RickJS comments on Ingredients of Timeless Decision Theory - Less Wrong

43 Post author: Eliezer_Yudkowsky 19 August 2009 01:10AM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Article Navigation

Comments (226)

Comments (226)

Sort By: Best

Show:

You are viewing a single comment's thread. Show more comments above.

Comment author: Eliezer_Yudkowsky 19 August 2009 03:24:46PM * 7 points [-]

Not by my standards.

Free will is about as easy as a problem can get and still be Confusing. Plenty of moderately good reductionists have refused to be confused by it. Killing off the problem entirely is more like dropping nuclear weapons to obliterate the last remnants of a dead horse than any great innovation within the field of reductionism.

There are non-reductionist philosophers who would think of reducing free will as a great and difficult achievement, but by reductionist standards it's a mostly-solved problem already.

Formal cooperation in the one-shot PD, now that should be interesting.

Comment author: RickJS comment score below threshold [+] (7 children)

Comment author: SilasBarta 09 September 2009 05:37:45PM 5 points [-]

While a good question, Eliezer_Yudkowsky has already thoroughly answered it in The True Prisoner's Dilemma.

His point there is, the values in the matrix are supposed to represent the participants' utility, rather than jail time, which accounts for your compassion for your friend. If it were simply prison sentences, your reasoning would apply, which is why EY says the true Prisoner's Dilemma requires convoluted, unusual scenarios, and why normal presentations of the PD don't make the situation clear.

Comment author: Furcas 09 September 2009 05:36:17PM * -1 points [-]

That Prisoner A is completely and utterly selfish is part of the Prisoner's Dilemma. If the prisoner's not selfish, it's not the Prisoner's Dilemma anymore.

EDIT: Of course, this is only true if the numbers in the matrix represent years spent in jail, not utilons.

Comment author: thomblake 09 September 2009 05:40:35PM -2 points [-]

inorite?!

Of course, this might still be muddy if you recast the payoff matrix in utilons, or (to abstract away less) adjust the "external" payoff matrices so that the "internal" payoff matrices match those of the original problem.

Comment author: RickJS 10 September 2009 02:05:11AM 0 points [-]

Inorite? What is that?

I suspect I'm not smart enough to play on this site. I'm quite unsure I can even parse your sentence correctly, and I can't imagine a reason to adjust the external payoff matrices (they were given by Wei Dai, that is the original problem I'm discussing) so the internal payoff mtrices match something. I'm baffled.

Comment author: Cyan 10 September 2009 02:41:45AM 1 point [-]

"inorite".

Comment author: thomblake 10 September 2009 01:05:47PM * 0 points [-]

See Cyan's comment below. Do not be dispirited by lolspeak.

Also, the reason to adjust the payoff matrices in the original problem is so that your 'internal' payoff matrices match those of Wei Dai's problem, or to put it another way, consider the problem in the least convenient possible world. Basically, the prisoner's dilemma is still there if you take the problem to be in utilons, which take into account things like your 'compassion' (in this case, valuing the reward given to the other person). I can't quite figure out what your formula for discounting is above, so let me simplify...

It would be remiss for me to not do the math, though it is not my forte:

Suppose the matrix represents jelly beans for you or the opponent, each worth 1 utilon. Further suppose that you get .25 utilons for each jelly bean the opponent gets, due to your 'compassion'. Now take this payoff matrix (in jellybeans):

375/500 -150/600 600/0 75/100

Which becomes in your 'internal' matrix (in utilons):

500/500 0/600 600/0 100/100

Now cooperation is dominated by defection for the 'compassionate' person.

Someone please note if my numbers don't work out - it's early here.

Comment author: RickJS 10 September 2009 11:23:12PM 1 point [-]

Ah. Thanks! I think I get that.

But maybe I just think I do. I thought I understood that narrow part of Wei Dai's post on a problem that maybe defeats TDT. I had no idea that compassion had already been considered and compensated out of consideration. And that's such common shared knowledge here in the LessWrong community that it need not be mentioned.

I have a lot to learn. I now see I was very arrogant think I could contribute here. I should read the archives & wiki before I post. I apologize.

<<Begins to compute an estimated time to de-lurk. They collectively write several times faster than I can read, even if I don't slow down to mull it over. Hmmm... >>

RickJS comments on Ingredients of Timeless Decision Theory - Less Wrong

Article Navigation

Comments (226)

Nearest Meetups

Virtual Study Room

Recent Comments

Recent Posts

Recent Rationality Quotes

Recent Wiki Edits

Top Contributors, 30 Days

Recent Karma Awards

Comments (226)

Top Contributors, 30 Days

You'll need to login or register to do that

(Don't worry, it only takes a few seconds)

Create

Login