What's Trust Got to Do With It? The Prisoner's Dilemma in Criminal Plea Negotiations by M. Nycole Hearon


The prisoner’s dilemma is a game theory concept that has become the standard for the non zero-sum game.”[1] It is a theory that presents one situation, with multiple options, that are either beneficial or detrimental to the involved parties. The classic prisoner’s dilemma generally presents in the criminal procedure context. Codefendants are separated and placed in different holding rooms to prohibit any collaboration.  The district attorney then tells each defendant that there are two options—to confess or not to confess. If one defendant confesses, the defendant confessing will go free and the other defendant will receive the full sentence. If the defendant does not confess, he will not go free, but will receive a small sentence for a minor charge.  If both defendants confess, each will receive minor punishment.[2] The hypothetical would appear on paper as thus:

Screen Shot 2013-09-03 at 9.07.33 AM

Under the prisoner’s dilemma theory, the decision each defendant makes to confess or not confess depends on how much he trusts his codefendant. Both defendants would benefit by just remaining quiet; however, if there were strong distrust between codefendants, then each would try to get the least amount of time possible by confessing. 

I first learned of the prisoner’s dilemma game theory in a negotiations class. It intrigued me instantly.  As a former prosecutor, I dealt with plea negotiations all day, every day.  In preparing the plea deals, I rarely took into account whether codefendants had any level of trust for each other.  Among other considerations, e.g., whether the defendants were in custody, the severity of the crime, and the likelihood of the case making it to trial, largely influenced my decisions on plea offers. In addition, opposing counsel and the judge presiding over the case had some influence. Frankly, there was no sense in pushing a case that would eventually plea out on the courthouse steps.  In cases with codefendants, if each had similar criminal backgrounds, they would receive the same offer. The offers were usually contingent on both accepting, and in some cases, one codefendant refused to accept the plea. This did not complicate the matter overmuch. If one codefendant agreed, the deal would remain and the other codefendant’s plea agreement would serve as leverage against the refusing codefendant in terms of evidence. Occasionally, I would request that one codefendant testify against the other for a lesser sentence.  Again, in making the offers, I never considered how much trust existed between the codefendants—I considered how much they did not want to be behind bars.

In criminal pleas, whether the other codefendants are trustworthy is rarely a dispositive factor.  Experience in the system, the presiding judge, the prosecutor on the case, underlying motivations of the defendants, defendants’ knowledge of the prosecutor’s underlying motivations, and whether the defendants are repeat players in the criminal system play the larger roles. Simply put, the original prisoner’s dilemma, which focuses on whether two defendants trust each other enough to cooperate to receive the best possible outcome in terms of sentencing, is better suited to academic parlor conversation than real life complex scenarios. 

 What Price Freedom?

In a study conducted in 1974 based on codefendants in robbery cases, researchers Brian Forst and Judith Lucianovic noted that neither the plea rate, conviction rate, nor incarceration rate was significantly higher in cases involving codefendants than in other cases.[3] The researchers’ premise, based on the prisoner’s dilemma theory, was that a prosecutor can use leverage against the codefendants to convince one or the other to testify against the other codefendant.[4]  The simple hypothesis is that a prosecutor would exploit the advantage gained by one confessor to obtain convictions on more serious charges by way of plea-bargaining.[5] Thus, in cases with codefendants, there should have been evidence of more plea bargains and a higher rate of conviction in these cases than in other similar cases where there was only one defendant.[6] Forst and Lucianovic focused their study on cohort criminal cases prosecuted in Washington, D.C. Of the 601 cases in which defendants were convicted, 227 of the cases involved codefendants.[7] The results of the convictions were compared with 1,894 cases in which solo defendants and codefendants had robbery arrests.[8]

The researchers found that plea bargaining was not more successful, or used more extensively, in cases involving codefendants than cases with solo defendants, whether in guilty pleas or guilty verdicts. As point of fact, the percentage of convictions resulting from pleas (74%) was comparable to solo defendant convictions resulting from pleas.[9] In comparison, Forst and Lucianovic hypothesized that prosecutors used plea bargaining leverage in cases where codefendants already had other robbery convictions. The result, however, indicated that the conviction rate was even less for repeat player codefendants.

Forst and Lucianovic added variables to their research, e.g. evidence and eyewitnesses, which would tend to make one defendant testify against the other.[10] Surprisingly, the results showed that the only variable that made a difference was if the prosecutor had tangible evidence that could tie the defendants to the crime; otherwise, the results were still similar to solo convictions gained from plea bargaining.[11] The researchers surmised that the codefendants were more afraid of retaliation from the other codefendant if one decided to testify against each other.[12] What appeared as solidarity to remain quiet between the codefendants was actually fear of bodily, possibly deadly, harm should a codefendant cooperate with the prosecutor.

Forst and Lucianovic’s research appears dated at first, but it is comparable to statistics kept by the United States Sentencing Commission in 2010 (literally over 30 years later) that had similar results. Forst and Lucianovic’s percentage for convictions of defendants who substantially aided the government in Washington, D.C. was 31.9 percent.[13] The rate for Washington, D.C. in 2010 for convictions based on help from defendants was 30.3 percent.[14]  The percentage difference is essentially negligible.  The study supported Forst and Lucianovic’s point that codefendants are likely to refrain from testifying against another codefendant for fear of reprisals from the other codefendant.

 Prosecutorial And Defense Considerations

Recidivism is high in criminal cases and the defendants become repeat players who, by being part of the system so often, are able to understand the resources available, or not available, to the prosecutor and defense counsel. Moreover, the threat of substantial bodily harm, or death, is likely to keep the codefendant silent. Thus, tacit cooperation by codefendants is the norm rather than the exception and their decision is not based on trust.

Whether or not the prosecution employs the prisoner’s dilemma technique depends on several factors, or bargaining chips. Most prosecutors are paid by salary (same pay regardless of the hours they work); thus, a good win-loss ration may not affect their salary per se, but it will affect whether or not they are promoted within the organization.[15] Plea-bargaining, then, benefits the prosecutor because it allows for a “win” ratio—thus ensuring convictions—while allowing prosecutors to pursue other cases that they can likely win in the courtroom. Furthermore, pushing for a trial, even in a strong case, is risky for the prosecutor because juries are highly unpredictable. What appears to be a “slam dunk” on paper may not play out the same in front of a jury. Witness testimony may distort evidence obtained or the judge may rule it inadmissible.[16] With plea negotiations, prosecutors do not have to worry about unpredictable juries and questionable police interrogation tactics.[17] Additionally, plea negotiations can also avert attention from faulty investigations because defense counsel is not cross-examining the arresting officer.[18]

For the public defender, working closing with prosecution and judges makes them more flexible in plea-bargaining.  Because public defenders, like prosecutors, are repeat players, they must yield to some pressure to make plea bargains so that that can get better deals for the clients who have committed crimes that are more serious.[19] This competitiveness of understanding incentives played out between prosecutors, public defenders, and defendants is not unknown or completely bad for the defendant. A defendant who was involved in the criminal system repeatedly better understands the plea-bargaining system more so than newly minted prosecutors and public defenders.[20] Ergo, repeat player codefendants are not worried about trust in making a deal, but they are considering all the factors that make a great plea-bargain a better choice than silence or going to trial.  Codefendants have just as much information available to them as the prosecutor and public defender.[21] But even if the defendant is not aware of witness testimony or expert witness reports, she generally has been through the system enough to know when the prosecutor is “puffing” up the amount of evidence against her to drive a hard bargain.

Defendants bargain against prosecutors’ uncertainty about trial; the self-interest of the prosecutor (win-loss ratio) and the amount of money the state is willing to spend on a case. Additionally, the defendants’ socio-economic status and culture factor into how they bargain. These factors, combined, are what author Stephanos Bibas termed the “structural-psychological perspective” of plea-bargaining.[22]

Uncertainty makes lawyers cling to known information; thus pushing for pleas by anchoring onto past information, such as the defendant’s “rap sheet”, the court docket and which judge will likely preside over the case. Self-interest pushes prosecutors to “under invest in individual cases and push clients to plead.”[23] Lawyers do not want to take the risk of going to trial and somehow damage their reputations if it ends in a loss. The result is a strong push for plea negotiations. Attorney self-interest, consciously or unconsciously, slants plea bargaining decisions.

Money, as a factor, buys the best defense counsel along with the best investigators who have more experience and knowledge. These players also have ongoing relationships with prosecutors.[24] The crushing workload faced by prosecutors and defense counsels puts more pressure on both to plea bargain.  Defendants who go to trial take what money is available; hence, less money is available to prosecute/defend other cases.[25]

The last factor, demographic variation, factors in favor of men, especially those who are overconfident and less intelligent.[26] They end up having better plea deals because they tend to be repeat offenders (usually violent and drug offenders) with some knowledge of the system that allows them to have relevant anchors, (as opposed to unknowns) when making a plea deal.[27] 

Keep It Simple Is Stupid

At the beginning of this discussion, I noted that the prisoner’s dilemma did not fit into a real world application. The dilemma rests on proving that codefendants that cooperate with one another in a non zero-sum situation have the best chance of the best outcome under the circumstances.  The original dilemma, however, is too simplistic. So much so that the theory has been expanded by other theorists and has included extended play by the participants.

The narrowness of the prisoner’s dilemma was discussed in Douglas G. Kelly’s, “Evolutionary Game Theory and the Prisoner’s Dilemma: A Survey.”[28] The original prisoner’s dilemma only allows for one “play”, which means the participants either cooperate or defect.[29] Kelly surmised that if each player saw their choices as presented according to the game (see Introduction, supra), each player would defect.[30]  Defection allowed the most personal gain for each defendant, regardless of trust.  Kelly wrote that Robert Axelrod created the Iterated Prisoner’s Dilemma in which two players (defendants) play ten rounds of the prisoner’s dilemma and the amount of points amassed by each player depended on whether each defected or cooperated with each play.[31] In Axelrod’s example, one player amassed 18 points, while the other obtained 33 points.[32] Had each player been cooperative, each would receive 30 points, but the player with the most points gained by defecting and using the other player’s willingness to cooperate.

Kelly further noted that people confront similar situations in everyday life: Late at night in a subway station, with no one around, why not jump over the turnstile to avoid paying? It saves you fare and costs the system almost nothing.[33] Or, when one traffic lane is closed o the freeway, drivers can increase the flow of traffic by if the drivers alternated entering the one lane.[34]  But as any driver knows, there is always a pile up in the closed lane as defectors force their vehicles into the open lane. Kelly’s argument, and this author agrees, is that the prisoner’s dilemma is never actually enacted in real life; but there is a series of prisoner’s dilemmas in which each player makes gains by cooperating with the potential for punishing the defectors.  Codefendants, especially repeat defendants are highly unlikely to defect. This is not due to a trust of their codefendant, but a trust that forces other than the prosecution will punish their defection. Again, self-interest is the player here for the codefendants and the other codefendants actions are not relevant. Moreover, repeat players can predict with some amount of certainty what resources, and motivators, are available to the prosecution and defense.


The prisoner’s dilemma is a poor tool to determine the behavior of criminal codefendants. It fails to address outside pressures, e.g., social status and monetary incentives, and also fails to take into account repeat players. While it points to the players’ motivation of self-interest, it fails to take into account the motivations of the prosecution or defense counsel.

Furthermore, the only way the game predicts human behavior is by employing a system where the game is repeated over and over again. Statistics show that even when the game is employed in real life situations, it rarely changes the overall outcome, e.g., convictions of codefendants. In real-time context, defendants’ self-interest is also tied to the self-interests of the prosecution and the defense. At best, the iterated prisoner’s dilemma may be used because most criminal defendants are repeat players in the system. Even then, defection is not optional for the defendant, but trustworthiness has nothing to do with it.

[1]Brian Forst & Judith Lucianovic, The Prisoner’s Dilemma: Theory and Reality, 5 J. of L. & Crim. Justice, 55 (1977).  A zero-sum game is one that includes finite possibilities and one player usually ends up with less points than the other.

[2]Id., at 57.

[3]Id., at 59.

[4]Id., at 57.





[9]Id., at 58.

[10] Id., at 60.

[11] Id.

[12] Id., at 61. (Emphasis added.)

[13] Id., at 59.

[14] U.S. Sentencing Comm’n, Sentences Relative to the Guideline Range by Circuit and District, Table 9 (2010), http://www.ussc.gov/Data_and_Statistics/Federal_Sentencing_Statistics/State_District_Circuit/2010/dcc10.pdf.

[15] Stephanos Bibas, Plea Bargaining Outside the Shadow of Trial, 117 Harv. L. Rev. 2463, 2471 (2004).

[16] Id., at 2472.

[17] Id.

[18] Id.

[19] Id., at 2479

[20] Id., at 2481.

[21]Id. at 2494.

[22] Id., at 2527.

[23] Id. Contrast, Rebecca Hollander-Blumoff, Getting to “Guilty”: Plea Bargaining As Negotiation, 2 Harv.  Negot.  L. Rev. 115, 126, n. 53 (1999). Each of the defense attorneys interviewed by the author claimed that their allegiance was to their clients. 

[24] Id.

[25] Id., at 2529.

[26] Id.

[27] Id.

[28] Department of Statistics and Operations Research, University of North Carolina Chapel Hill (2008).

[29] Id., at 3.

[30] Id.

[31] Id., at 7. The Iterated Prisoner’s Dilemma was played in our class, and notably, one group was able to amass the most points by cooperating against another player who defected in each round. The cooperation was realized, however, only after the third play. A straightforward prisoner’s dilemma would not lead to the same results.

[32] Id.

[33] Id., at 4.

[34] Id.

© Conflict Change Consulting Ltd.  2014