My previous post concerned a new experimental decision study by Kenju Kamei and me in which groups of subjects played games of voluntary contribution to joint accounts, interacting with partners of their own selection. One purpose of the study was to re-confirm that having a choice of partners encourages individuals to invest in reputations for cooperativeness. The other was to go beyond past studies by investigating what participants would take away from their past experiences—experiences we gave them by letting them play several self-contained multi-period games. Specifically, some participants might see signs that they should have cooperated more at the outset of their earlier games to attract better partners. Wiping the slate clean and letting all start afresh might lead these participants to be more cooperative the second than the first time around. A countervailing factor is that if the first game had a known final period—as was indeed the case—then quite a few who had been acting cooperatively up until then might have switched to non-cooperation in the final period. Seeing these “end game” defections by others might discourage previously trusting subjects from cooperating when a new series of plays began, or at least it might encourage them to try to “defect” earlier than others next time around.
Our experiment included six different treatments, generated by varying both (i) how beneficial mutual cooperation is relative to mutually selfish behavior, and (ii) how much information about one’s past play is conveyed at each partner choice stage. With respect to (i), in three treatments, subjects could earn 70 percent more each period by fully cooperating than by not cooperating at all, while in three others, the potential gain from cooperation was only 30 percent. With respect to (ii), two of the treatments offered “100 percent information”; that is, when you were deciding what ranks to give each of five potential partners, you were being shown the average level of cooperation of each of them so far in the phase (the self-contained series of interactions) and the players were also identified to you by one digit ID numbers that stayed the same throughout the phase before randomly changing when the next one began. In two other treatments, each period had only a 50 percent chance of being counted in the “past average contribution” shown at later periods’ partner choice stages. And in the last two treatments, there was no (i.e., 0 percent) chance that potential partners’ past average contributions would be shown, so subjects could rely only on what they knew about specifically ID’d individuals by having played with them during a period within the current phase.
The main takeaway of the experiment was that when information was good (either 50 percent or 100 perccent) and when returns from cooperation were high (70 percent gains), participants engaged in more cooperation in later than they did in earlier phases—i.e., their experiences taught them to be more cooperative.
Toward the end of the previous post, I suggested that modern life might resemble a series of such phases or sets of games because people may live in different places and hold different jobs at different points in their lives. When you have ongoing interactions with people—including relationships as simple as dealing with a plumber or housepainter or car mechanic—they have an incentive to act in a trustworthy and fair manner so as to get your repeat business and have you recommend them to friends. But when they know you’re about to move away, or when they themselves are preparing to retire and close up shop, they might do less careful work or overcharge, or they may worry that you won’t pay for the work. Some of these “end game problems” are avoided in ordinary life simply because one doesn’t know in advance when the last period of an interaction has arrived. Other end game problems are hard to avoid; for instance, how can you motivate a spouse to look after you if you become disabled or suffer from dementia? Recognizing that such problems could arise leads us to look for partners who we hope will be intrinsically motivated, and to try to build emotional bonds strong enough to withstand temptations to neglect us in the lurch.
Of course, our simple laboratory decision games, in which players are kept essentially anonymous to each other and only a modest amount of money is at stake (typical earnings totaled around $25), can hardly capture the sorts of emotional bonds that form in real life. Still, it’s worth noting that even in the treatments in which subjects could know nothing of past behaviors except those of their own past partners during the current phase (our “0 percent reputational information” condition), there were pairs of partners who managed to form ongoing bonds with one another and to achieve levels of cooperation comparable to those in treatments with better information.
A less close analogue to the situation modeled by our experiment would involve not individuals moving from setting to setting within a given lifetime, but the passing on of insights gleaned by one generation to the members of the next one. If your grandparents learned from their experiences that it paid in the long-run to be trustworthy and to do their fair share in interactions with others, then they may have tried to pass that lesson along to your parents who, if they had found their own experience also supportive, may have likewise attempted to pass it along to you. What kind of norms are being passed along will depend on whether we live in the kind of world in which trustworthiness and cooperation pay off, or one in which the trustworthy and cooperative are regularly taken for suckers. The answer could vary by time and place. In some societies—today’s Denmark?—children’s upbringings may emphasize trust and cooperation, which become self-fulfilling expectations, whereas in others—Somalia?—the corresponding message space may be filled by cautions to trust only close kin, with correspondingly negative impact on behavior.
We may be almost as far from knowing how to help grow a normatively healthy society today as we are from knowing how to grow a healthy brain or to prevent one from developing dementia. But even though both today’s social science and today’s biological science will look primitive fifty years from now if scientific progress continues, thoughtful observers then may look back at work in both fields and find ways in which we’re now laying the groundwork for the advances of their day. Our research already hints strongly that understanding how trusting and cooperative orientations or their opposites are nurtured need not remain beyond human reckoning forever.