Participant Faces from a Repeated Prisoner's Dilemma

The authors (ES and TS) conducted a computerized laboratory experiment in an experimental economics laboratory using a modified Prisoner’s Dilemma known as Friend or Foe (e.g. see List, 2006. The Review of Economics and Statistics, 88(3), 463-471). 96 young adults aged 18 to 25 years old (51 men, 45 women) gave permission to be video recorded at intervals throughout the experimental procedure under standardized videographic conditions and for their recordings and experiment data to be made available for later research. All individuals video recorded were students at Chapman University in Orange, CA. We provide a dataset described below based on behavioral game and survey data collected from participants over a computer network, media from video recordings, and media analysis measures.

Description of Experimental Task: Participants were provided a printed copy of the instructions (Instructions.pdf) explaining the following experimental procedure. Participants were randomly paired with one other person, with whom they anonymously interacted over a computer network throughout the experimental session. Participants were told that at no time would their or other participants’ true identity or videorecorded images be revealed to participants in the experiment session. Participants were told to expect a minimum of two rounds of game interactions with possibility for more rounds. Each round required the participants to first privately state their (non-binding) intention for the camera, saying either  “I intend to Split” or “I intend to Take All” (which we videorecorded, see Media: Videos of Stated Intent), then to choose an interaction strategy (either “Split” or “Take All”). Each round we provided participants the results of their pair’s game interaction on their computer screens. We videorecorded participants discovery, viewing, and reaction to the interaction results (see Media: Videos of Viewed Results). After the last round of interaction, we collected response data using an exit survey.

Game and Survey Data: Data includes participants’ stated intent and game behavior for each round, emotional reactions (each on a scale of 1 to 5) self-reported after learning the results of each round (measured with a modified 20 item PANAS), and the following data collected during an exit survey: responses (each on a scale of 0 to 4) to six questions used for scoring the Revised Life Orientation Test (from Scheier et al., 1994. Journal of Personality and Social Psychology, 67, 1063-1078), age, gender, first language, and English language fluency.

Media: Videos of participants were taken using computer display mounted Logitech C920 1080P HD digital cameras, at computer terminals in individual cubicles. The cameras were aimed at participants who sat against a background of either a gray carpeted cubicle or a brown wall under standardized diffuse lighting conditions. For video recordings in this data set, participants were instructed to wear microphone-earphone headsets, to keep hair away from their face, and to sit upright and look at the computer screen they were facing. Camera-to-head distance was controlled by the cubicle space and chair position, and factory default camera settings were held constant.

Video of Stated Intent. 204 eight-second videos with audio were taken after an on-screen prompt instructed participants to make a spoken non-binding statement of intent. During this phase, we played grey noise over a public address system that masked ambient sound from beyond the individual cubicle within which the video recording equipment was situated. In combination with well insulated over-ear headphones worn by participants, the grey noise was helpful for preventing interference or leakage of other participants’ statements.

Video of Viewed Results. 204 eight-second videos with audio were taken that capture the moment of participants discovery, viewing, and reaction to the interaction results. The interaction results appear and remain visible to participants on their computer screens from around 2.5 seconds (no earlier than 2 seconds) into the 8 second videos.

Thin Slice Video. 204 videos (two to three seconds in length) without audio were trimmed from full Videos of Stated Intent. These thin slice videos capture the 2-3 second interval directly following the concluded stated intent.

Photograph. 204 photographs (640 x 480 pixel .jpg files, 24 Bit depth) including a participant’s full face were captured from video using the VLC media player (3.0.11) “snapshot” tool.

Media Analysis Measures:

Stated Intent. The particular statement of intent (either “I intend to Split” or “I intend to Take All”) unambiguously recognized by the authors upon inspection of the sound and appearance of each Video of Stated Intent’s recorded statement, is noted in the attached Media.csv file.

Skin Coloration in Photographs. Using Adobe Photoshop 2020, skin patches (51 × 51 pixels) were selected from the left and right cheeks of each photograph face image. These cheek areas were selected so that they included only skin surfaces but never the edge of a face, ears, lips, or eyelashes. The light (L*),red (a*), and yellow(b*) values that are relevant to human color perception were then measured by centering a "color sampler tool" in Photoshop that averaged over a 51x51 pixel area the CIELab color space values. L*,a*,b* values for right and left cheeks in Photographs are noted in the attached Media.csv file.

Facial Width to Height in Photographs: To measure facial width to height ratio (fWHR) in photographs, we used tools in Photoshop 2020: first we rotated the photos as necessary using image rotation so that pupils were levelled horizontally, we record the degree of positive or negative rotation used, then we used the frame tool to measure the upperface “height” distance between the upper edge of upper lip and lower edge of brows and the “width” distance between left and right zygion (byzygomatic width) of face images. These methods are consistent with Carre and McCormick (2008, Proceedings of the Royal Society B‐ Biological Sciences, 275, 2651–2656.) methods for measuring the height of upperface and byzygomatic width of face images. The facial width and height in Photographs are noted in the attached Media.csv file.

Data and files: We provide audio-video and image files of Media along with the open-access .csv files detailed below. The above data is organized by the following three data structures in .csv file format.

Fields.csv is an open-access file includes Game and Survey Data organized by participants and rounds in the experiment. Fields_codebook.csv is a codebook explaining the column headers and participants’ data in the Fields.csv file.

Media.csv is an open-access file that catalogues the names and types of Media files provided and includes Media Analysis Measures. Media_codebook.csv is a codebook explaining column headers and participants’ data in the Media.csv file.

MappingFile.csv is a restricted file linking the participant data in the Fields.csv and Media.csv files. Please contact the authors for this access to this key at http://doi.org/10.5281/zenodo.4321814 if you plan to work at the intersection of these data or replicate published work based on these data.

Instructions.pdf is the instructions document provided to all participants in the experiment from which the above data was collected.