The Predictioneer’s Game: Using the Logic of Brazen Self-Interest to See and Shape the Future

by Bruce Bueno De Mesquita

  When my firm advises on a lawsuit, we are always asked how much of the documentation we want to read. Presumably the attorneys hope to give us a reality check, just as they have done for their clients and with themselves. Usually the documentation we are given the opportunity to read is prodigious. There are stacks and stacks of paper. Fortunately, our answer reliably is that we really do not need to read the documents. The merits of the case don’t matter very much once negotiations begin—the merits are inherent in the impasse.

  How can this be? Remember the information we seek in expert interviews. None of it is about how meritorious anyone’s position is. It’s all about calculating how much they care about the result and ferreting out how much they care about getting personal credit. Business managers often care a lot about the result; lawyers often care a lot about credit that results in getting more business and building their reputations for the next suit.

  In the last chapter I introduced one issue that was part of a more complex web of problems confronting a large firm embroiled in litigation with the U.S. Department of Justice. Here was the table I used to reference the scale of outcomes:

  Scale Position

  Meaning

  100

  Multiple felony charges including several specific, severe felonies

  90

  Several specific, severe felony counts but no lesser felonies

  80

  One count of the severe felony plus several lesser felonies

  75

  One count of the severe felony but no other felonies

  60

  Multiple felony counts but none of the severe felonies

  40

  Multiple misdemeanor counts plus one lesser felony

  25

  Multiple misdemeanor counts and no felonies

  0

  One misdemeanor count

  Now I’d like to follow the process through in this example to enable you to see how an outcome can be engineered.

  The second appendix contains the information on the plea issue as obtained from the defendant’s experts, including in-house lawyers, outside lawyers, and corporate executives. Naturally, it is masked to protect anonymity. As is evident from a glance at the appendix, just on this one issue—and there were many others in this litigation—the list of interested parties was extensive. Not only did the defendant have a long list of attorneys and corporate executives with an interest in trying to shape the charges brought against the firm, so too did the community that had been affected by the firm’s actions, as well as various segments of the federal government. Far from being unusual, this is typical of a large, potentially costly, and even devastating litigation.

  The long list of involved parties means the game was much more complicated than anyone could possibly keep straight in their head. Up to now you may have looked at examples and thought “I can work this out in my head,” but no one can work through this complicated a problem without the help of a computer. That is exactly where the added value comes in from having a trustworthy algorithm.

  We’ve all heard stories about evil corporations defrauding people, flouting safety, addicting people to their products, avoiding taxes, polluting the environment, running sweatshops, and God knows what else. That’s what this looked like to me at first blush. The company in question, my client, was accused of having done really terrible things that prompted not only civil action but criminal complaints as well. They were accused pretty much of having destroyed a local community for profit. And yet they seemed like such nice, friendly, soft-spoken, genuinely good people. They had pictures of their children and grandchildren in their wallets. They drove modest cars, ate in normal restaurants, and watched the usual run-of-the-mill TV sitcoms and reality shows. Could the situation really have been as awful as it was portrayed? Could the cast of characters hiring my consulting firm really be the soulless monsters described in the media?

  As is almost always the case, reality was a lot more nuanced and complicated than the charges suggested, and the people involved were not the satanic ogres portrayed in the local press. Terrible things had happened, but it was far from obvious that the company was responsible, culpable, or negligent or that it harbored the slightest bit of intent to do any harm, for profit or for any other reason. In fact, as dramatic as the news stories were about the case, reality was much different, and it seems—as we will see—that the U.S. attorney, if not his or her office, understood that.

  Mind you, I am not trying to justify some of the awful things that happened. My partner and I strongly urged our client to take steps outside the litigation to help the community involved, just out of a sense of concern for the people’s well-being. They welcomed our advice and acted on it. They wanted to do something for the people who were harmed and had already been contemplating some such action, even though their attorneys urged them not to, fearing this would be interpreted as an admission of guilt. Our advice just served to tip the balance in favor of their doing what was right over what was expedient. That decision, however, was related to humane behavior, not to justice.

  Justice requires that we distinguish between bad outcomes and bad intentions or willful ignorance. I don’t believe it’s fair to blame people when things turn out badly unless there is proof (and not just innuendo) that they intentionally chose actions or inaction when a reasonable person could foresee the bad consequences of their decision. It is best to judge people based on what they reasonably could know and expect before they did things, not based on what we know later, after the situation has played itself out. But of course I am no lawyer, so my view of justice may be way off compared to how the American judicial system thinks about things. After all, it is not a lawyer’s job to get at the truth, it is a lawyer’s job to make the best case possible for the client. That, I suppose, is my job too when I’m wearing my consultant’s hat instead of my professor’s hat.

  Anyway, the unfolding drama required a big stage. It involved at least a metaphorical cast of thousands. Still, the final decision process revolved around a few star performers, many of whom were most reluctant to see their names in bright lights. The lead players who longed for anonymity included the board of directors of my client’s firm, some of whom were pretty actively engaged in discussions over how to handle the issues; the president and the CEO of the relevant unit of the corporation; and the senior in-house attorney. The senior outside attorneys were also crucial players in the unfolding drama.

  On the other side, the U.S. attorney, his/her staff, some of the line attorneys in the Department of Justice and in ABC (a government agency whose name is masked to ensure anonymity for the client), the head of the relevant local government, and plaintiff’s counsel didn’t mind seeing their names in lights at all; in fact, a few of them relished the thought. They were star players as well.

  Getting the lead actors to agree on a settlement was the task at hand. Otherwise, the cases were going to go to trial. Probably the corporation and its representatives would have come out pretty well in terms of a judgment, but not before they had been dragged through the mud day in and day out during the trial. That was the scenario expected by the client. It wasn’t a pretty picture. They had been working on the drama’s script for several years with no sense of progress but with deep concerns that catastrophe lay just around the corner.

  Keenly aware of the aura of doom and gloom that pervaded all discussions, my partner and I set out to define the issues and to turn the model loose in order to get a first impression of the lay of the land. We were curious about whether the situation was as hopeless as the client thought. The model’s initial estimate of what would happen—our weighted mean and median are the same in this case—equals 60 on the outcome scale. Sixty is equivalent to pleading guilty to multiple felony counts not including any of the severe felonies. This initial prediction was viewed as good news by the defendant, a ray of sunshine in their overcast view of the situation.

  The firm’s most senior executives—not just bit players
in the company—were looking at having to plead guilty to at least one count of a severe felony as well as several lesser felonies. So the initial estimate revealed the possibility of a better-than-expected outcome for the client. That was the good news. But every silver lining needs a cloud, and this was no exception. The simulation of the negotiating game that followed from that initial estimate bore out the defendant’s gloomy expectation. The initial estimate was more optimistic than the model’s conclusion after it had simulated the consequences of the predicted interactions among the players. Remember that the changes from initial positions predicted in the game led also to a more accurate prediction about the expected final decision. In other words, the motivations and power of the plaintiffs and their cause suggested that my client would lose ground as negotiations wore on. The aura of doom and gloom was quick to return.

  As figure 6.1 shows, my model predicted the negotiations would follow a complex path, first looking very encouraging and then turning sour. This figure illustrates what I meant earlier about this game being like multidimensional chess. If all that the players cared about was getting the result they advocated, this would be no harder than a round-robin chess tournament. But egos enter into negotiations and so enter into the negotiating game. Some players will take big risks to try to win big. Others are more concerned about not losing than they are about winning. That means figuring out which players are choosing their moves to get the plea they favor, which are picking moves that will get them the most credit for finding a settlement (or blocking one), and which are ambivalent about these competing desires. Chess isn’t this messy. Imagine trying to win at chess when the rules for winning change with each opponent, as they do in the negotiating game!

  FIG. 6.1. The Unengineered Plea Path

  According to the model, the give-and-take in settlement discussions would persist through eight meetings between the defendant’s representatives and the U.S. attorney or his/her representatives. At the end of the eight exchanges of views and arguments, the game indicates that the cost of continued negotiations was no longer going to be worth the small changes in the expected agreement. That agreement was just about at 80 on the plea scale—that is, one count of the severe felony plus several lesser felonies.

  The result anticipated by most of my client’s lawyers and expected by the board, and the plea predicted by the end of the game, were the same. If my model couldn’t improve on this, then my consulting partner and I would have nothing to offer the firm that hired us. We would be just another expense.

  However, the game uncovered something not anticipated by most of the attorneys (one senior outside attorney had the ultimate result dead right from the outset, although he/she did not quite see why it would arise or how to get there). The model’s simulation indicated that early discussions with the U.S. Attorney’s Office and others would suggest that a powerful coalition of interests was going to form around a lesser set of charges—almost like a gambit in chess, designed to suck the client and the U.S. attorney into a move that would be used against them later. The figure, which displays the predicted upshot of discussions at each stage in the negotiations, anticipates that the second and third presentation of arguments by the defendant’s side were likely to soften the U.S. attorney enough so that he/she would contemplate agreeing to 50, a better mix of multiple misdemeanor and felony counts not including a severe felony. This was, in fact, the position believed to be held by the U.S. attorney at the outset of negotiations, according to our interviews with the client’s representatives and attorneys. However, the initial analysis also showed that with subsequent rounds of discussion, the U.S. attorney could be and would be persuaded to take a tougher, not a softer, stance.

  Why the move to a tougher stance after showing openness to a more moderate settlement? The model indicated that the hard-line attorneys in ABC and in the office of the U.S. attorney were going to execute a gambit. The ABC attorneys and the Department of Justice lawyers apparently were out to make names for themselves as tough guys who could bring evil corporations to their knees. The analysis suggested that they thought this was the perfect case to do it. Their gambit aimed to set up the U.S. attorney so that he/she would reveal softness early in settlement discussions. Then later they would pounce on this softness in order to embarrass the U.S. attorney politically, compelling him/her to take a tougher stance when it counted, in the end, lest he/she be tarred publicly with the ignominious tag of being soft on business malfeasance. That would not have played out well in the affected community or in the Department of Justice. To be sure, these lawyers were really rolling the dice, taking a big risk if their gambit failed. But they had every reason to think it would succeed—and probably it would have, had my client not been using a tool like the negotiating game.

  Even though (according to the model) the U.S. attorney was willing to go for a mix of misdemeanors and lesser felonies, the logic showed that the hard-liner gambit would work. He/she would abandon a moderate stance, choosing instead to go for severe felonies. This was the U.S. attorney’s way of solving the tough choice between pursuing the outcome he/she thought was right and following the gambit’s path, thereby avoiding careerist costs and keeping the support of others in the Department of Justice and in the affected community. Seeing that the U.S. attorney had significant political exposure and careerist ego involvement in the case, it became apparent that we could find ways to gratify his/her ego and bollix the hard-liners’ gambit.

  The simulations uncovered several interesting patterns that opened the opportunity to engineer the outcome. First, as noted, the U.S. attorney took a tough stance according to the simulations because of pressure from attorneys at ABC and within the Department of Justice and in response to arguments from several stakeholders in the affected community. They were strongly committed to their point of view, and the U.S. attorney cared about being perceived by them as interested in helping them get justice as they saw it. This was particularly interesting because the U.S. attorney’s own view of what could be supported as punishment through the judicial process was considerably less. The U.S. attorney’s view involved none of the severe felonies and showed openness to at least some misdemeanors. The U.S. attorney, facing hard-line pressure, was willing to trade off some sense of legal correctness for political correctness and its attendant personal credit.

  To engineer a better result, I simulated what might happen if the defendants altered their bargaining position from that of being prepared to accept numerous misdemeanors and maybe one lesser felony, a posture predicted to end in their caving in to one or more severe felonies. I looked at what would happen if they offered more concessions up front and also if they offered less. I also looked at how they could maneuver to get some important hard-liners to make arguments that would make those hardliners look foolishly extreme to the U.S. attorney, turning the gambit on its head. In examining alternative strategies I took advantage of another insight gained from the base analysis: the U.S. attorney tilted more toward eagerness to make a deal than to sticking to a particular position. Also, it was evident that the defendant’s strategy had to create leverage against the pressure exerted by the hard-liners who wanted a plea involving severe felonies.

  It turned out that the best strategy for the defendant involved two shifts from the approach they had planned as reflected in the data they gave me about themselves. First, the one outside counsel who favored pleading guilty to one lesser felony and numerous misdemeanors needed to convey unity with the rest of the defendant’s team in endorsing a plea to misdemeanors only. Although this one attorney had the right settlement in mind—that was the ultimate agreement on this aspect of the case—he/she could not so much as hint at this flexibility during the initial meetings with the U.S. attorney, and he/she did not.

  This attorney acted out the scripted part perfectly. Jack Nicholson, great as his performance was in A Few Good Men, paled in comparison to the performance of the attorney who had to fake a commitment to a position he/she did not r
eally believe in. Since he/she led negotiations on this issue, his/her ability to be convincing was critical, and convincing he/she was.

  Of course, getting an attorney or anyone else to act contrary to their beliefs is no small order. It takes great faith that the model’s logic should be allowed to trump personal intuition. As an old client used to say when introducing me to his colleagues, “Check your intuition at the door.” The greatest value of a model is when it provides an insight that is contrary to the decision makers’ expectations—when it correctly urges them to check their intuition at the door. It takes a courageous person to defy one’s own beliefs and follow the lead derived from a computer model, since after all we never know what is or is not correct until after the fact. All we know is the model’s track record for accuracy (but then everyone thinks his problem is unique) and whether the logic for the proposed action is persuasive. Fortunately, in this case the outside attorney being asked to change his/her approach had worked with my firm before on other cases. In fact, this attorney is the very person who persuaded the client to use my company’s services. He/she had seen the model, as he/she put it, “work its magic” before, so this attorney had no problem agreeing to act out the part as written by the model.

  The second maneuver that was required was considerably more challenging to “sell” to the defendant. The company’s directors were naturally very concerned about this matter and were eager to find a solution. The simulations—remember, all of this analysis is happening before discussions with the U.S. attorney have begun—showed that the directors were so fearful of how the case was likely to unfold that they would cave in to the mounting pressure from the opposition hard-liners by agreeing to numerous felony counts including one count of the severe felony—that is, the hard-liners’ chess gambit was going to work. To produce no severe felonies, getting instead multiple misdemeanors and one lesser felony as the plea, it was necessary to control the reaction of the board when the U.S. attorney pressed hard for an outcome the board was prepared to live with. The strategy for them was simple to articulate but hard to do: they had to take the position that they would not negotiate or authorize any discussion of felonies at all, risking the ire of the U.S. attorney and a breakdown in discussions.