Social dilemmas are often shaped by actions involving uncertain returns only achievable in the future, such as climate action or voluntary vaccination. In this context, uncertainty may produce non-trivial effects. Here, we assess experimentally — through a collective risk dilemma — the effect of timing uncertainty, i.e. how uncertainty about when a target needs to be reached affects the participants’ behaviours. We show that timing uncertainty prompts not only early generosity but also polarised outcomes, where participants’ total contributions are distributed unevenly. Furthermore, analysing participants’ behaviour under timing uncertainty reveals an increase in reciprocal strategies. A data-driven game-theoretical model captures the self-organizing dynamics underpinning these behavioural patterns. Timing uncertainty thus casts a shadow on the future that leads participants to respond early, while reciprocal strategies appear to be important for group success. Yet, the same uncertainty also leads to inequity and polarisation, necessitating the inclusion of new incentives handling these societal issues. 

The data was collected through behavioral economic experiments performed in a laboratory. Participants remain anonymous throughout the experiment and interact with other through a tablet computer. The experiment was divided into a control and 2 treatments. The control (treatment 1 - NU) uses a similar setting to Milinski et al. 2008, where participants play a collective risk dilemma (CRD) with 90% of risk and 10 rounds. Participants interact in groups of 6 subjects to which they are randomly assigned before the experiment starts. Jointly they are required to contribute to a public account from their endowment (40 EMUs) to achieve a total of 120 EMUs by the end of the experiment. If they achieve or surpass this value, participants keep in their account whatever was not contributed, otherwise their final payoff will be 0 EMUs with 90% of probability.

The 2 treatments test two levels of timing uncertainty. In treatment 2 (LU) the setting remains the same, but the total number of rounds is uncertain. Participants know that the game has at least 8 rounds, on average 10 and that after round 7 there is 1/3 of chances that the experiment will end in the next round. In this way, at the end of round 8 (and all other rounds that may follow) a virtual dice with 6 faces is thrown and if the results is either 1 or 2 the experiment ends.

Treatment 3 (HU) has the same design, with a change in the variation of the distribution that defines the duration of the experiment. In this treatment the experiment takes at least 6 rounds and after round 5 there is 1/5 chances that the experiment will end in the next round. In this case, at the end of round 6 (and all the other rounds that may follow) a virtual dice with 10 faces is thrown and, if the result is 1 or 2 the experiment ends.

Please refer to the file "Readme.md" for information on the meaning of each feature (column) in the dataset.