NOTE: End date changed to 9/30/2021 per NSSC (Ed., 4/1/21)

NOTE: End date changed to 3/31/2021 per NSSC (Ed., 5/21/2020)

NOTE: End date changed to 6/30/2020 per NSSC (Ed., 10/10/19)

NOTE: End date shows 6/30/2019 in NSSC (Ed., 4/2/19)

NOTE: End date changed to 5/17/2019 per D. Arias/HRP (Ed., 3/22/18)

NOTE: Element change to Human Factors & Behavioral Performance; previously Behavioral Health & Performance (Ed., 1/17/17)

Executive Summary of Phase 2: Through our Phase 2 efforts, we make one contribution to team risk in long-distance space exploration missions. We finalized the development of the computational model and published the report detailing the methodology and results of computational modeling (Deliverable 2).

Executive Summary of Phase 3: Through our Phase 3 efforts, we make seven contributions to team risk in long-distance space exploration missions. First, we used the CREWS ABM developed and demonstrated how empirically driven ABMs enable the recommendation of countermeasures that can support teams, in this case, to promote positive and dissuade negative social relationships within teams. We considered a “re-pairing” countermeasure, in which we reassigned which pairs of crew members are assigned to collaborate based on simulation results from the ABM. We tested the effectiveness of these ABM-recommended countermeasures among four member crews completing 45-day missions in the HERA LDSE analog (Deliverable 3). Our second contribution was to demonstrate the validation, reliability, and impact of the predictive team model as well as practical and research recommendations (Deliverable 4). The third contribution is to examine the crew performance over time (Deliverable 4). We found that behavioral team performance (cognitive conflict and psychomotor tasks) increases over time, whereas conceptual team performance (creativity and intellective tasks) declines. The fourth and fifth contributions are to examine which crew network patterns predict team performance (Deliverable 5 & Deliverable 6). One of the findings was that densely configured positive working networks (i.e., “Who do you enjoy working with?”) are associated with high performance on tasks, such as executing tasks (i.e., simulation tasks in which two crew members use a joystick to fly a transit vehicle to collection sites, while the other two crew members use virtual reality goggles to complete Extra Vehicular Activity exploring an asteroid’s surface). The sixth contribution is to examine the effect of team size on the performance, motivation, and effort of individual teammates (Deliverable 7). We found that there are many considerations for crew size, but solely from a team perspective, there may be some advantages to smaller teams (e.g., more influenced by teammates' ideas, greater contributions to performance, less social loafing, greater viability). The seventh contribution is that we developed a prototype dashboard for the detection of crew social relations, available here: http://nasa.soc.northwestern.edu/ (Deliverable 8).

NOTE: End date changed to 9/30/2021 per NSSC (Ed., 4/1/21)

NOTE: End date changed to 3/31/2021 per NSSC (Ed., 5/21/2020)

NOTE: End date changed to 6/30/2020 per NSSC (Ed., 10/10/19)

NOTE: End date shows 6/30/2019 in NSSC (Ed., 4/2/19)

NOTE: End date changed to 5/17/2019 per D. Arias/HRP (Ed., 3/22/18)

NOTE: Element change to Human Factors & Behavioral Performance; previously Behavioral Health & Performance (Ed., 1/17/17)

In the past year, we continued to make progress on each of our three research aims (as identified in previous task book and annual reports): (1) to develop an agent-based model of team composition for LDSE based on empirical data linking key model inputs (e.g., individual difference variables, network relational factors, task characteristics) to team functioning (e.g., social integration, team processes, team cohesion, team conflict) in LDSE-relevant contexts; (2) to conduct virtual experiments using characteristics and relationships identified in Phase I to identify the team functioning patterns that arise under different member compositions, and create a predictive model of team composition; and (3) to conduct an initial validation of the model we developed in LDSE analogs.

In addition to the above tasks, we have also worked to incorporate additional data into our models and analyses, in order to establish credibility and robustness of our results. This has included both the incorporation of new data from HERA (Campaign 6), as well as adjusting models originally developed using HERA to also incorporate data from NEK SIRIUS (Nezemnyy Eksperimental’nyy Kompleks / Scientific International Research In a Unique terrestrial Station) analog missions. These steps help ensure that our model is appropriate given the availability of new data.

Progress on Research Aim #1: Our first research aim is to develop an agent-based model of team composition for LDSE based on empirical data linking key model inputs (e.g., individual difference variables, network relational factors, task characteristics) to team functioning (e.g., social integration, team processes, team cohesion, team conflict) in LDSE-relevant contexts. Over the reporting period, we continued to make incremental developments and refinements on the agent-based model (ABM) we developed of social relationships within crews. We split our model into two parts: one part focusing on positive and negative relationships (task affect and hindrance) between crew members, and the other part focusing on claiming and granting leadership between crew members. Within each of the two parts, we implemented code to model the interaction between different types of ties: interactions between positive and negative relationships, and interactions between relationships claiming leadership and relationships granting leadership.

Progress on Research Aim #2: Our second research aim is to conduct virtual experiments using characteristics and relationships identified in Phase I to identify the team functioning patterns that arise under different member compositions, and create a predictive model of team composition. During the past reporting period, using the new model fit, we began conducting these virtual experiments using computational simulation. In these virtual experiments, we examine how different hypothetical team compositions (team member’s demographics, relationships, task schedules) would perform during LDSE according to our model. Using optimization methods, we have developed a way to select high-performance teams based out of a pool of different potential crew members.

Progress on Research Aim #3: Our third research aim is to conduct an initial validation of the model developed in LDSE analogs. This year, we analyzed results of the validation test of the model conducted in Campaign 5, and began compiling the results for presentation and publication. Our ABM was used to preemptively recommend “best” and “worst” pairings of crew members for the Rover and Phobos tasks in mission. Using surveys gathered over the course of each mission, we were able to evaluate the outcomes of using both “best” and “worst” pairings (the change in social relationships, their satisfaction with each pairing, viability). By performing new statistical analyses, we were able to demonstrate the viability of using our ABM to identify best pairings in advance of a future mission. These results will be presented at the upcoming International Astronautical Congress in October 2021.

NOTE: End date changed to 3/31/2021 per NSSC (Ed., 5/21/2020)

NOTE: End date changed to 6/30/2020 per NSSC (Ed., 10/10/19)

NOTE: End date shows 6/30/2019 in NSSC (Ed., 4/2/19)

NOTE: End date changed to 5/17/2019 per D. Arias/HRP (Ed., 3/22/18)

NOTE: Element change to Human Factors & Behavioral Performance; previously Behavioral Health & Performance (Ed., 1/17/17)

In year 4, we have made significant progress on our three research aims: (1) develop an agent-based model of team composition for LDSE based on empirical data linking key model inputs (e.g., individual difference variables, network relational factors, task characteristics) to team functioning (e.g., social integration, team processes, team cohesion, team conflict) in LDSE-relevant contexts; (2) conduct virtual experiments using characteristics and relationships identified in Phase I to identify the team functioning patterns that arise under different member compositions, and create a predictive model of team composition; and (3) conduct an initial validation of the model developed in Phase 2 in the Human Exploration Research Analog (HERA) and NASA Extreme Environment Mission Operations (NEEMO) using specific manipulations of key factors (e.g., compositions; situational characteristics).

Research Aim #1: Our first research aim is to develop an agent-based model of team composition for LDSE based on empirical data linking key model inputs (e.g.,individual difference variables, network relational factors, task characteristics) to team functioning (e.g., social integration, team processes, team cohesion, team conflict) in LDSE-relevant contexts. During year four, we developed an approach to train and validate our model off of empirical data using genetic search algorithms and approximate bayesian computation. This allowed us to test our models based on the observational data collected from the HERA crews, as well as identify new parameters and effects that should be added into our model in order to improve performance. By looking. at the out of sample performance of our model, we were able to confirm which components did and did not result in successful predictions.

Research Aim #2: Our second research aim is to conduct virtual experiments using characteristics and relationships identified in Phase I to identify the team functioning patterns that arise under different member compositions and create a predictive model of team composition. We developed a framework for how to conduct these experiments going forward. In order to design meaningful experiments, we identified a way to explore the effects of possible interventions that would improve crew functioning and performance. We applied the model to examine “What If” scenarios about hypothetical new team compositions, to extrapolate about what would happen in such crews.

Research Aim #3: Our third research aim is to conduct an initial validation of the model developed in Phase 2 in the Human Exploration Research Analog (HERA). To support this aim, we have been collecting data on trait constructs, sociometric constructs, and team dynamics measures, with many of these constructs and measures being collected over time. In Campaign 5, we began to use the model to implement countermeasures -- the pairing of crew members on different tasks (Rover and Phobos) -- our ABM was used to recommend pairing assignments for these tasks in advance of the start of each mission. We identified a “good” pairing and a “bad” pairing based on our predictions, and alternated between using these pairings on the task in-mission. Using new data collected through these missions, we were able to demonstrate that the “good” pairings identified based on our model helped to improve team functioning.

NOTE: End date changed to 5/17/2019 per D. Arias/HRP (Ed., 3/22/18)

NOTE: Element change to Human Factors & Behavioral Performance; previously Behavioral Health & Performance (Ed., 1/17/17)

This past year, we have made significant progress on our three research aims: (1) develop an agent-based model of team composition for LDSE based on empirical data linking key model inputs (e.g., individual difference variables, network relational factors, task characteristics) to team functioning (e.g., social integration, team processes, team cohesion, team conflict) in LDSE-relevant contexts; (2) conduct virtual experiments using characteristics and relationships identified in Phase I to identify the team functioning patterns that arise under different member compositions, and create a predictive model of team composition; and (3) conduct an initial validation of the model developed in Phase 2 in the Human Exploration Research Analog (HERA) and NASA Extreme Environment Mission Operations (NEEMO) using specific manipulations of key factors (e.g., compositions; situational characteristics).

Research Aim #1: Our first research aim is to develop an agent-based model of team composition for LDSE based on empirical data linking key model inputs (e.g., individual difference variables, network relational factors, task characteristics) to team functioning (e.g., social integration, team processes, team cohesion, team conflict) in LDSE-relevant contexts. This year we developed a functional version of the model based on theories of team composition, team functioning, and social networks. We used data from Campaign 3 to validate the model. We also developed a prototype of a decision aid dashboard (TEAMSTAR) that runs the ABM (agent-based model) in the background and could be used by NASA personnel to make informed decisions about crew composition.

Research Aim #2: Our second research aim is to conduct virtual experiments using characteristics and relationships identified in Phase I to identify the team functioning patterns that arise under different member compositions, and create a predictive model of team composition. This year we began testing the model for its fit with the positive-affect, negative-affect, information-sharing, and coordination networks.

Research Aim #3: Our third research aim is to conduct an initial validation of the model developed in Phase 2 in the Human Exploration Research Analog (HERA) and NASA Extreme Environment Mission Operations (NEEMO). To support this aim, we have been collecting data on trait constructs, sociometric constructs, and team dynamics measures, with many of these constructs and measures being collected over time. As part of this effort, we developed and administered a team task batter to track team performance on multiple dimensions over time.

NOTE: Element change to Human Factors & Behavioral Performance; previously Behavioral Health & Performance (Ed., 1/17/17)

In year 2 of a 3 year programmatic effort, we have made significant progress on our three research aims: (1) develop an agent-based model of team composition for LDSE based on empirical data linking key model inputs (e.g., individual difference variables, network relational factors, task characteristics) to team functioning (e.g., social integration, team processes, team cohesion, team conflict) in LDSE-relevant contexts; (2) conduct virtual experiments using characteristics and relationships identified in Phase I to identify the team functioning patterns that arise under different member compositions, and create a predictive model of team composition; and (3) conduct an initial validation of the model developed in Phase 2 in the Human Exploration Research Analog (HERA) and NASA Extreme Environment Mission Operations (NEEMO) using specific manipulations of key factors (e.g., compositions; situational characteristics).

Research Aim #1: Our first research aim is to develop an agent-based model of team composition for LDSE based on empirical data linking key model inputs (e.g., individual difference variables, network relational factors, task characteristics) to team functioning (e.g., social integration, team processes, team cohesion, team conflict) in LDSE-relevant contexts. This year we developed a first functional version of the model based on theories of team composition, team functioning, and social networks.

Research Aim #2: Our second research aim is to conduct virtual experiments using characteristics and relationships identified in Phase I to identify the team functioning patterns that arise under different member compositions, and create a predictive model of team composition. This year we began testing the model for its fit with the positive-affect, negative-affect, information-sharing, and coordination networks.

Research Aim #3: Our third research aim is to conduct an initial validation of the model developed in Phase 2 in the Human Exploration Research Analog (HERA) and NASA Extreme Environment Mission Operations (NEEMO). To support this aim, we have been collecting data on trait constructs, sociometric constructs, and team dynamics measures, with many of these constructs and measures being collected over time. For the HERA mission, there were three main data collection phases: before simulation, during simulation, and at the end of simulation.

In order to effectively address our research questions and ultimately build a predictive model of team composition, we first developed a refined conceptual model. This model incorporates the dynamic nature of LDSE, accounts for how individual differences may affect team functioning over an extended length, and allows for the important role of relational ties between team members (e.g., failures in social integration such as subgrouping and isolation) in mission success.

Our conceptual model focuses on the recursive developmental trajectory of a teaming episode and the emergence of teaming capital. We define a teaming episode as a period of interaction between two or more members in which there is collaboration towards a common goal. Examples of teaming episodes include working to repair a broken robot, preparing for an EVA (extravehicular activity), and even social activities, such as gathering around the table for lunch. We are specifically concerned with how relational ties develop overtime in these teaming episodes; examples include positive affective ties, negative affective ties, informational ties, and behavioral ties. The positive and negative affective ties will help to determine to what extent a team is socially integrated or if there are subgrouping or isolation risks for a particular composition. Informational and behavioral ties will help to determine if teams are establishing effective processes to be successful in LDSE missions. Together the team member attributes and relations form the crew’s teaming capital. Teaming capital is the team member knowledge, skills, abilities, and other characteristics (KSAOs) and relational (e.g., positive and negative) ties between team members from which team members can draw for effective functioning and performance.

The first six months of our grant included extensive preparation for data collection in the 2016 HERA campaign (Campaign 3). We completed data collection on mission 1 (C3M1) and are currently collecting data on the second mission of the 2016 HERA campaign (C3M2). During the HERA campaigns we administered two unique and proprietary measures, in the form of games in order to test team performance. The first game, Project Red Design, asked HERA participants to collaborate with students at Georgia Tech to build a well on Mars. The second game, Project Red Relay, had the same group of participants attempt to route messages to each other in the most efficient way possible. Relationships established in Project Red Design would help complete the tasks in Project Red Relay.

In addition to collecting data in HERA, we are conducting a meta-analysis of the extant team composition literature coded for fidelity and contextual moderators related to LDSE. We are using this meta-analysis to help identify key composition variables and configuration of variables that affect team processes, emergent states, and performance. Along with HERA data, the meta-analytic effects will help to guide the rules of the ABM. During the first year, we made substantial progress on the meta-analysis and have completed extensive literature searches, coded 12,648 dependent effect sizes from 517 sources, and prepared the data sets for analysis.

Our initial ABM was focused on how crew members (e.g., agents) operate within the teaming episodes model, specifically the extent to which individual difference predict the relational ties that develop between team members. The relational ties form patterns between crew members that, in conjunction with the individual differences, contribute to the effectiveness of teaming episodes. A LDSE mission is depicted as a series of teaming episodes. A mission can be “played out”; success across teaming episodes defines effective team composition.

We continued to develop this ABM to focus on the effective team composition of HERA participants, using information from the HERA playbook to indicate the tasks the crew is engaged in, as well as the directed and valued links between them. This model will continue to be developed as we include HERA data and effects from the meta-analysis. We will also add interactive relations, both between crew member (e.g., agent; referred to as turtles in the ABM program used) attributes and between crew member attributes and contextual features (e.g., task switching).

Data collection for HERA 2016 and the analyses for the meta-analysis are still underway, but we did uncover some preliminary results by conducting simulations with the ABM. Moving forward, we will conduct analyses on HERA, complete our analyses for the meta-analysis, expand the empirical basis of the ABM (e.g., text mine oral histories), and refine our ABM. Our more refined agent-based model will include how team member attributes interact with one another as well as how team member attributes interact with the context.