2. Background

In Enhancing Human-Agent Teaming by ieeebrain2 Comments

Arwen H. DeCostanza*, Amar R. Marathe*, Addison Bohannon*, A. William Evans*, Edward T. Palazzolo**, Jason S. Metcalfe*, and Kaleb McDowell*
*Army Research Laboratory, **Army Research Office

As the vision for Internet of Things promises a world of interconnected devices that anticipate our needs, we can expect future military teams to be equipped with an array of intelligent and networked agents that anticipate the needs of the team and make decisions both independently and in coordination with their human teammates. This will differ fundamentally from the human-centered team concepts on which military doctrine and organizational psychology are built. At present, military teams train and operate with a host of advanced technology (e.g., night vision technology, smart weapon technology), but at their core, the technology endows human team members with greater individual capabilities and does not effectively change the dynamics of human teamwork. Merely increasing individual member capabilities may not enhance the short- and long-term emergent properties of teamwork that are critical to team performance. This issue will become increasingly critical for future teams that are expected to require close cooperation, coordination, and communication between dynamic assemblages of human and agent (e.g., robots, intelligent assistants, and intelligent sensors) teammates, particularly when the teams are not trained together. Further, both Soldiers and intelligent agents (both envisioned to have varied levels of training, experience, and operational capabilities) will likely be required to adapt to new styles of work and interactions over the courses of their deployment that induce novel challenges that the teams must overcome. In envisioning this future of human-agent teaming, we hold 3 critical assumptions.

  1. First, human capabilities and human team members within the future operating environment will continue to be necessary. The specific capabilities required of humans will change from what they are today, but technology will not replace humans completely.
  2. Second, the roles of humans and the nature of the interactions they have with autonomy will change. The concept of technology being a tool for humans will be superseded by technology as mentored actors in the environment, teammates with unique non-human skills, and technology that augments fundamental human capabilities.
  3. Third, the training required for humans and groups to be effective will change dramatically. Training will have to enable humans and teams to handle the dynamic and rapid evolution of technology, as well as the shift in critical analysis and action from humans to intelligent technology.

Advancements across many different domains are considered critical to this future human-agent teaming vision. Piekarski et al. (2016) document many of the challenges from an intelligent systems perspective, including: control of large, distributed autonomous teams with varying levels of autonomy and intelligence; combining autonomous agents, sensors, and tactical super-computing to establish distributed networked intelligent systems; and methods for heterogeneous teams to carry out tasks under dynamic and varying conditions. Similarly, Suri et al. (2016) describe research challenges related to fundamental understanding of how to learn and devise complex models of the Internet of Battlefield Things (IOBT), including goals, networks, information, and analytics that enable intelligent command and control, and battlefield services. In 2017, the US Army Research Laboratory began 2 large-scale collaborative research programs to address these challenges.

Neither of these current efforts focuses directly on optimizing solutions to enhance human strengths and mitigate human weaknesses, nor do they directly address development of the inherent teamwork (states and processes) between humans and autonomous agents that is extremely critical to this future vision. Here, we open discussion about one promising approach to addressing the larger problem of coordination and cooperation in human-agent teams—individualized and adaptive technologies that promote effective teamwork in teams of humans and intelligent agents. This report posits a number of critical capabilities necessary to realize the future human-agent teaming vision. Within this discussion, several reoccurring topics are critical to discuss upfront, including teamwork and individualized, adaptive technologies.

What do we mean by teamwork?
Many researchers have argued that team performance is a multilevel process that includes both individual taskwork performance processes and individual- and team-level teamwork processes (Kozlowski and Klein 2000; Salas et al. 2007). Marks et al. (2001) provided definitions of taskwork and teamwork to distinguish the 2, suggesting that “taskwork represents what it is that teams are doing, whereas teamwork describes how they are doing it with each other” (p. 357). When we refer to human-agent teaming and human-agent teamwork, we are focused on the team-level states and processes that influence performance and effectiveness (e.g., cohesion, shared mental models, shared situation awareness, coordination, and communication), rather than individual taskwork.

What do we mean by technology? Agent?
To this point, it may be unclear what we mean when we use technology, agent, intelligent agent, and teammate. We consider the team to be composed of human team members as well as distributed sensors, robots, UAVs, autonomous vehicles, intelligent assistants, and other advanced technologies that can perform taskwork as part of the larger team, while we reserve the term technology for those devices, software, protocols, and other interventions that target the members of the team with the goal of improving team processes. It is entirely possible that a technology will also be a team member, which we refer to as an agent. We use the term technology when referring to its role as assisting in team performance as opposed to satisfying its role in the team (i.e., completing its assigned taskwork).

Technologies of the future should be individualized and adaptive.
Military technologies and doctrine prioritize the interchangeability of operators. Although this leads to robust performance, assumptions about average operator capabilities limit system capabilities and also likely constrain high-performing individuals from using the full extent of their own and the system’s capabilities. Future technologies need to be adaptive and individualized, accounting for individual’s capabilities and limitations in real-time to achieve greater performance. In allowing individual agents to behave in manners that are consistent with their own strengths rather than imposing uniform behaviors, individual performance should be dramatically improved. Shifting to this paradigm also enables technological solutions to target particular individual capabilities and performance that improve team-level properties and performance.

Within this context, we focus here on science and technology to enable the interactions and interdependencies between heterogeneous members of human-agent teams, and specifically, on influencing individual team members with the goal of enhancing emergent team properties contributing to effective performance, and limiting emergent team properties contributing to less effective performance. The report posits 3 broad, intertwined areas that frame the technical and scientific challenges: 1) Individualized, Adaptive Technologies for Teamwork, which focuses on technologies that can adapt to individuals to optimize systems of interdependent agents for the purposes of enhancing overall team performance; 2) Adaptive Implementation, which focuses on the adaptive application of individualized technologies based on the dynamics of task and environmental context and team members (e.g., state, knowledge, skills, abilities); and 3) Training for Mutual Adaptation and Complex Teaming, which focuses both on how training must change for humans and agents to team in highly adaptive, highly intelligent technology contexts, as well as how individualized technologies can support on-the-job training.


  1. Marks’ distinction is a good one — but often gets lost — in all research publications, we can’t let others change our words (e.g., to avoid repetition).

    Actually, we need t-shirts with the taskwork/teamwork verbage. People may actually get that difference if we continue to promote it.

    1. Thanks for your reply – teamwork vs. taskwork is a continual discussion in our lab meetings and writing this paper with an interdisciplinary team was very thought-provoking for me. I’m excited to do more work in this area!

Leave a Comment