interpret them. Further detail on my interpretation of the NAP must be acquired in-game.

I feel it is worth noting that if applied concepts in game theory (such as the prisoners dilemma) stop working under the conditions of the participants (one, some, all) being aware of the field and its implications, it merely indicates either the theory was incomplete already, or must be adjusted for situations such as the above one where one/some/all know the theory. In the field we call this perfect and imperfect information.

Stated differently, it doesn't matter whether people know or don't know the underlying theories and rules.

Yeah, also, I'm gonna play.

Can you clarify this? By "killer", do you mean one who killed in the previous round, or one who has ever killed? That is, would you kill a Grudger (one who always kills against an opponent who has previously killed, even after the opponent stops killing)?

I don't know if the programs knew the outcomes or not (they were paired off for entire games, instead of switching around like we are). But of course we all know the outcomes of these games, so there's no point in pretending we don't. If an opponent killed in the last round (unless they killed a killer), or killed me last time we were paired, I will kill. A slightly better strategy is "Tit for Tat with forgiveness," so I might occasionally forgive, or trade with a killer, anyway. This allows a killer to occasionally benefit from immediately rehabilitating.

Did the programs know of the outcomes of other games in iterated prisoners dilemma?  Is that different here?  If so are you playing tit for tat socially or individually?