Fog Creek Software
Discussion Board




The Prisoner's Dilemma

There have been a few threads on honesty & trust. That reminds me of the Prisoner's Dilemma'. For the uninitated, there is a description here  http://plato.stanford.edu/entries/prisoner-dilemma/

My interpretation (which will I will be slaughtered for) is this:

Mutual trust usually pays off. If you don't know a person and you have to interact with them, your most productive course of action is to trust them until they prove untrustworthy and never trust them again if you are 'betrayed'. Obviously, you don't bet the farm on that first interaction. You can take this a stage further. If someone proves untrustworthy, remove them completely from your sphere of influence with as much vigour as possible within appropriate ethical and legal constraints.

Grifters play this game, of course. They seed your trust with a small payoff and then hit you up for your life savings, so I don't simply interpret 'person' in a literal way. I tend to group situations and classes of behaviour too.

Works for me. Comments ? Flames ? Practical experiences ?

Woodentongue
Friday, March 19, 2004

I believe this is called common sense.

masterOfTheObvious
Friday, March 19, 2004

"They seed your trust with a small payoff and then hit you up for your life savings, so I don't simply interpret 'person' in a literal way."

It's interesting you mention this in light of the threats on the Apprentice. Many lauded the generous actions of Nick (I think that's his name -- not good with names) when he gave back the entire $250 advertising fee because one half of the signs fell off for one half of the day. While many think this would create a sense of trust, my impression is precisely the opposite: It was such an unreasonable, unbusinesslike "gift" that I think anyone with any business acumen would either have no faith in the business abilities of the individual, or it would lay a seed of distrust as they wait for the other shoe to drop and the con to be pulled.

.
Friday, March 19, 2004

>> "Works for me. Comments ? Flames ? Practical experiences?"

I hereby flame thee!

Adam N
Friday, March 19, 2004

But seriuosly, you can easily rig the payoffs in the prisoner's dilemma in such a way as to make backstabbing pay for some.  Welcome to corporate America, baby!!!

Adam N
Friday, March 19, 2004

For what it is worth I would advice against going to deep into game theory. The knowledge you will gain will not make you a happy puppy.

Just me (Sir to you)
Friday, March 19, 2004

Not to get off topic, but Nick knew exactly what he was doing... He wasn't the PM for that week, and it was a good bet that he wouldn't get fired for "doing what he thought was the ethical thing to do". On the other hand, he built a relationship with the client for when he may need their business in the future.

As far as ethics in the workplace, just do whatever lets you sleep well at night. Personally, I like to think that I have a competence level in my profession that enables me to act with dignity.

genius
Friday, March 19, 2004

"On the other hand, he built a relationship with the client for when he may need their business in the future."

My point was precisely the opposite -- in the real world, where people are jaded and cynical (AKA realistic), he basically _sabotaged_ the relationship rather than build it. In the business world people expect predictable, fair behaviours where they can understand your motivation. Whenever someone acts unexplainably in your favour, you either lose respect for their abilities/product (which is why desperation is deadly in sales), or you become extremely skeptical of everything they do in the future.

.
Friday, March 19, 2004

"Whenever someone acts unexplainably in your favour, you either lose respect for their abilities/product (which is why desperation is deadly in sales), or you become extremely skeptical of everything they do in the future.
"

I agree. I had a somewhat similar reaction to Nick's action.  The owner was willing to take much much less.

Giving someone TOO much too easily makes them wonder why you gave in so easily.

I.e., if Nick readily gave back the minimum he had to, then that looks expedient: please the customer asap and move on. Likewise, if there as a lot of haggling and Nick *finally* relented and gave back more than he had to, then it looks like the customer wore him down. (I think this is not as productive as the former example).

BUT... if Nick readily give back more than he has to, it might make the customer suspicious.

Mr. Analogy
Friday, March 19, 2004

I'd love to have seen the reactions if Nick's team had lost by $75. I'm pretty sure Nick would've been fired then.

Giving a full refund is a long-term strategic action. This is not a long-term strategic game.

Philo

Philo
Friday, March 19, 2004

Actually, the really silly thing about Nick giving the guy cash back, is that the guy asked for a *credit*!  He didn't ask for cash back, and it didn't sound like he wanted a full credit, just credit for time not used.

So Nick wasted time going over there, when he could've given the guy a credit right then and there on the phone.  Meanwhile, depending on how Trump's people accounted for sales, they might have gotten full credit for present-day sales of future services.  (After all, the other team got credit for selling future rides.)

Even in the worst case, they would've only been out the portion of sales that were credited, not the full amount.  Nick's actions just seemed like strategic self-aggrandizement to me.

Phillip J. Eby
Friday, March 19, 2004

This all seems like a case of people who don't live in the real world deciding what ought to be done. About that prisoner's dilemma. You don't rat. You don't become a fink. It's a darn good principle that needs no approval from an ivory tower. About trusting. You only trust as much as is reasonable. You never completely trust unless you have no other options.  No one will every be able to program a computer to make these decisions. Never. This is why character counts.

Me
Friday, March 19, 2004

There's a point at which you turn human behaviour into a mathematical formula. At that point it's time to go play with your children, or sit in the sun on a beach or in a bar and look at attractive women. It's time to reconnect with your body.

Statistics in human behaviour are only important when the outcome is statistical, which it should never be. Human interaction should be messy and emotional, and you should love every minute of it.

www.MarkTAW.com
Friday, March 19, 2004

I love you, man!!!

Jack of all
Friday, March 19, 2004

>Mutual trust usually pays off. If you don't know a person and you have to interact with them, your most productive course of action is to trust them until they prove untrustworthy and never trust them again if you are 'betrayed'.

<Me_Putting_on_my_philosophers_hat>

You have just advanced a very old argument, (at least 150 years old).

http://www.objectivistcenter.org/articles/rdonway_dostoevsky-nietzsche-ayn-rand.asp

These old style socialists/anarchists (see Chernyshevsky, Bakunin) have voiced these opinions ages ago.

The problem (apart from others that socialism/anarchy has with human nature) is that this argument is insufficient to construct a general modell of human behavior.

You see, your argument works ok, if other, rational minded people agree with it. If people do not (motivated by things as the dynamics of group behaviour, nationalism, bigotry and other 'features of human nature') then things do not work out as planned.

</Me_Putting_on_my_philosophers_hat>

Michael Moser
Saturday, March 20, 2004

Woodentongue:-
I agree. You gotta start a relationship with co-workers placing trust in them. But if they fail you, you have to ensure that can't happen again and the best way to do that is to cut them out till they can earn your trust again (if possible)

nakedCode
Saturday, March 20, 2004

There was a program some time ago that tested a wide variety of strategies to this game, and the one that won was amazingly simple - tit for tat.  It starts out being nice, and then just does whatever the other player did most recently.  It didn't always win, but it had the highest average score among the many strategies tested.

Just an interesting (I thought) tidbit.

Aaron F Stanton
Saturday, March 20, 2004

The summary of 'mutual trust pays off' is true, but only partially true. The real summary is that mutual trust pays off if you are likely to further interact with the same party. Otherwise, you're best off acting in your own self-interest.

A good book on how game theory applies to human behavior is Axelrod's 'The Evolution of Cooperation.'

A reader
Sunday, March 21, 2004

Tit-for-tat not only "doesn't always win" -- it *never* wins in any single series of interactions with an individual. So how can it win overall? Because, for a wide variety of strategies X, when tit-for-tat plays X *both* tit-for-tat and X do better than X generally does in interaction with other strategies.

The "real summary" is correct when considering the nice clean isolated interactions that occur in theoretical studies like Axelrod's tournaments. Out in the real world, though, you can have a reputation. So the issue is not only "Will I be interacting with this individual again?" but "Will I be interacting again with this individual, or with anyone who might have heard about my interactions with this individual?".

Gareth McCaughan
Monday, March 22, 2004

I agree with woodentounge.

A lot of people argue that I am naive because of my willingness to do business on the strength of a handshake.

I have lost both professionally and personally due to this misguided trust, but I like to think it was money/effort well spent.

You screw me over, and you reveal your true colours. Cheaper to find out now than in x years time when the stakes could be much higher.

Tapiwa
Monday, March 22, 2004

Good points. Doesn't The Prisoner's Dilemma assume each person is acting in an informational vacuum? I.e. no information about the other player except what moves they take.

Now, some of the alternative methods concocted for the tournament may have included a measure of reputation, but cannot include intuition because intiution is based on variables completely outside of the scope of the theory. (Actually, I could debate both sides of this, but let's say intuition is completely unquantifiable).

Put another way, and in terms we can all understand (ha ha, I joke).  Would you be surprised if Nick betrayed somene in the Boardroom? What about Troy? Do you think Troy would betray Kwame? What exactly do you base your opinions on? Entirely quantifiable past decisions based on the players, or something less definable?

IMHO you'll be tempted to explain your stance using facts, but it'll be a gut decision, the facts just back it up, and if you had the opposite gut decision, you could also find facts to back it up.

Or to put it another way, The Prisoners Dilemma is played like Joker Poker, playing the odds with no interaction. Real Poker is played face to face, and there's a good reason for it.

www.MarkTAW.com
Monday, March 22, 2004

> Doesn't The Prisoner's Dilemma assume each person is acting in an informational vacuum?

The "Tit-for-tat" strategy assumes that I can recognise you, and remember the results of our most recent previous transaction: so that if you betrayed me last time then I will non-cooperate with you this time.

Christopher Wells
Monday, March 22, 2004

'Tit-for-tat' differs from the OP's "and never trust them again if you are 'betrayed'": because if you cooperate this time when Tit-for-tat non-cooperates, then Tit-for-tat will cooperate next time: Tit-for-tat only has a short-term, not a long-term, memory.

In the long run, Tit-for-tat can be betrayed one more time that it is betrayed: but, because it's forgiving, it can end up in a long-term mutually profitable cooperative relationship even with an entity that betrayed once near the begining, if its counterpart makes ammends by cooperating (suffering one betrayal in restitution) and then continuing to cooperate.

Another aspect of Tit-for-tat is that its first move (when it has no prior information) is "cooperate": therefore instances of Tit-for-tat permanently cooperate with each other.

Tit-for-tat is said to be a robust strategy.

Christopher Wells
Monday, March 22, 2004

In the prisoners dilemma you have no reliable feedback as to what the other guy is doing.

In real life you rely on your lawyer to make the decisons for you :)

Stephen Jones
Monday, March 22, 2004

In real life do a number six on any one that wants a lawyer to be morally responsible for him.

Tapiwa
Monday, March 22, 2004

> The "Tit-for-tat" strategy assumes that I can recognise you

Yes, you operate in a vacuum except for the completely binary nature of the other person's decision. Have you ever been talked into doing something you didn't want to? It could never happen in The Prisoner's Dilemma.

www.MarkTAW.com
Monday, March 22, 2004

Ethics or morals Tapiwa have nothing to do with it.

That's why you get a lawyer to make the decisions for you :)

Stephen Jones
Tuesday, March 23, 2004

*  Recent Topics

*  Fog Creek Home