Team Cognition as Interaction

Interactive Team Cognition

Interactive team cognition aligns with recent views of individual cognition (e.g., embodied cognition, activity theory) that recognize that cognition can reside outside of the head (Chemero, 2009; Nardi, 1996). Cognition can be found in artifacts such as checklists and digital calendars and, more generally, in the context surrounding the individual. For a team, this context includes the other team members. Interactions with the other team members, often in the form of explicit communication, are instrumental to cognitive processing at the team level. Through interactions, team members coordinate cognitively with each other, integrating ideas and creating new knowledge. Consider, for example, a response to a burning building by a team of firefighters. One firefighter outside the building sees a blocked fire escape on the third floor. This person then communicates this information by radio to the firefighter inside who is searching for victims on the third floor and who now heads toward the part of the building where the blocked fire escape is located. Noises then direct this firefighter to three coworkers trapped in a stairwell leading to the exterior fire escape. This joint processing of cues has directed the search in an efficient and effective manner. Many examples exist of situations in which information was not communicated to the right team members in a timely manner (e.g., the response to Hurricane Katrina).

The theory of interactive team cognition posits that interactions, often in the form of explicit communications, are central to team cognition. More specifically,

Empirical Support for Interactive Team Cognition

A series of studies conducted in the author’s laboratory have supported interactive team cognition. Most of these studies have taken place in the context of a three-person task that simulates control of an unmanned aerial system engaged in a surveillance mission (Cooke & Shope, 2005). The three team roles are pilot, photographer, and mission planner. Each role has access to some unique and overlapping information, and the three team members must communicate (via voice or text chat) to take as many photos of designated ground targets as possible. Team and individual task performance is measured, as is team knowledge (of the task and of the team) and team process (coordination, communication, team situation assessment, and expert judgments of process). Across studies, we have seen that differences in team process relate to differences in team performance and that this relationship is stronger than that between team knowledge and team performance.

In this task, the measure of team performance is a composite score based on how well the team achieves its goals. Performance points are lost for poor or missed photos of targets, the amount of fuel and film used, and route violations. We have found, as shown in Figure 1, that teams’ performance scores improve across the first four missions and then tend to level off, reflecting a team learning curve (Cooke, Kiekel, & Helm, 2001). Interestingly, throughout our experiments, any changes in team knowledge (e.g., increases in knowledge similarity among team members) seem to take place very early (by the end of the first mission), whereas team process (communication patterns, coordination, interaction behaviors) continues to improve over the first four missions. The data suggest that as teams improve their scores, they are also improving the way they coordinate information passing—or the “push and pull” of information—across team members. In other words, experienced teams provide information to and request information from the right team member at the right time.

OPEN IN VIEWER

Fig. 1.

Team performance scores for each of three training conditions across nine 40-minute missions with an 8- to 10-week retention interval between Missions 5 and 6. Cross-trained teams had members trained on all three positions. Procedural teams were prescribed a rigid set of rules and procedures to which they were to adhere. Perturbed teams were blocked from being able to follow the rules and procedures that they were trained to follow and had to work around perturbations in order to succeed. Adapted from “Training Adaptive Teams” by J. C. Gorman, N. J. Cooke, and P. G. Amazeen, 2010, Human Factors, 52, p. 302. Copyright 2010 by the Human Factors and Ergonomics Society. Adapted with permission.

Teams not only learn as a unit, they also forget as a unit, as indicated by the performance drop after the 8- to 10-week interval between Missions 5 and 6 in Figure 1. The degree to which teams forget is best predicted by quality of interactions prior to the break, rather than drops in individual performance (Cooke, Gorman, Duran, Myers, & Andrews, 2013). Those teams with better team process before the break experience less of a performance decrement after the break.

Most of the teams in these experiments were composed of members who had never met each other before arriving at the laboratory. In one study (Cooke, Gorman, Duran, & Taylor, 2007), we collected data from a few teams with members who had worked together extensively, but not in the unmanned aerial system task environment. These teams achieved performance scores significantly higher than teams with unfamiliar members, and the advantage persisted throughout the study. The highest-scoring team was a team of three males who played an Internet-based video game together for months prior to the experiment. We reasoned that there was some interaction process (concerning how to push and pull information) that was developed in the game environment that transferred to our unmanned aerial system environment. Based on the team members’ backgrounds, it is also clear that task knowledge played little role in their superior performance.

We have also found that mixing up team members, as opposed to having teams stay together, results in teams that are more adaptive and resilient to change (Gorman & Cooke, 2011). The same result can be obtained by perturbing the simulated experience to block familiar interaction patterns, thereby encouraging the development of new ones. In these cases (i.e., perturbation training in Gorman, Cooke, & Amazeen, 2010; see Fig. 1), we see performance benefits at critical test missions over step-by-step procedural training or cross-training various roles. Changes in coordination patterns that are observed after mixing or perturbation are not accompanied by a change in knowledge (Gorman et al., 2010). In sum, the data suggest that for this task, team cognition in the form of interactions was tied to team performance, more so than shared mental models.

Implications and Future Directions

Conceptualizing team cognition as interactions among teammates has a number of interesting implications. First, because we can observe many interactions among team members, we can directly observe team cognition, and such observations may shed light on less transparent individual cognition. Second, unlike knowledge-based measures, measures of interactions can be obtained unobtrusively. In fact, the richness in communication data and physical interactions makes them ideal sensors for the cognitive activity of teams. Over time, this activity is likely to exhibit patterns and fluctuate with anomalies as team experience is gained and as the environment is perturbed. Third, the training and technological interventions that we might develop to facilitate team cognition would be interaction-based and not knowledge-based (e.g., practice interactions with intelligent agents).

What are the implications of interactive team cognition and the findings associated with it for the shared-mental-model approach? It is not that shared mental models or some distribution of knowledge across team members is unimportant. We view knowledge as a prerequisite for effective performance, but it is just a starting point for the action-oriented task that we used. According to the data described here, teams with ideal knowledge but ineffectual interactions will not be effective as a team, yet teams with sketchy knowledge, but effective interactions, may succeed.

However, in other tasks that require knowledge building (e.g., among teams of scientists; National Research Council, 2015), the balance may shift, with shared knowledge playing the leading role and team interactions supporting knowledge building. Future research is needed to identify the role of shared knowledge or interaction across a variety of team tasks. Further, Gorman (2014) has suggested that individual team members’ knowledge and interaction or coordination come together in teamwork. Coordinative linkages shape individual thought and action, and in turn individuals have a role in coordination. These kinds of interdependencies between knowledge and interaction and between individuals and teams require examination at a systems level, an area requiring further investigation.

There are other research gaps as well. Communication measures are useful in many settings, but not all team interaction involves explicit communication. Additional unobtrusive measures of team interaction are needed, along with analytical capabilities to reflect the dynamics of the interactions. Once we can accurately monitor and assess the team’s state, we can provide feedback to the team and possibly do so in real time. We can further use this feedback to improve team performance through training and technological interventions.

Taking a broader perspective on team cognition by considering cognitive processing as well as knowledge opens the door to additional measures, a richer understanding of team dynamics, and possibilities for new interventions to improve teamwork.