Rats Cooperate in Prisoners' Dilemma | Hacker News Reader

Rats Cooperate in Prisoners' Dilemma

blog.the-scientist.com

37 pointsmhb15 comments

"The authors of the PLoS study noted that when experimenters observed low cooperation rates, the animals had been food deprived. Fully satiated rats, on the other hand, freely cooperated and easily solved the Prisoner’s Dilemma. These results show that the primordial drive for food in a hungry animal simply clouds judgement."

Rather than say "hunger clouds judgement", it's more likely that hunger leads to short-term optimization, and the prisoner's dilemma is all about long-term optimization where you will usually get another chance to cooperate or not.

If you're hungry, and you cooperate and get chumped... well you may not eat again. ;)

Maybe this is why some people who are content with their lives, happy to not constantly need more, feel like we should collaborate more than compete. Those who are greedy and need more, more, more are more typically hard capitalists. Those who have enough are more socialist.

I have more than I need and I don't mind sharing. Bill Gates doesn't mind sharing. Warren Buffett doesn't mind sharing. But the bankers who are continuing to take money even when they have billions in the bank lobby congress to make it harder for credit card holders to escape debt with bankruptcy.

We've replaced the food hunger with power hunger. There's plenty of food to go around. Plenty of space. But our hunger for money, power, and fame causes irrationally selfish behavior.

I think this is completely wrong. Bill gates and warren buffet don't mind sharing, but they do want you to honor your contracts with them and purchase your goods. Just because they don't lend people money means that they those who cannot afford their goods just don't purchase them. Whereas banks when banks expect you to pay your debts they are "greedy and don't want to share". Also don't forget Buffet is a banker too, how much of goldman sachs and other banks does he own?

"Studies conducted in other labs previously concluded that rats didn’t grasp how to succeed in the Prisoner’s Dilemma. The authors of the PLoS study noted that when experimenters observed low cooperation rates, the animals had been food deprived. Fully satiated rats, on the other hand, freely cooperated and easily solved the Prisoner’s Dilemma. These results show that the primordial drive for food in a hungry animal simply clouds judgement."

The same goes for employees by either (1) feeding them or (2) making sure they eat.

Either way, I consistently noticed a significant increase in cooperation for tasks at hand when either (1) I fully paid for food--like pizza--to share with the team, or (2) informed everyone I was ordering food and to tell me what they want and collected money.

The only difference in behavior observed between the two examples above is when I performed #1, there was always an increase in the number of "what else can I do" or "do you need anything" and more frequent inquiries about the nature of progress which would then lead to the "what else..." and "do you need..."

I was under the impression that in Prisoners' Dilemma, the punishment for the "sucker" was supposed to be more extreme than the punishment if both prisoners rat on each other.

Yeah, this is not the Prisoner's Dilemma! The way they described it the table looks like this:

  Player 1, Player 2

     C     D
    ____ ____
  C|1, 1|0, 1|
  D|1, 0|0, 0|

In this case, both cooperate and defect are Nash equilibrium strategies, which is NOT the case for a Prisoner's Dilemma.

In order for this to be the Prisoner's Dilemma, only defect must be in NE, see: http://en.wikipedia.org/wiki/Prisoner%27s_dilemma

The actual payoff matrix the experimenters used is here: http://www.plosone.org/article/slideshow.action?uri=info:doi...

It does follow the rules of prisoner's dilemma, as far as I can tell.

(Found via the link to the original paper, http://www.plosone.org/article/info:doi/10.1371/journal.pone...)

It is, but if both remain silent they get a smaller penalty than if both rat on each other.

>rat on each other.

Pun intended?

"Past experiments hinted that humans are the only species capable of figuring out how to win the game"

What about flocks of birds and other cooperative groups?

I believe the key here is "figuring out" - choosing a particular behaviour in a setup not found in natural setting. If a particular behaviour is hardcoded in DNA (e.g. bees building hexagonal cells), you cannot say it's been figured out.

Well, maybe by the evolution but not the animal.

The usual explanation of those behaviors is that they are "emergent", i.e. individual agents are "dumb" but the dynamics of the whole system leads to complex behaviors. Another example is ant colonies.

Human experimenters are full of themselves. See recent experiments in self awareness in crows / apes / etc.

The vivesectors would have it that it's ok to use animals because they are lower forms of life and yet assert that doesn't affect the results!

I'd really like to see some sort of setup where rat is pitted against man and other species.

Why duplicate reality?

Reply on news.ycombinator.com