Evolution of Cooperation in the Laboratory
Evolution of Cooperation in the Laboratory William M. Baum and Peter J. Richerson University of California, Davis One way to validate the theory that cultural evolution is a Darwinian process is to produce cultural evolution in the laboratory. To demonstrate cultural evolution, one should show variation, transmission, and selection of behavior in groups. One earlier experiment showed the feasibility of studying evolution of traditions in groups (Baum, Richerson, Efferson, & Pacciotti, 2004). Another earlier experiment showed that written anonymous messages were effective in increasing contributions in a public-goods game (Baum, Pacciotti, Richerson, Lubell, & McElreath, 2012). We studied microsocieties that played a public-goods game in one generation and left advice for the next generation. The groups could punish anyone who contributed too little to the public good, and we varied the cost of punishment from zero (free), to 1 cent, and to 5 cents. Punishing was altruistic in the sense that it was costly to the individual but beneficial to the group. Coordinated altruistic punishment evolved across generations when the cost was low (1 cent) and not when the cost was high (5 cents). METHOD Participants: 71 groups of 10 undergraduates at UC Davis, 18 yo or older. One-time participation only. Procedure: (adapted from Chaudhuri et al., 2006) Before each session, instructions were read aloud describing the public-goods game, the method and cost of punishment, the possibility of writing messages to be read aloud to the group, and the leaving of advice for the next group. All contributions, punishment, and messages were anonymous, because all actions occurred inside cardboard boxes (see Figure 1). The game lasted for 10 rounds, contributions to the public account from zero to 50 cents were allowed on each round, and punishment could be stipulated for anyone contributing less than a set criterion. If a participant chose to punish, anyone who contributed less than the criterion was fined 10 cents. (Thus, several people had to punish for the punishment to be effective.) The sum of all contributions to the public account was doubled and then distributed equally among the 10 participants. The sum, double the sum, the share, and the number of people choosing to punish were announced after each round and were written on a white board for all to see. After each round, any written messages were read aloud to the group. At the end of 10 rounds, each participant wrote advice for the next generation, and this advice was read aloud to the next generation before the first round. The first generation received no advice or, in the replications, received advice from the free punishment condition that strongly advised punishing. Five series were conducted: one series of free punishment (14 generations); one series of 1-cent punishment (14 generations); one series of 5-cent punishment (14 generations); one replication series of 1-cent punishment (16 generations); and one replication series of 5-cent punishment (13 generations). Fig. 1
RESULTS Variation Figure 2 shows how contributions and frequency of choosing punishment varied across generations in the 5 series. When punishment was free, contributions and punishment were high and frequent, occurring at a relatively invariant rate. When punishment cost 5 cents, contributions tended to be lower, and punishment occurred at a low rate. When punishment cost 1 cent, punishment varied considerably. In both series, punishment re-evolved from zero. Accordingly, we focus on 1-cent punishment. Transmission Figure 3 shows the effect of advice in one generation on the frequency of punishment in Round 1 of the next generation. The more advice to “punish” the more punishment was chosen, and the more advice not to punish (“nopunish”) the less punishment was chosen. Advice was moderately but clearly efficacious. Selection Figure 4 shows dynamics within generations, from one round (i) to the next (i+1). Top graph: Contributions affected frequency of punishment. Punishing was maximal when contributions were submaximal. When contributions were maximal, punishment fell off. When contributions were very low, less punishing occurred and more defection. The fitted curves are quadratic functions. Bottom graph: We found no evidence that punishment affected contributions when examined in the overall data. In some generations, but by no means all generations, punishment strongly drove contributions, as shown in Figure 5. More analysis is required.
Fig. 4
Fig. 5
Fig. 2
Fig. 3
DISCUSSION AND CONCLUSION Altruistic punishment evolved as a group practice when the cost was non-zero but not too high (1 cent). Five participants had to punish to completely remove the advantage of contributing zero. Groups met the required 5 people about a third of rounds, and sometimes even averaged across rounds (Figure 2). Advice from one generation affected the frequency of punishment in the next generation. The more advice to punish, the more choices to punish, and the more advice not to punish, the fewer choices to punish (Figure 3). Thus, transmission occurred. If punishing was maintained by its effect on contributions, then we should observe that effect. Evidence for such selection was weak (Figure 4), at least in a fine-grained analysis from round to round, probably because: (a) advice was efficacious; and (b) most participants tended to choose punishment intermittently. In some groups, evidence for selection was strong (Figure 5). More analysis is required on individual participants’ strategies and messages. REFERENCES Baum, W. M., Paciotti, B., Richerson, P., Lubell, M., & McElreath, R. (2012). Cooperation due to cultural norms, not individual reputation. Behavioural Processes, 91, 90-93. Baum, W. M., Richerson, P. J., Efferson, C. M., Paciotti, B. M. (2004). Cultural evolution in laboratory microsocieties including traditions of rule giving and rule following. Evolution and Human Behavior, 25, 305-326. Chaudhuri, A., Graziano, S., & Maitra, P. (2006). Social learning and norms in a public goods experiment with intergenerational advice. Review of Economic Studies, 73, 357380.
RESULTS Variation Figure 2 shows how contributions and frequency of choosing punishment varied across generations in the 5 series. When punishment was free, contributions and punishment were high and frequent, occurring at a relatively invariant rate. When punishment cost 5 cents, contributions tended to be lower, and punishment occurred at a low rate. When punishment cost 1 cent, punishment varied considerably. In both series, punishment re-evolved from zero. Accordingly, we focus on 1-cent punishment. Transmission Figure 3 shows the effect of advice in one generation on the frequency of punishment in Round 1 of the next generation. The more advice to “punish” the more punishment was chosen, and the more advice not to punish (“nopunish”) the less punishment was chosen. Advice was moderately but clearly efficacious. Selection Figure 4 shows dynamics within generations, from one round (i) to the next (i+1). Top graph: Contributions affected frequency of punishment. Punishing was maximal when contributions were submaximal. When contributions were maximal, punishment fell off. When contributions were very low, less punishing occurred and more defection. The fitted curves are quadratic functions. Bottom graph: We found no evidence that punishment affected contributions when examined in the overall data. In some generations, but by no means all generations, punishment strongly drove contributions, as shown in Figure 5. More analysis is required.
Fig. 4
Fig. 5
Fig. 2
Fig. 3
DISCUSSION AND CONCLUSION Altruistic punishment evolved as a group practice when the cost was non-zero but not too high (1 cent). Five participants had to punish to completely remove the advantage of contributing zero. Groups met the required 5 people about a third of rounds, and sometimes even averaged across rounds (Figure 2). Advice from one generation affected the frequency of punishment in the next generation. The more advice to punish, the more choices to punish, and the more advice not to punish, the fewer choices to punish (Figure 3). Thus, transmission occurred. If punishing was maintained by its effect on contributions, then we should observe that effect. Evidence for such selection was weak (Figure 4), at least in a fine-grained analysis from round to round, probably because: (a) advice was efficacious; and (b) most participants tended to choose punishment intermittently. In some groups, evidence for selection was strong (Figure 5). More analysis is required on individual participants’ strategies and messages. REFERENCES Baum, W. M., Paciotti, B., Richerson, P., Lubell, M., & McElreath, R. (2012). Cooperation due to cultural norms, not individual reputation. Behavioural Processes, 91, 90-93. Baum, W. M., Richerson, P. J., Efferson, C. M., Paciotti, B. M. (2004). Cultural evolution in laboratory microsocieties including traditions of rule giving and rule following. Evolution and Human Behavior, 25, 305-326. Chaudhuri, A., Graziano, S., & Maitra, P. (2006). Social learning and norms in a public goods experiment with intergenerational advice. Review of Economic Studies, 73, 357380.
