Total Posts:20|Showing Posts:1-20
Jump to topic:

TIT for TAT

Yvette
Posts: 859
Add as Friend
Challenge to a Debate
Send a Message
8/5/2010 2:02:59 PM
Posted: 6 years ago
A while back I made a thread about bonobo kindness, and used the title of a book I was just starting to read. Vacation's had me nearly fully through it, and it's fleshed out/reshaped what (I admit as) my previous opinions. I'm still of the same basic opinion (that morals are rooted in evolution) but my understanding is different.

The book (The Moral Animal) built on some solid proof: males need mostly sex and females need mostly resources in order to pass on genes, status provides access to both, human emotions generally (even when not immediately obvious) lead them along the path to reproductive success in the ancestral environment, etc.

The basic idea of the book was this: evolution is capable of creating instincts which drive individuals, and that's what emotions are.

According to the book, computer simulations in one experiment pitted various programs for maximizing access to resources against each other. Cooperation netted the most resources, but cost a little, cheating cost nothing but could get the other program's resources if the other program cooperated. The program with the most resources at the end "won", and the winner was the simplest method: cooperate with everyone the first time you meet them, continue to cooperate until they cheat. IE, if another program cheats on the first meeting, the winning program never cooperated with them again.

Now, imagine this program, called TIT FOR TAT, can communicate the cooperation level of others, and that the programs are capable of punishing cheaters, and you've got yourself a moral system. Performing beneficial acts improves how much people are willing to cooperate with you, regularly cheating does the opposite. Emotions fueling desires for fairness and justice, feelings of rage, guilt, gratitude, affection, etc, generally follow this system.

Some examples from the book include people feeling guiltier about wrongdoing when there can be social consequences/knowledge of the act, feelings of affection increasing towards those who help the individual more, loss of affection once there are better reproduction prospects elsewhere, etc.

Really a great book, I recommend it. I'm not sure where this idea that evolution exists and accounts for the instincts of all animals and some human instincts but not the ones we hold most dear comes from.
In the middle of moving to Washington. 8D

"If God does not exist, then chocolate causing cancer is only true for the society that has evidence for that." --GodSands
tkubok
Posts: 5,044
Add as Friend
Challenge to a Debate
Send a Message
8/5/2010 2:20:10 PM
Posted: 6 years ago
I dont have to read the book, this concept is fairly old. Although its the first time hearing about an actual simulation, but for the most part, most of this is common logic, and the other parts are filled in by sociobiology. Its quite funny too, when you get around to reading some of the behaviors of animals.
Yvette
Posts: 859
Add as Friend
Challenge to a Debate
Send a Message
8/5/2010 3:18:46 PM
Posted: 6 years ago
At 8/5/2010 2:20:10 PM, tkubok wrote:
I dont have to read the book, this concept is fairly old. Although its the first time hearing about an actual simulation, but for the most part, most of this is common logic, and the other parts are filled in by sociobiology. Its quite funny too, when you get around to reading some of the behaviors of animals.

It's an old concept, yea. But it's not a very accepted one, surprisingly. Even Objectivists who accept evolution like Ragnar_Rahl don't accept the concept.
In the middle of moving to Washington. 8D

"If God does not exist, then chocolate causing cancer is only true for the society that has evidence for that." --GodSands
mattrodstrom
Posts: 12,028
Add as Friend
Challenge to a Debate
Send a Message
8/5/2010 3:45:02 PM
Posted: 6 years ago
At 8/5/2010 3:42:09 PM, Puck wrote:
Robotics exercises disagree. :P Robot builds that compete with resources will invariably end up as winner takes all.

are they also competing with wolves?

and other animals?

and the weather?
"He who does not know how to put his will into things at least puts a meaning into them: that is, he believes there is a will in them already."

Metaphysics:
"The science.. which deals with the fundamental errors of mankind - but as if they were the fundamental truths."
mattrodstrom
Posts: 12,028
Add as Friend
Challenge to a Debate
Send a Message
8/5/2010 3:51:45 PM
Posted: 6 years ago
At 8/5/2010 3:45:02 PM, mattrodstrom wrote:
At 8/5/2010 3:42:09 PM, Puck wrote:
Robotics exercises disagree. :P Robot builds that compete with resources will invariably end up as winner takes all.

are they also competing with wolves?

and other animals?

and the weather?

btw... I didn't read the OP with the whole communicating thing... so I was just talking bout resource sharing :(

not the game theory stuff

my bad.
"He who does not know how to put his will into things at least puts a meaning into them: that is, he believes there is a will in them already."

Metaphysics:
"The science.. which deals with the fundamental errors of mankind - but as if they were the fundamental truths."
Yvette
Posts: 859
Add as Friend
Challenge to a Debate
Send a Message
8/5/2010 4:14:13 PM
Posted: 6 years ago
At 8/5/2010 3:42:09 PM, Puck wrote:
Robotics exercises disagree. :P Robot builds that compete with resources will invariably end up as winner takes all.

That was...rather vague, and claimed a lot. Can you try specifying a little? Winner takes all doesn't really happen in a natural environment, cooperation tends to lead to larger rewards than competition, and as the experiment I mentioned showed, within competition, the cooperators are the winners, not the cheaters.
In the middle of moving to Washington. 8D

"If God does not exist, then chocolate causing cancer is only true for the society that has evidence for that." --GodSands
Puck
Posts: 6,457
Add as Friend
Challenge to a Debate
Send a Message
8/5/2010 4:17:18 PM
Posted: 6 years ago
At 8/5/2010 4:14:13 PM, Yvette wrote:
At 8/5/2010 3:42:09 PM, Puck wrote:
Robotics exercises disagree. :P Robot builds that compete with resources will invariably end up as winner takes all.

That was...rather vague, and claimed a lot. Can you try specifying a little? Winner takes all doesn't really happen in a natural environment, cooperation tends to lead to larger rewards than competition, and as the experiment I mentioned showed, within competition, the cooperators are the winners, not the cheaters.

Naw, your claim was that simulations prove co-op was the natural result. Robotic exercises where robots compete for resources, show that cooperation isn't necessarily the natural end product of competition.
tkubok
Posts: 5,044
Add as Friend
Challenge to a Debate
Send a Message
8/5/2010 6:19:26 PM
Posted: 6 years ago
At 8/5/2010 3:42:09 PM, Puck wrote:
Robotics exercises disagree. :P Robot builds that compete with resources will invariably end up as winner takes all.

No no, not until we are capable of programming a robot advanced enough for proper social interaction and capable of learned behavior. In other words, an advanced AI, which no programmer has ever accomplished.
tvellalott
Posts: 10,864
Add as Friend
Challenge to a Debate
Send a Message
8/5/2010 6:45:48 PM
Posted: 6 years ago
At 8/5/2010 6:19:26 PM, tkubok wrote:
At 8/5/2010 3:42:09 PM, Puck wrote:
Robotics exercises disagree. :P Robot builds that compete with resources will invariably end up as winner takes all.

No no, not until we are capable of programming a robot advanced enough for proper social interaction and capable of learned behavior. In other words, an advanced AI, which no programmer has ever accomplished.

AI is getting better and better. I don't know if we will see code that can write it's own code (Think) but I'm sure in a decade we will be seeing programs the likes of which we can only imagine. Look at the progression of video game enemies. Compare old school Doom with MGS3.
My first computer had 4gb of HD space, now I can buy an 8gb USB drive for 10 bucks.
"Caitlyn Jenner is an incredibly brave and stunningly beautiful woman."

Muh threads
Using mafia tactics in real-life: http://www.debate.org...
6 years of DDO: http://www.debate.org...
lastrequest691
Posts: 339
Add as Friend
Challenge to a Debate
Send a Message
8/5/2010 7:29:50 PM
Posted: 6 years ago
Computers are never wrong.

So the Computer Simulation are not wrong.
"That song was absolutely waste of talent; you sounded like a wounded animal and who told you to play the guitar by yourself." Simon Cowell
Puck
Posts: 6,457
Add as Friend
Challenge to a Debate
Send a Message
8/5/2010 7:32:01 PM
Posted: 6 years ago
At 8/5/2010 6:19:26 PM, tkubok wrote:
At 8/5/2010 3:42:09 PM, Puck wrote:
Robotics exercises disagree. :P Robot builds that compete with resources will invariably end up as winner takes all.

No no, not until we are capable of programming a robot advanced enough for proper social interaction and capable of learned behavior. In other words, an advanced AI, which no programmer has ever accomplished.

At least in some exercises I've seen, the robots did 'learn' as to what constituted more viable methods for obtaining energy at the expense of competitors.
SportsGuru
Posts: 1,648
Add as Friend
Challenge to a Debate
Send a Message
8/5/2010 7:41:42 PM
Posted: 6 years ago
At 8/5/2010 7:29:50 PM, lastrequest691 wrote:
Computers are never wrong.

So the Computer Simulation are not wrong.

code is only as good as the programmer, aka someone who can fully be wrong

if simulations were never wrong, wars would end MUCH quicker
Yvette
Posts: 859
Add as Friend
Challenge to a Debate
Send a Message
8/5/2010 10:56:37 PM
Posted: 6 years ago
At 8/5/2010 4:17:18 PM, Puck wrote:
At 8/5/2010 4:14:13 PM, Yvette wrote:
At 8/5/2010 3:42:09 PM, Puck wrote:
Robotics exercises disagree. :P Robot builds that compete with resources will invariably end up as winner takes all.

That was...rather vague, and claimed a lot. Can you try specifying a little? Winner takes all doesn't really happen in a natural environment, cooperation tends to lead to larger rewards than competition, and as the experiment I mentioned showed, within competition, the cooperators are the winners, not the cheaters.

Naw, your claim was that simulations prove co-op was the natural result. Robotic exercises where robots compete for resources, show that cooperation isn't necessarily the natural end product of competition.

That's...not actually what I, or the book, claimed at all.

It claimed that cooperation proved to be the most effective method as opposed to just screwing over everyone around you. And the way of evolution is that those who are most successful pass on their genes. Cooperation isn't necessarily the natural end product of competition, it's just the most effective way to deal with it, and the TIT FOR TAT method is not only generally used by humans but one that was shown to do it most effectively while minimizing risk.
In the middle of moving to Washington. 8D

"If God does not exist, then chocolate causing cancer is only true for the society that has evidence for that." --GodSands
belle
Posts: 4,113
Add as Friend
Challenge to a Debate
Send a Message
8/6/2010 4:57:21 AM
Posted: 6 years ago
At 8/5/2010 10:56:37 PM, Yvette wrote:
At 8/5/2010 4:17:18 PM, Puck wrote:
At 8/5/2010 4:14:13 PM, Yvette wrote:
At 8/5/2010 3:42:09 PM, Puck wrote:
Robotics exercises disagree. :P Robot builds that compete with resources will invariably end up as winner takes all.

That was...rather vague, and claimed a lot. Can you try specifying a little? Winner takes all doesn't really happen in a natural environment, cooperation tends to lead to larger rewards than competition, and as the experiment I mentioned showed, within competition, the cooperators are the winners, not the cheaters.

Naw, your claim was that simulations prove co-op was the natural result. Robotic exercises where robots compete for resources, show that cooperation isn't necessarily the natural end product of competition.

That's...not actually what I, or the book, claimed at all.

It claimed that cooperation proved to be the most effective method as opposed to just screwing over everyone around you. And the way of evolution is that those who are most successful pass on their genes. Cooperation isn't necessarily the natural end product of competition, it's just the most effective way to deal with it, and the TIT FOR TAT method is not only generally used by humans but one that was shown to do it most effectively while minimizing risk.

the important part that i think you're leaving out are the conditions in which such a situation holds. if both cooperate, theres a small reward for both. if one cooperates and the other doesn't, the defector gets a huge reward and the lone cooperator gets nothing, or even a punishment. and if they both defect they get nothing. tit for tat is the most successful strategy in that situation that anyone has ever been able to invent. i am guessing with puck's robots, the incentives weren't set up that way at all (ie there was no punishment for defecting). that would change things completely. since cooperation is so helpful to people, punishing defectors for screwing us over takes on more significance with us.
evidently i only come to ddo to avoid doing homework...
belle
Posts: 4,113
Add as Friend
Challenge to a Debate
Send a Message
8/6/2010 4:58:04 AM
Posted: 6 years ago
also is this the richard wright book? i was tihnking about reading it but his other book (non-zero) kind of annoyed me
evidently i only come to ddo to avoid doing homework...
tkubok
Posts: 5,044
Add as Friend
Challenge to a Debate
Send a Message
8/6/2010 5:20:26 AM
Posted: 6 years ago
At 8/6/2010 4:57:21 AM, belle wrote:
the important part that i think you're leaving out are the conditions in which such a situation holds. if both cooperate, theres a small reward for both. if one cooperates and the other doesn't, the defector gets a huge reward and the lone cooperator gets nothing, or even a punishment. and if they both defect they get nothing. tit for tat is the most successful strategy in that situation that anyone has ever been able to invent. i am guessing with puck's robots, the incentives weren't set up that way at all (ie there was no punishment for defecting). that would change things completely. since cooperation is so helpful to people, punishing defectors for screwing us over takes on more significance with us.

But youre ignoring consistency. If two people constantly cooperate, they will produce consistently positive results. If one person screws the other over, sure, one of them will get better results than the other, but then, there would be no "Next time". Its like a Russian defecting to the USA. Once you do that, you cannot defect back to Russia. Its a one time thing. When we talk about cooperation vs screwing someone over, in the long run, you will probably gain more through cooperation.
Yvette
Posts: 859
Add as Friend
Challenge to a Debate
Send a Message
8/6/2010 3:03:57 PM
Posted: 6 years ago
At 8/6/2010 4:58:04 AM, belle wrote:
also is this the richard wright book? i was tihnking about reading it but his other book (non-zero) kind of annoyed me

Robert Wright, yea. Haven't read his other book but this one was good.
In the middle of moving to Washington. 8D

"If God does not exist, then chocolate causing cancer is only true for the society that has evidence for that." --GodSands
belle
Posts: 4,113
Add as Friend
Challenge to a Debate
Send a Message
8/6/2010 3:17:54 PM
Posted: 6 years ago
At 8/6/2010 5:20:26 AM, tkubok wrote:
At 8/6/2010 4:57:21 AM, belle wrote:
the important part that i think you're leaving out are the conditions in which such a situation holds. if both cooperate, theres a small reward for both. if one cooperates and the other doesn't, the defector gets a huge reward and the lone cooperator gets nothing, or even a punishment. and if they both defect they get nothing. tit for tat is the most successful strategy in that situation that anyone has ever been able to invent. i am guessing with puck's robots, the incentives weren't set up that way at all (ie there was no punishment for defecting). that would change things completely. since cooperation is so helpful to people, punishing defectors for screwing us over takes on more significance with us.

But youre ignoring consistency. If two people constantly cooperate, they will produce consistently positive results. If one person screws the other over, sure, one of them will get better results than the other, but then, there would be no "Next time". Its like a Russian defecting to the USA. Once you do that, you cannot defect back to Russia. Its a one time thing. When we talk about cooperation vs screwing someone over, in the long run, you will probably gain more through cooperation.

this too. in one-off games the best strategy is to defect. its only if you play over and over with the same person/people that cooperating pays off.
evidently i only come to ddo to avoid doing homework...