README ------ These files represent data analyzed to produce the statistics and figures for *Resilient cooperators stabilize long-run cooperation in the finitely repeated Prisoner’s Dilemma*. ------ **actions.csv** represents all ~375,000 player decisions collected over the course of the experiment. It contains the following fields: - **session**: participants were divided into two experiment sessions which were conducted at either 1pm ET or 3pm ET - **dropout**: whether this player eventually dropped out of the experiment prior to completion - **day**: the day of the experiment, from 1 to 20; not all players participated on all days - **game**: the game count within each day, from 1 to 20; some players may have missed games within a day - **round**: the round number of the game, from 1 to 10; an incomplete game will contain fewer than 10 rounds of data - **gameId**: a unique random string for each 2-player game - **player**: a unique random string representing each participant - **action**: 1 for cooperation, 0 for defection - **opp_action**: the opponent's decision during the same round, included for ease of analysis - **payoff**: the payoff this participant received during this round - **time**: the time elapsed since the start of the round - **total_time**: the time elapsed since the start of the game Note that the time fields may contain measurement error due to latency between the participants' browsers and the server. Decisions on the first round may also be slower than in other rounds, as players are idle for a short amount of time before random rematching. ------ **surveys_anon.csv** is the result of a debriefing and survey conducted for all players when they either dropped out of the experiment or completed it. It is particularly useful for understanding qualitatively how participants thought of the game. In this file, **player** is a randomized ID that is the same as in the actions file above. For exact wording of the survey questions, please consult page 20 of the [Supplementary Information](https://www.nature.com/article-assets/npg/ncomms/2017/170113/ncomms13800/extref/ncomms13800-s1.pdf) of our paper, or the experiment [source code](https://github.com/TurkServer/long-run-cooperation/blob/master/client/views/survey.html). Note that we have not included players' ZIP (postal) codes as this has been demonstrated to be sufficient to deanonymize MTurk workers along with their age and gender. For players who dropped out of the experiment, we also asked them why they stopped playing, with their response in the field **stopped**. As we referred to cooperate and defect neutrally as *Choice 1* and *Choice 2*, respectively, you will see this terminology used by participants as well. ------ Also included are **.ipynb** files (IPython/Jupyter notebooks) representing the code that was used to analyze the data above and generate the figures in the paper. These are using Python 2.7 and will need to be run in an appropriate environment (we suggest Anaconda/Miniconda). Note that these notebooks also contain other code for analyses that were not included in the paper, which may or may not be useful to you. If you require additional data from the experiment that is not represented here, please contact Andrew Mao.