Pri­soner’s Dilemma

The Pri­soner’s Dilemma is a well-stud­ied game in game the­ory, where sup­pos­edly ra­tio­nal in­cen­tive fol­low­ing leads to both play­ers stab­bing each other in the back and be­ing worse off than if they had co­op­er­ated.

The origi­nal for­mu­la­tion, via Wikipe­dia:

Two mem­bers of a crim­i­nal gang are ar­rested and im­pris­oned. Each pris­oner is in Soli­tary Con­fine­ment with no means of com­mu­ni­cat­ing with the other. The pros­e­cu­tors lack suffi­cient ev­i­dence to con­vict the pair on the prin­ci­pal charge, but they have enough to con­vict both on a lesser charge. Si­mul­ta­neously, the pros­e­cu­tors offer each pris­oner a bar­gain. Each pris­oner is given the op­por­tu­nity ei­ther to be­tray the other by tes­tify­ing that the other com­mit­ted the crime, or to co­op­er­ate with the other by re­main­ing silent. The pos­si­ble out­comes are:
If A and B each be­tray the other, each of them serves two years in prison
If A be­trays B but B re­mains silent, A will be set free and B will serve three years in prison
If A re­mains silent but B be­trays A, A will serve three years in prison and B will be set free
If A and B both re­main silent, both of them will serve only one year in prison (on the lesser charge).

The “stay silent” op­tion is gen­er­ally called Co­op­er­ate, and the “be­tray” op­tion is called Defect. The only Nash Equil­ibrium of the Pri­soner’s Dilemma is both play­ers defect­ing, even though each would pre­fer the co­op­er­ate/​co­op­er­ate out­come.

A pop­u­lar var­i­ant is the Iter­ated Pri­soner’s Dilemma, where two agents play the Pri­soner’s Dilemma against each other a num­ber of times in a row. A sim­ple and suc­cess­ful strat­egy is called Tit for Tat—co­op­er­ate on the first round, then on sub­se­quent rounds do what­ever your op­po­nent did on the last round.

See also: Co­or­di­na­tion/​Cooperation

In­tro­duc­tion to Pri­son­ers’ Dilemma

Scott Alexander
30 Jun 2012 0:54 UTC
35 points
5 comments5 min readLW link

2014 iter­ated pris­oner’s dilemma tour­na­ment results

30 Sep 2014 21:23 UTC
64 points
57 comments6 min readLW link

Pri­son­ers’ Dilemma with Costs to Modeling

Scott Garrabrant
5 Jun 2018 4:51 UTC
173 points
20 comments7 min readLW link

Pri­soner’s Dilemma Tour­na­ment Results

6 Sep 2011 0:46 UTC
105 points
171 comments11 min readLW link

The True Pri­soner’s Dilemma

Eliezer Yudkowsky
3 Sep 2008 21:34 UTC
106 points
114 comments4 min readLW link

Ro­bust Co­op­er­a­tion in the Pri­soner’s Dilemma

7 Jun 2013 8:30 UTC
73 points
146 comments7 min readLW link

Clas­sify­ing games like the Pri­soner’s Dilemma

4 Jul 2020 17:10 UTC
73 points
22 comments6 min readLW link

Black­mail, Nukes and the Pri­soner’s Dilemma

10 Mar 2010 14:58 UTC
20 points
20 comments2 min readLW link

Co­op­er­at­ing with agents with differ­ent ideas of fair­ness, while re­sist­ing exploitation

Eliezer Yudkowsky
16 Sep 2013 8:27 UTC
48 points
45 comments4 min readLW link

A Differ­ent Pri­soner’s Dilemma

14 Apr 2018 15:54 UTC
25 points
1 comment5 min readLW link

Real World Solu­tions to Pri­son­ers’ Dilemmas

Scott Alexander
3 Jul 2012 3:25 UTC
46 points
88 comments7 min readLW link

Pri­soner’s Dilemma (with visi­ble source code) Tournament

7 Jun 2013 8:30 UTC
51 points
235 comments2 min readLW link

The Epistemic Pri­soner’s Dilemma

18 Apr 2009 5:36 UTC
37 points
46 comments2 min readLW link

Pri­soner’s dilemma tour­na­ment results

9 Jul 2013 20:50 UTC
32 points
124 comments1 min readLW link

The con­tinued mi­suse of the Pri­soner’s Dilemma

23 Oct 2009 3:48 UTC
29 points
70 comments2 min readLW link

Paper: Iter­ated Pri­soner’s Dilemma con­tains strate­gies that dom­i­nate any evolu­tion­ary opponent

2 Jun 2012 20:50 UTC
27 points
19 comments2 min readLW link

Fixed-Length Selec­tive Iter­a­tive Pri­soner’s Dilemma Mechanics

13 Sep 2011 3:24 UTC
25 points
14 comments15 min readLW link

Pri­soner’s Dilemma on game show Golden Balls

21 Apr 2012 0:31 UTC
24 points
32 comments1 min readLW link

The Coun­ter­fac­tual Pri­soner’s Dilemma

21 Dec 2019 1:44 UTC
20 points
9 comments2 min readLW link

The Truly Iter­ated Pri­soner’s Dilemma

Eliezer Yudkowsky
4 Sep 2008 18:00 UTC
20 points
85 comments1 min readLW link

Pri­soner’s Dilemma as a Game The­ory Laboratory

25 Aug 2011 14:30 UTC
17 points
47 comments3 min readLW link

[LINK] Can­tor’s the­o­rem, the pris­oner’s dilemma, and the halt­ing problem

30 Jun 2013 20:26 UTC
13 points
9 comments1 min readLW link

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

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

Reflex­ive Or­a­cles and su­per­ra­tional­ity: pris­oner’s dilemma

24 May 2017 8:34 UTC
12 points
1 comment4 min readLW link

Pri­soner’s Dilemma vs the Afterlife

24 Sep 2013 16:59 UTC
11 points
69 comments2 min readLW link

Other pre­spec­tive on re­solv­ing the Pri­soner’s dilemma

4 Jun 2013 16:13 UTC
11 points
34 comments1 min readLW link

Another Iter­ated Pri­soner’s Dilemma Tour­na­ment?

25 May 2012 14:16 UTC
11 points
23 comments1 min readLW link

The True Epistemic Pri­soner’s Dilemma

19 Apr 2009 8:57 UTC
9 points
72 comments2 min readLW link

New pris­oner’s dilemma and chicken tournament

14 Sep 2011 8:00 UTC
6 points
13 comments2 min readLW link

Agent-Si­mu­lates-Pre­dic­tor Var­i­ant of the Pri­soner’s Dilemma

15 Dec 2015 7:17 UTC
5 points
34 comments2 min readLW link

Iter­ated Pri­soner’s Dilemma in soft­ware patents

22 Jul 2013 20:22 UTC
3 points
8 comments2 min readLW link

Pavlov Generalizes

20 Feb 2019 9:03 UTC
68 points
4 comments7 min readLW link

Player of Games

29 Aug 2018 21:26 UTC
54 points
3 comments12 min readLW link

Re-for­mal­iz­ing PD

28 Apr 2009 12:10 UTC
30 points
63 comments2 min readLW link

[Question] A way to beat su­per­ra­tional/​EDT agents?

Abhimanyu Pallavi Sudhir
17 Aug 2020 14:33 UTC
6 points
13 comments1 min readLW link

An­nounc­ing the 2014 pro­gram equil­ibrium iter­ated PD tournament

31 Jul 2014 12:24 UTC
24 points
63 comments1 min readLW link

Most Pri­soner’s Dilem­mas are Stag Hunts; Most Stag Hunts are Bat­tle of the Sexes

14 Sep 2020 22:13 UTC
101 points
18 comments9 min readLW link