Starting Over

Does life tend to harden us and make us more cynical and self-serving, or do some of us, at least, learn from experience that a little kindness helps ourselves as well as others?

In the early years of experimental economics, some were skeptical that the cooperative behaviors often displayed by university student subjects were representative of a broader population including older, non-student adults. Perhaps students at a given school share a sense of camaraderie. And maybe students are more idealistic and fair-minded than older adults.

Yet most attempts to see whether the “pro-social” behaviors displayed in economics experiments by university students would stand up when the subjects were “grown-ups” found that those older subjects were if anything more pro-social (more trusting, more cooperative) than the student ones. Something about their experiences seemed to be teaching people to be nicer, not more selfish, with the passage of time.

This post concerns one recent attempt to see what people take from their experiences of social interaction.

In my earlier post “When Nice Guys Finish First,” I described a decision experiment in which subjects interact in a dilemma situation called a voluntary contribution game. In each of a series of periods or rounds, each subject receives several units of experimental currency and has to decide what to keep for herself and what to put in a group fund. The dilemma is that all group members earn more if all put all of their money into the group fund, but if what others have put in is taken as given, each earns more the more she keeps for herself. Since the experiment prevents the subjects from entering into enforceable agreements, traditional economic reasoning predicts that self-interest will drive them toward a worse outcome than is available to them via moral norms, trust, or cooperation.

The “Nice Guys Finish First” experimental treatment that my research partners and I had conducted was one in which the subjects were given a say over who they would interact with as the game progressed. Virtually everyone wanted to have as partners those who contributed the most to the group fund, but since partner choice was mutual, cooperators ended up playing together and earning more than the more selfishly-behaving types.

While this was one of the first experiments to rigorously establish the benefits of earning a cooperative reputation, and while it seemed to illustrate well why having a say over who we deal with helps motivate us to adopt better behavior, it left at least two interesting questions unanswered. First, the subjects who had been less cooperative at the outset weren’t necessarily less nice. They may simply have been less optimistic that others would cooperate. But the design of the experiment made it difficult for them to overcome the consequences of their initial pessimism, because their early low contributions to their group fund remained part of the public information about them for the duration of their session, making it hard for them to compete for the most cooperative counterparts. We wondered whether a wiping clean of the slate might have led such individuals to compete for a good reputation from the outset, having learned that it pays to be nice.

An opposing force might be at work, though, if all subjects first finished a finite number of interactions whose end they were all well informed of. This would be an effect of what we call “end game” behavior. Clever individuals might act cooperatively to share the benefits of cooperation with like-seeming partners for as long as new partner choices remained possible for all, but when the last such choice had been taken and there was no remaining incentive to be nice, they might “cash in” on their reputations by putting nothing in the project and enjoying the benefits of any more selfless or naïve group members they could make “suckers” of. If a brand new game of the same type began, without any carrying over of reputation from the first one, wouldn’t the witnessing of such chicanery not lower trust by causing subjects to try to “defect” from cooperation one step ahead of one another?

My former student and now collaborator Kenju Kamei, presently of Bowling Green State University, recently conducted a new set of experiments with me to investigate whether either or both of these forces—learning that it pays to be cooperative, and learning to get the jump on defectors—would be at work if something similar to the old “Nice Guys” treatment were played in several sequences with fully clean breaks. We simplified matters by having the subjects play in the smallest possible groups, that is as pairs. Each had nine potential partners during 40 periods of play, and in each single period expressed a preference among a randomly selected five individuals from among those nine. The 40 periods were broken up into four independent sequences of 10 periods each. A subject’s “reputation” stayed with her from period 1 to 10, 11 to 20, and so on, but after every ten periods, the slate was wiped clean by having the computer scramble identifying information as to who was who and show past history from within the current ten period sequence only (for example, in period 24, only your average contribution in periods 21 through 23 would be known by those rating you as a partner).

In our new working paper (short title “Play it Again”), we report that within each ten period sequence, play closely resembles that in the “Nice Guys” experiment (Page et al., “Voluntary Association...” The Economic Journal, 2005), at least in those treatments in which there is a large potential payoff from cooperation and in which information on what one has done thus far in the sequence is fully available. And in those treatments, both of our two suspicions about what might happen if individuals could start fresh received some support, but the first one especially. That is, rather than learning with experience that it’s best to try to exploit others, most subjects seemed to be learning that investing in a reputation for cooperativeness paid off. Average contributions to the group fund—called “joint account” given that groups contained two players only—actually rose from one sequence to the next. But this was mainly the case in the first seven periods of each sequence. “Cashing in” by free riding on a cooperating partner tended to start a little earlier in the later than in the earlier sequences. Despite that countervailing force, most subjects were contributing positively to their joint account in all but the very last period of the sequence, in those treatments in which information and potential cooperative gains were high. Some even contributed their entire endowment in a known last period, in fact.

Does modern life in any way resemble a series of “finitely repeated games”? It may, in the sense that we might move from one to another nearly non-overlapping world as we progress through it. We may spend our college years in a different social circle than those of high school, then experience a series of moves to new communities and jobs. As we near the last of our interactions with a given circle, are we tempted to give up our pretense of niceness, secure that our reputation won’t follow us? Or might we have taken niceness aboard as an internalized trait that we hesitate to depart from out of concern for our own self-image? These questions are of course far too big to answer in so brief a space, or to fully answer in any case with such stylized and simple studies. But we think our hundreds of subjects’ hours, and the hundreds of hours we ourselves spent analyzing and writing up their behaviors, do shed some light on these matters. Please stay tuned for a “Play it again” sequel in this same space.