RSS

New­comb’s Problem

New­comb’s Prob­lem is a thought ex­per­i­ment in De­ci­sion The­ory ex­plor­ing prob­lems posed by hav­ing other agents in the en­vi­ron­ment who can pre­dict your ac­tions.

From New­comb’s Prob­lem and Re­gret of Ra­tion­al­ity:

A su­per­in­tel­li­gence from an­other galaxy, whom we shall call Omega, comes to Earth and sets about play­ing a strange lit­tle game. In this game, Omega se­lects a hu­man be­ing, sets down two boxes in front of them, and flies away.
Box A is trans­par­ent and con­tains a thou­sand dol­lars.
Box B is opaque, and con­tains ei­ther a mil­lion dol­lars, or noth­ing.
You can take both boxes, or take only box B.
And the twist is that Omega has put a mil­lion dol­lars in box B iff Omega has pre­dicted that you will take only box B.
Omega has been cor­rect on each of 100 ob­served oc­ca­sions so far—ev­ery­one who took both boxes has found box B empty and re­ceived only a thou­sand dol­lars; ev­ery­one who took only box B has found B con­tain­ing a mil­lion dol­lars. (We as­sume that box A van­ishes in a puff of smoke if you take only box B; no one else can take box A af­ter­ward.)
Be­fore you make your choice, Omega has flown off and moved on to its next game. Box B is already empty or already full.
Omega drops two boxes on the ground in front of you and flies off.
Do you take both boxes, or only box B?

One line of rea­son­ing about the prob­lem says that be­cause Omega has already left, the boxes are set and you can’t change them. And if you look at the pay­off ma­trix, you’ll see that what­ever de­ci­sion Omega has already made, you get $1000 more for tak­ing both boxes. This makes tak­ing two boxes a dom­i­nant strat­egy and there­fore the cor­rect choice. Agents who rea­son this way do not make very much money play­ing this game.

The gen­eral class of de­ci­sion prob­lems that in­volve other agents pre­dict­ing your ac­tions are called New­comblike Prob­lems.

New­comb’s Prob­lem and Re­gret of Rationality

Eliezer Yudkowsky
31 Jan 2008 19:36 UTC
88 points
605 comments10 min readLW link

New­comblike prob­lems are the norm

So8res
24 Sep 2014 18:41 UTC
53 points
111 comments7 min readLW link

Con­fu­sion about New­comb is con­fu­sion about counterfactuals

AnnaSalamon
25 Aug 2009 20:01 UTC
45 points
42 comments2 min readLW link

Null-box­ing New­comb’s Problem

Yitz
13 Jul 2020 16:32 UTC
25 points
10 comments4 min readLW link

You May Already Be A Sinner

Scott Alexander
9 Mar 2009 23:18 UTC
44 points
37 comments3 min readLW link

Coun­ter­fac­tual Mugging

Vladimir_Nesov
19 Mar 2009 6:08 UTC
55 points
295 comments2 min readLW link

A few mis­con­cep­tions sur­round­ing Roko’s basilisk

Rob Bensinger
5 Oct 2015 21:23 UTC
60 points
130 comments5 min readLW link

Self-con­firm­ing pre­dic­tions can be ar­bi­trar­ily bad

Stuart_Armstrong
3 May 2019 11:34 UTC
45 points
11 comments5 min readLW link

You’re in New­comb’s Box

HonoreDB
5 Feb 2011 20:46 UTC
43 points
176 comments4 min readLW link

New­comb’s prob­lem hap­pened to me

Academian
26 Mar 2010 18:31 UTC
43 points
99 comments3 min readLW link

A model of UDT with a halt­ing oracle

cousin_it
18 Dec 2011 14:18 UTC
41 points
102 comments2 min readLW link

Parfit’s Es­cape (Filk)

G Gordon Worley III
29 Mar 2019 2:31 UTC
40 points
0 comments1 min readLW link

New­comb’s Prob­lem vs. One-Shot Pri­soner’s Dilemma

Wei_Dai
7 Apr 2009 5:32 UTC
13 points
16 comments1 min readLW link

Oper­a­tional­iz­ing New­comb’s Problem

ErickBall
11 Nov 2019 22:52 UTC
36 points
23 comments1 min readLW link