Artificial Social Intelligence for Successful Teams (ASIST) Study 1 Falcon Testbed Dataset

<p> Artificial Social Intelligence for Successful Teams (ASIST) has a total of 4 studies, with a program overview on this website:<a href="https://artificialsocialintelligence.org/">https://artificialsocialintelligence.org/</a>
ASIST consists of 13 institutions and more than 300 individual contributors. Arizona State University (ASU) and Aptima as TA3 jointly developed the testbed, integrated requests from the performer teams, and collected the human subject dataset.</p>

<p>The ASIST Study 1 Falcon Testbed Dataset was developed as part of 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 withhold) advice to team members that improves team states (e.g., motivation), processes (e.g., synchronization), and mission outcomes (e.g., game score). These agents—called Artificial Social Intelligence Advisors (ASI Advisors)—draw measurements of team states and processes from components known as Analytic Components (ACs). In this initial exploratory stage, Study 1 involved one participant per trial, with ASI Advisors passively monitoring the player's actions, inferring their training condition, and predicting their next actions.</p>

<p>This material is based upon work supported by the Defense Advanced Research Projects Agency (DARPA) under Contract No. HR001119C0130. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the DARPA.</p>

<p>The Study 1 consent form included an agreement to keep the data confidential until the program ended in 2023, whereas the consent forms for other studies did not. As a result, Study 1 data were published later than the data from Studies 2, 3, and 4.</p>

<p><strong>Data Overview</strong></p>

<p>The dataset was collected between 2020-08-25 and 2020-10-20. It consists of data from 78 participants, each associated with a set of 42 files, including survey data, competency testbed logs (also known as trial messages), screen recordings, audio files, audio transcripts, observation annotations, and processed eye-tracking data. All files in the 'studywise' folder pertain to experimental materials and study design, while all files in the 'trials' folder contain the trial data for each participant.</p>

<p>A readme file (README_ASIST Study 1 Falcon Dataset.txt) describes the dataset contents in detail. </p>

<p>People involved with sample collection, processing, analysis, and/or submission in 2020: Nancy J. Cooke (ASU), Verica Buchanan (ASU), Chris Corral (ASU); Anagha Mudigonda (ASU), Myke Cohen (ASU), Aaron Teo (ASU), Lovein Thomas (ASU), Jeska Clark (ASU), Xiaoyun Yin (ASU), Federico Scholcover (ASU), Lixiao Huang (ASU), Jared Freeman (Chief Scientist at Aptima), JCR Colonna-Romano (Software Architect at Aptima), Matt Wood (Senior Scientist at Aptima). [If anyone is missing, please report to Lixiao Huang]</p>

<p> We would like to thank all the performer teams (TA1: CMU, DOLL, USC, SIFT, UAZ, CRA; TA2:UCF, IHMC, Cornell, CMU, Rutgers, Gallup) <p>

<p> We also would like to thank the following individuals for their valuable contributions to the annotation efforts in this dataset: Adarsh Pyarelal (UAZ), Anagha (ASU), Chinmai Basavaraj (UAZ), Fade Eadeh (CMU-TA2), Hayley Borck (SIFT), JiHyun Jeong (Cornell), Jessica Williams (UCF), Lixiao Huang (ASU), Loren Champlin (UAZ), Meghavarshini Krishnaswamy (UAZ), Paulo Soares (UAZ), Peter Lovett (UAZ), Rhyse Bendell (UCF), Salena Ashton (UAZ), Savannah Boyd (UAZ), Seongjin Park (UAZ), and Shreeya Jain (UAZ). <p>