The rise in artificial intelligence capabilities in autonomy-enabled systems and robotics has pushed research to address the unique nature of human-autonomy team collaboration. The goal of these advanced technologies is to enable rapid decision-making, enhance situation awareness, promote shared understanding, and improve team dynamics. Simultaneously, use of these technologies is expected to reduce risk to those who collaborate with these systems. Yet, for appropriate human-autonomy teaming to take place, especially as we move beyond dyadic partnerships, proper calibration of team trust is needed to effectively coordinate interactions during high-risk operations. But to meet this end, critical measures of team trust for this new dynamic of human-autonomy teams are needed. This article seeks to expand on trust measurement principles and the foundation of human-autonomy teaming to propose a “toolkit” of novel methods that support the development, maintenance, and calibration of trust in human-autonomy teams operating within uncertain, risky, and dynamic environments.

We examined the effects of communication style on human performance, workload, situation awareness, and trust in a robot in a human-robot team. In a 2 x 2 mixed factor study, participants were teamed with a simulated robot to conduct a cordon-and-search style task. Participants were assigned to a communication style (Directive vs. Non-directive; between subjects), and both groups experienced varied periods of task load (high vs. low task load; within-subjects). Results indicate that task load influenced the participants' task performance more than communication style. However, there were some differential effects on response time and workload due to communication style. Participants in the Non-directive group did not report a higher workload than those in the Directive condition, even though objective measures of workload (i.e., eye-tracking measures) indicated they experienced a higher workload. These results may be due to the presence of feedback inherent in the differing communication styles.

Cohesion is an important property of teams that can affect individual teammates and team outcomes. However, cohesion in teams that include autonomous systems as teammates is an underexplored topic. We examine the extant literature on cohesion in human teams, then build on that foundation to advance the understanding of cohesion in human-autonomy teams, both similarities and differences. We describe team cohesion, the various definitions, factors, dimensions and associated benefits and detriments. We discuss how that element may be affected when the team includes an autonomous teammate with each description. Finally, we identify specific factors of human-autonomy interaction that may be relevant to cohesion, then articulate future research questions critical to advancing science for effective human-autonomy teams. Relevance Statement:The human team literature has provided a foundation onto which human-autonomy team research can build, but the team dynamics, and subsequent states, established in multi-human teams are expected to differ in human-autonomy teams. This manuscript focuses on cohesion, one such state and synthesises elements of human team cohesion and human-autonomy interaction to detail expectations for cohesion in human-autonomy teams. These expectations can serve as a launch point for future research.

Each of twenty participants teamed with a learning capable agent to conduct a threat classification task. The agent’s reasoning and learning transparency varied across four scenarios. Access to agent reasoning transparency improved task performance as assessed by percent of correct classifications. Agent learning transparency of inferred knowledge improved task response time and reduced cognitive workload. However, when the human was burdened with directly teaching the agent, task completion time and perceived workload increased dramatically, while satisfaction in task performance decreased. These findings indicate that when teamed with learning capable agents, human performance and workload are best supported when the autonomy can derive its needed information with minimal human input.

The future of robotics, envisioned by the US military, is made up of humans teamed with autonomous, intelligent robots. A shift in human–robot interaction (HRI), from the teleoperation of robotic systems to a more teamwork-oriented interaction, concurrently changes the informational requirements of the actors in the interaction. Consequently, military research into robotics is exploring what information fulfills those changed informational requirements, how robots can convey that information to human teammates, and what information do robots need to acquire from those human teammates. In this chapter, we discuss how the conception of a military robot is anticipated to change, how that change influences the interaction between humans and robots, and some of the different lines of research being done to support future HRI.

Robots’ increased autonomous capabilities necessitate human-robot communication. Exploration of this communication, including the pattern of and the content of such, is relevant to the development of these robots and the understanding of how they can interact with human teammates. This study compares two different patterns of communication and two approaches to transparent interaction, looking at their effects on human team members’ attitudes towards robots with which they are communicating. Participants found robots using a bidirectional communication pattern to be more animate, likeable, and intelligent than robots using a unidirectional communication pattern.

Agent transparency is an important contributor to human performance, situation awareness (SA), and trust in human-agent teaming. However, agent transparency's effects on human performance when the agent is unreliable have yet to be examined. This paper examined how the transparency and reliability of an autonomous robotic squad member (ASM) affected a human observer's task performance, workload, SA, trust in the robot, and perceptions of the robot. In a 2 (ASM transparency) x 2 (ASM reliability) within-subject design experiment, participants monitored a simulated soldier squad that included an ASM as it traversed a simulated training environment, while concurrently monitoring the environment for targets. There was no difference in participants' performance on the target detection task, workload, or SA due to either ASM transparency or reliability. ASM reliability influenced participant trust and perceptions of the robot. Results suggest that reliability may be a stronger influence on the human's perceptions of the robot than transparency. Robot errors had a profound and lasting effect on the participants' perception of the robot's future reliability and resulted in reduced confidence in their assessments of the robot's reliability. These findings could have important implications for the continued use of automated systems when the user is aware of system errors.

Human-robot interaction requires communication, however what form this communication should take to facilitate effective team performance is still undetermined. One notion is that effective human-agent communications can be achieved by combining transparent information-sharing techniques with specific communication patterns. This study examines how transparency and a robot’s communication patterns interact to affect human performance in a human-robot teaming task. Participants’ performance in a target identification task was affected by the robot’s communication pattern. Participants missed identifying more targets when they worked with a bidirectionally communicating robot than when they were working with a unidirectionally communicating one. Furthermore, working with a bidirectionally communicating robot led to fewer correct identifications than working with a unidirectionally communicating robot, but only when the robot provided less transparency information. The implications these findings have for future robot interface designs are discussed.

In a human-automation interaction study, automation assistance level (AL) was investigated for its effects on operator performance in a dynamic, multi-tasking environment. Participants supervised a convoy of manned and unmanned vehicles traversing a simulated environment in three AL conditions, while maintaining situation awareness and identifying targets. Operators' situation awareness, target detection performance, workload and individual differences were evaluated. Results show increasing AL generally improved task performance and decreased perceived workload, however, differential effects due to operator spatial ability and perceived attentional control were found. Eye-tracking measures were useful in parsing out individual differences that subjective measures did not detect. At the highest AL, participants demonstrated potentially complacent behaviour, indicating task disengagement. Practitioner Summary: The effect of varying automation assistance level (AL) on operator performance on multiple tasks were examined in a within-subjects experiment. Findings indicated a moderate AL improved performance, while higher levels encouraged complacent behaviour. Effects due to individual differences suggest that effective AL depends on the underlying characteristics of the operator.

Effective collaboration between humans and agents depends on humans maintaining an appropriate understanding of and calibrated trust in the judgment of their agent counterparts. The Situation Awareness-based Agent Transparency (SAT) model was proposed to support human awareness in human-agent teams. As agents transition from tools to artificial teammates, an expansion of the model is necessary to support teamwork paradigms, which require bidirectional transparency. We propose that an updated model can better inform human-agent interaction in paradigms involving more advanced agent teammates. This paper describes the model's use in three programmes of research, which exemplify the utility of the model in different contexts - an autonomous squad member, a mediator between a human and multiple subordinate robots, and a plan recommendation agent. Through this review, we show that the SAT model continues to be an effective tool for facilitating shared understanding and proper calibration of trust in human-agent teams.

Effective collaboration between humans and agents depends on humans maintaining an appropriate understanding of and calibrated trust in the judgment of their agent counterparts. The Situation Awareness-based Agent Transparency (SAT) model was proposed to support human awareness in human–agent teams. As agents transition from tools to artificial teammates, an expansion of the model is necessary to support teamwork paradigms, which require bidirectional transparency. We propose that an updated model can better inform human–agent interaction in paradigms involving more advanced agent teammates. This paper describes the model's use in three programmes of research, which exemplify the utility of the model in different contexts – an autonomous squad member, a mediator between a human and multiple subordinate robots, and a plan recommendation agent. Through this review, we show that the SAT model continues to be an effective tool for facilitating shared understanding and proper calibration of trust in human–agent teams.

A goal for future robotic technologies is to advance autonomy capabilities for independent and collaborative decision-making with human team members during complex operations. However, if human behavior does not match the robots’ models or expectations, there can be a degradation in trust that can impede team performance and may only be mitigated through explicit communication. Therefore, the effectiveness of the team is contingent on the accuracy of the models of human behavior that can be informed by transparent bidirectional communication which are needed to develop common ground and a shared understanding. For this work, we are specifically characterizing human decision-making, especially in terms of the variability of decision-making, with the eventual goal of incorporating this model within a bidirectional communication system. Thirty participants completed an online game where they controlled a human avatar through a 14 × 14 grid room in order to move boxes to their target locations. Each level of the game increased in environmental complexity through the number of boxes. Two trials were completed to compare path planning for the condition of known versus unknown information. Path analysis techniques were used to quantify human decision-making as well as provide implications for bidirectional communication.

We developed the Situation awareness-based Agent Transparency (SAT) model to support human operators’ situation awareness of the mission environment through teaming with intelligent agents. The model includes the agent's current actions and plans (Level 1), its reasoning process (Level 2), and its projection of future outcomes (Level 3). Human-inthe-loop simulation experiments have been conducted (Autonomous Squad Member and IMPACT) to illustrate the utility of the model for human-autonomy team interface designs. Across studies, the results consistently showed that human operators’ task performance improved as the agents became more transparent. They also perceived transparent agents as more trustworthy.

We examined how varying the transparency of agent reasoning affected complacent behavior, in the form of incorrect acceptances of an agent’s recommendations, in a route selection task. We were particularly interested in how participants’ eye movements might disambiguate whether the incorrect acceptances were due to complacency or incorrect information processing. Participants guided a threevehicle convoy safely through a simulated environment of which they had a limited amount of information, while maintaining communication with command and monitoring their surroundings for threats. The intelligent route-planning agent assessed potential threats and suggested changes to the convoy route as needed. Each participant was assigned to one of three agent reasoning transparency conditions. While access to agent reasoning did appear to reduce complacent behavior in one condition, performance in the other conditions indicated potential complacent behavior. An area of interest analysis, reviewed in conjunction with the performance data, indicated the reason behind the participants’ behavior was different between these two conditions. While in the non-transparent condition participants were likely engaging in complacent behavior, in the highly transparent condition it is more likely they were overwhelmed by the amount and/or type of information, resulting in difficulty assimilating the information to support their decision-making task.

We examined how varying the transparency of agent reasoning affected operator workload in a route selection task, and how the differing measures of workload compared in assessing and understanding cognitive workload. Participants guided a three-vehicle convoy safely through a simulated environment of which they had a limited amount of information, while maintaining communication with command and monitoring their surroundings for threats. The intelligent route-planning agent assessed potential threats and suggested changes to the convoy route as needed. Each participant was assigned to one of three agent reasoning transparency conditions. Contrary to our hypothesis, NASA-TLX Global workload measures indicated that workload decreased slightly as access to agent reasoning increased. However, psychophysical measures of workload disagreed with NASA-TLX global results. Comparison of individual NASA-TLX workload factors with the psychophysical measures indicated that performance satisfaction was highest in the intermediary transparency condition, and the addition of ambiguous information in the highest transparency condition increased effort and resulted in increased complacent behavior. Recommendations for future workload analysis are offered.