Artificial Social Intelligence for Successful Teams (ASIST) Study 3

The ASIST Study-3 dataset was developed in a human subjects research study designed to assess the capability of artificial intelligence to instantiate a Machine Theory of Teams, and apply it to generate and issue (or withold) advice to team members that improve team state (e.g., motivation), process (e.g., synchronization), and mission effects (e.g., game score). These agents -- called Artificial Social Intelligence -- draw measurements of team state and process from agents called Analytic Components. These take their input from survey responses and behaviors of a three-person team executing an urban search and rescue task in Minecraft. This material is based upon work supported by the Defense Advanced Research Projects Agency (DARPA) under Contract No. HR001119C0130. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Defense Advanced Research Projects Agency. We have partitioned the full dataset into folders that support research in specific areas. A readme file in each folder (e.g., readme_audio.txt) describes the folder's contents in detail. (1) Data in the studywide folder will be of interest to researchers who conduct any analysis with any data from ASIST Study-2, because these files contain data that describe the study overall, the data used to evaluate AI, or the coding of data. (2) Data in the surveys folder will be of interest to researchers who study individual differences and their effects on behavior. (3) Data in the testbedmessages folder will be of interest to researchers who study individual and team behavior or who use any other components of this dataset, because these are machine- and human-readable text (json) records of the state and behaviors of study participants, and of the state of the task environment. (4) Data in the transcriptions folder will be of interest to researchers who study language use. The audio source of these imperfect machine transcriptions can be found in study video files and audio files. (5) Data in the audio folder will be of interest to researchers who study language use, or who wish to validate, contextualize, or specify transcriptions, testbed messages, and certain survey data. (6) Data in the video folder will be of interest to researchers who study machine vision, or who wish to validate, contextualize, or specify transcriptions, testbed messages, and certain survey data. (7) Data in the analysis folder will be of interest to those seeking examples of analyses developed by ASIST program performers, either to understand the data better, to identify opportunities for further analysis, or to build on analysis code. (8) Data in the methods folder (and in the studywide folder) will be useful to those seeking to reproduce the human subjects experiment.