Fig. 1. Each phase of a typical public goods game. In phase 1, players each start with a personal pot of money. In phase 2, they decide how much of their own money to give to a group pot. In phase 3, the experimenter multiplies whatever was contributed; in this example, contributions are multiplied by 1.5. In phase 4, the multiplied money is divided evenly to every person. The tension of the game arises because the group as a whole does best if everyone contributes, yet individual people do best if they share in others’ contributions without contributing themselves.

Fig. 2. The disaster game. Everyone starts with personal money (top panel). They each decide how much to contribute to a group threshold. If the combined contributions are less than the threshold, everyone loses their remaining money (bottom left panel). If the combined contributions are more than or equal to the threshold, everyone keeps their remaining money. (Contributions are lost regardless.)

Fig. 3. The disaster game. Everyone starts with personal money (top panel). They each decide how much to contribute to a group threshold. If the combined contributions are less than the threshold, everyone loses their remaining money (bottom left panel). If the combined contributions are more than or equal to the threshold, everyone keeps their remaining money.

Fig. 4. The decision each player faced. A player starts with a personal pot of money. If they make a “certain contribution,” that money is given directly to the group’s threshold. If they make a “risky contribution,” there is a 50% chance the money will be doubled before going to the group’s threshold and a 50% chance the money will disappear and nothing will go to the group’s threshold. If they “defect,” then they keep their personal money, and nothing goes to the group threshold.

Fig. 5. The percentage of players making risky contributions in the lab with high stakes. Error bars are standard errors of the mean.

Fig. 6. The percentage of players making risky contributions in the lab with low stakes. Error bars are standard errors of the mean.

Fig. 7. The percentage of players making risky contributions online with high stakes. Error bars are standard errors of the mean.

Fig. 8. The percentage of players making risky contributions online with low stakes. Error bars are standard errors of the mean.

Fig. 9. The percentage of players making risky contributions online with high uncertainty in the risky contribution. Error bars are standard errors of the mean.

Fig. 10. The disaster game. Everyone starts with personal money (top panel). They each decide how much to contribute to a group threshold. If the combined contributions are less than the threshold, everyone loses their remaining money (bottom left panel). If the combined contributions are more than or equal to the threshold, everyone keeps their remaining money. Contributions disappear regardless of whether the threshold is met.

Fig. 11. Among a large set of groups, groups contribute to meet one another’s thresholds. However, help is not reciprocal. If Group A contributes for Group B, then Group B does not contribute for Group A. Group B contributes to help a random other group and Group A is helped by a different, random other group.

Fig. 12. The decision each player faced. A player starts with a personal pot of money. If they make a “certain contribution,” that money is given directly to the group’s threshold. If they make a “risky contribution,” there is a 50% chance the money will be doubled before going to the group’s threshold and a 50% chance the money will disappear and nothing will go to the group’s threshold. If they “defect,” then they keep their personal money, and nothing goes to the group threshold.

Fig. 13. The percentage making certain, risky, or no contributions at each threshold. Error bars are standard errors of the mean.

Fig. 14. The percentage making risky contributions at each threshold in each sample. Error bars are standard errors of the mean.

Fig. 15. The percentage of players in each sample making the most useful choice before and after the opportunity to buy information, at each threshold. Error bars are standard errors of the mean.

Fig. 16. The basics of the backfire game. Four citizens each independently decide how much to give to a threshold to prevent disaster. A policymaker decides whether to enact geoengineering. If the policymaker uses geoengineering, it could succeed and meet the group’s threshold. But if geoengineering fails, it also backfires. When it backfires, citizens lose money and continue to face the same challenge of preventing disaster.

Fig. 17. The proportion using geoengineering at each probability that geoengineering succeeds. Error bars are standard errors of the mean.

Fig. 18. The proportion who used geoengineering at each probability that geoengineering succeeded, in each condition. Error bars are standard errors of the mean.

Fig. 19. The proportion who used geoengineering at each probability that geoengineering succeeded. Results are broken down by whether players believed their group would succeed without geoengineering. Error bars are standard errors of the mean.

Fig. 20. Total carbon emissions and carbon emissions per capita for the top ten highest-emitting countries in 2020.