Chapter for Handbook of Applied Cognition, Second Edition Nancy J. Cooke, 1, 3 Jamie C. Gorman, 2, 3 and Jennifer L. Winner 1, 3
Arizona State University East1
New Mexico State University2
Cognitive Engineering Research Institute3
Nancy J. Cooke, Ph.D.
Applied Psychology Unit
Arizona State University East
7001 E. Williams Field Rd., Bldg. 140
Mesa, AZ 85212
Research and theory on cognitive structure and process occurring inside an individual’s head have dominated the last fifty years of scientific psychology. In this regard, cognition has primarily been attributed to the individual and not to a team or group. For the most part, applications of cognitive theories and findings have, as a result, tended to be individual-centric (e.g., cognitive processes involved in using a computer; individual decision making; situation awareness of a single pilot). However, the world has not kept pace. The growing complexity of technological systems and the associated cognitive burden of operating, maintaining, diagnosing, and overseeing such systems has necessitated teams or groups of humans. Problems posed to cognitive engineers in settings ranging from military systems to aviation and nuclear power plant systems depend increasingly on multiple humans interacting with each other and interfacing with complex technology. What does cognitive psychology have to say about these important problems? Do individually oriented cognitive theories and methods apply to teams? Are teams somehow different from a collection of individuals? How can we measure, assess, and design for team cognition? How can team cognition be applied to increase team effectiveness? These are some of the questions to be explored in this chapter.
First we will define team cognition in order to position this topic within the area of applied cognition and to provide some focus for the remainder of the chapter. Team researchers have distinguished between teams and groups, and in turn, between team research and applications and group research and applications. In this chapter we follow convention and define a team as a special type of group. Specifically, Salas, Dickinson, Converse, and Tannenbaum (1992) define a team as "a distinguishable set of two or more people who interact dynamically, interdependently, and adaptively toward a common and valued goal/object/mission, who have each been assigned specific roles or functions to perform, and who have a limited life span of membership" (p. 4). Thus teams are special types of groups that are interdependent and that have specific roles for different team members. A surgical unit is a team, but a jury is a group, as is a typing pool. A pilot, and co-pilot, and flight attendant comprise a team by this definition, but the faculty of a psychology department functions more as a group, as do religious and political groups. Command-and-control tasks and military planning tasks involve teams.
One dimension that tends to distinguish groups from teams is degree of homogeneity in regard to cognitive requirements of individuals. Teams tend to be more heterogeneous than groups by virtue of a division of labor that is tied to the definition of teams and necessitated by the requirements of increasingly complex systems. Related to the notion of a division of labor, heterogeneous team members complement each other, though they may not be very similar (like vinegar and oil). Nonetheless, they can work well together in the service of some higher-level functionality, such as a common or valued goal.
Much research has been done in the organizational management, computer supported cooperative work, and social psychology arenas on small group behavior and decision making (e.g., Kerr & Tindale, 2004). The research in this area tends to be on groups, not teams, and so some theories and findings may not apply. On the other hand, there are probably as many theories and findings that do apply to both groups and teams. Thus, in this chapter, we will consider the work on groups that may inform the theories, measures, and applications of team cognition.
Given the Salas, et al. (1992) definition of a team, what is team cognition? In this chapter we define team cognition as cognitive activity that occurs at a team level. For instance, planning is a cognitive activity and team members can carry out planning independently. But this does not by itself constitute team cognition, in the sense that planning has not been carried out via collaborative planning by two or more team members with a common goal. Thus, the team-level stipulation necessitates that there is interaction among the individuals. For example, by our definition, team situation awareness is not the respective situation awareness of each individual team member (Endsley & Jones 1997), but necessarily something based on interaction and probably something emergent.
The difference between a linear aggregate of individuals and interacting individuals is a very important, and somewhat controversial, distinction in team cognitive research (Cooke & Gorman, in press). Team cognition is different than the sum of the cognition of individual team members. It is the result that emerges during functional interactions of team members in pursuit of a common and valued goal. So how do we measure team cognition? How do we design technology to support this team-level cognition? How do we train teams to support team-level cognition? These are questions addressed in the remainder of this chapter. Answers to these questions are increasingly important as teams continue to supplant individuals as the human element in highly complex technological systems (e.g., socio-technical systems; Eason, 1988).
The sections in this chapter are organized in a nontraditional sequence that to the authors, characterizes the historical roots of this new field. Though several precursors to the study of team cognition such as team performance in the industrial/organizational tradition, management in the business tradition, and small group decision making in the social psychology tradition are not at all new, the study of team cognition as defined in the preceding paragraphs is no more than fifteen years old. Unlike other applied research topics for which application of well-developed theory is the end goal, for team cognition, application has been the main driver. The field was sparked by a need to improve the cognitive activity of teams. Measurement, empirical work, and theoretical development followed. Of course, existing theory from the precursor disciplines was also influential in the early days of the field and in reality, the interplay between application and theory is really more cyclical than linear, in this case the push stemming from application is noteworthy. Therefore, we have decided to structure this chapter accordingly. We start with the applied need, working our way to the developments in measurement and empirical work, and finally culminating in the continually evolving theory of team cognition.
THE NEED FOR A SCIENCE OF TEAM COGNITION
We often think of applications in cognitive psychology as the natural product of good cognitive theories. As the story goes, a theory is born, grows, is tested, is further developed, and eventually matures all in the pristine circumlocution of the lab. When ready the theory is (sometimes) pushed out into the world of application, often looking for a job and sometimes finding that it is poorly qualified to fit the existing need. This story does not characterize team cognition. While there were theories and sciences of teams and of group decision making they did not completely address the problem of team cognition. And there was a problem to be addressed.
Indeed much of the research on team cognition was fueled by a handful of disasters that occurred in a ten-year period and that pointed out problems with team performance in cognitively demanding arenas. The Three Mile Island and Chernobyl nuclear power plant accidents of 1979 and 1986, respectively both involved problems with the response of the operating crews in the control rooms (Gaddy & Wachtel, 1992). On January 28, 1986 faulty decision making at the organizational level (i.e., teams of teams) resulted in the mistaken and tragic launch of space shuttle Challenger (Vaughan, 1996). Then in July of 1988 the USS Vincennes, a US Naval warship, mistakenly shot down an Iranian airbus full of passengers (Collyer & Malecki, 1998). This incident, like the others, was tied to a complex web of causes and preexisting system weaknesses (Reason, 1997), but also like the others, was partially attributed to coordination problems in the command-and-control decision making.
These events of the 1980s highlighted the complexity of our human-technological systems, and not only in terms of the physical system, but also in terms of the socio-cognitive system. It became clear that human decision making and other cognitive activities were occurring in the context of a complex socio-technical system and that this type of cognition, so foreign from the isolated and artificial lab studies that dominate mainstream cognitive psychology, demanded attention. The naturalistic decision making movement fully appeared on the scene in the 1990s to address cognitive processing in complex dynamic environments (Zsambok & Klein, 1997). The cognitive engineering and decision making technical group of the Human Factors and Ergonomics Society also had its start in the 1990’s. The TADMUS (Tactical Decision Making Under Stress) research program of the Navy was initiated in 1990 as a direct result of the Vincennes incident with specific Naval applications as the target of this research (Cannon Bowers & Salas, 1998). This was not the case of a well-developed theory of team cognition looking for a problem, but a problem, or set of problems, in need of a theory. The TADMUS program fueled some of the first research on team cognition. It is significant that these research efforts were tied heavily to application—to enhancing team effectiveness in complex cognitive environments. Because an understanding of team cognition is a prerequisite to improving team effectiveness, the theoretical growth has occurred in parallel, but the research has been driven primarily by applied needs. Thus, applications are seen here not as the end result of various research programs, but as drivers of those programs.
Fifteen years have passed since the disasters and ensuing programs that sparked work on team cognition. Theoretical perspectives, methodologies, and research findings have progressed significantly and in parallel with specific applied solutions. As the field progresses, so do the applications. With disasters such as the Vincennes looming in the recent past, there was no time to wait for the science of team cognition. Applications to improve team decision making, for instance, were needed “yesterday.” Thus, the need spawned some initial applied solutions that were based on preliminary conceptions of team cognition primarily drawn from precursor and related disciplines (i.e., social psychology, management, information processing) and applied by those in industrial/organizational psychology, military psychology, and human factors. Efforts toward these early applications incited many research questions regarding measurement and concepts of team cognition. At the same time the applications evolve with the science of team cognition. Thus, in the following sections we describe some of the early applications of team cognition, saving the more recently applied products of this evolution for the end of the chapter.
In applied psychology and specifically in industrial/organizational psychology, there has been a long tradition of applying psychology to training and specifically, applying the study of teamwork to improving team training effectiveness (Ford, Kozlowski, Kraiger, Salas, & Teachout (1997). To address the need for training teams in more cognitively laden activities, team training programs began to focus more heavily on the cognitive training of teams.
For instance, Crew Resource Management (CRM) training programs (Helmreich, Merritt, & Wilhelm, 1999; Salas, Burke, Bowers, & Wilson, 2001) incorporate important cognitive skills such as team coordination, communication, and resource allocation. CRM training has been predominant in the aviation community and CRM training programs have been recently implemented in other team domains such as transportation and medicine (Salas, Wilson, Burke, & Wightman, in press).Evaluations of CRM training programs have been generally positive.
Cross training involves the training of individuals in the job skills associated with other team member positions. In jobs that are heavily cognitive, many of the skills are apt to be cognitive in nature. Cross training is generally effective, but the concept of team cognition and in particular, shared mental models and interpositional knowledge were drawn upon as explanations for the success of this approach (Cannon-Bowers, Salas, Blickensderfer, & Bowers, 1998; Cooke, Kiekel, Salas, Stout, Bowers, & Cannon-Bowers, 2003; Volpe, Cannon-Bowers, Salas, & Spector, 1996).
Other training strategies were readily adapted to cognitive tasks and include team leadership training and team self-correction which teaches team members to identify and correct problems in the team without help from an outside instructor (Salas & Cannon-Bowers, 2001)
In order to study team cognition in a laboratory, it became clear that rich synthetic environments were required that provided empirical testbeds for teams. Although there were team testbeds in place, there was a need for cognitive simulators for teams. Some initial testbeds grew out of the TADMUS program and were based on tactical decision making by crews on ships (e.g., TANDEM, Johnston, Poirier, & Smith-Jentsch, 1998).
In the mid to late 1990s there was a flurry of work devoted to the development of synthetic task environments for teams (Schiflett, Elliott, Salas, & Coovert, 2004). The Dynamic Distributed Decision Making Task (DDD, Kleinman, Young, & Higgins, 1996) was developed as a synthetic version of an Airborn Warning and Control System (AWACS) platform. This synthetic environment has since morphed to simulate a wide range of team tasks (e.g., a snowmobile-based search and rescue mission). Cooke and Shope (2002, 2004) developed a synthetic task environment to simulate Unmanned Aerial Vehicle (UAV) ground control by a team. This testbed was unique in its experimenter-friendly design (i.e., affording measurement and manipulation) and its applicability to heterogeneous team tasks in which team members all have very different jobs to do (i.e., navigator, pilot, photographer).
The act of developing these simulations required and generated extensive knowledge about the cognitive team tasks in the field. Questions were also raised about the concept of simulator fidelity (Cooke & Shope, 2004; Salas, Bowers, & Rhodenizer, 1998). As military doctrine changed and concepts such as network-centric warfare became common place, the applicability of these testbeds as potential training and interface testing environments also became clear.
Much of the work on decision aids and collaborative tools for teams and groupware had its start in business management and human-computer interaction. This work has been labeled CSCW (Computer Supported Cooperative Work), GDSS (Group Decision Support Systems), and groupware (Olson, Malone, & Smith, 2001). These applications were designed to facilitate business meetings, collaborative writing, and organizational decision making. More specific applications of groupware focused on facilitating the integration of multiple perspectives across a broad decision making domain (e.g., product development teams, Monplaisir, 2002) by improving the ability for collaboration from a distance.
Groupware researchers have experimented with various communication modes for distributed groups such as video, audio, and text (Daft & Lengel, 1986). However, while groupware can often lead to increased decision quality, it may have a negative impact in terms of time needed to make decisions, as well as overall group member consensus and sense of satisfaction (cf. cohesiveness; McLeod, 1992). Indeed the advantages of various CSCW applications seem dependent on the type of team and task (DeSanctis & Poole, 1991).
Some Unanswered Questions
These initial applications for improving team cognition were valuable in identifying research questions and guiding future research and development efforts in team cognition. As before, team cognition is a good example of the problem-driven approach of cognitive engineering.
Interests in team cognition, combined with successful tests of cross training, gave rise to the concept of shared mental model and the idea of interpositional knowledge. But there were also questions about scalability of cross training to large heterogeneous teams in highly complex settings such as shipboard command-and-control. Would cross training across all positions be practical? Is there another more streamlined approach (Cooke, et al., 2003)? Is the “shared mental models for all” an optimal goal state?
Groupware applications provided valuable insight into how distributed decision making may be facilitated by technology, but it also spawned questions about the costs and benefits of co-located versus distributed environments. With groupware technology also came an interest in ways to facilitate team situation assessment and group interaction processes associated with group decision making. Indeed, these were some of the same issues important those interested in team cognition. However, other questions were raised about differences between small groups and teams and between the business meeting and more heterogeneous, interdependent tasks.
One very important question kept surfacing; the question of measurement. How should we measure team cognition? The answer to this question was central for designing and evaluating the success of training programs, supportive tools and technologies, and even for understanding individual and team differences that may lead to applications in team composition and staffing (Morgan & Lassiter, 1992). Further, the measurement problem is not merely about eliciting team cognition from experts to build systems, but it is also about assessment and diagnosis. How can we measure team cognition so that deficiencies and strengths of a team can be assessed and diagnostic information can be provided to target specific team-level cognitive skills? Accordingly, one of the biggest measurement challenges is to consider how assessment and diagnosis could occur in an embedded, automated, and real-time fashion in an operational setting. We could then monitor team cognition, assess it on the fly, and provide diagnostically appropriate interventions, all in real time.
MEASURING TEAM COGNITION
Leveraging cognitive work at the individual level researchers have applied and adapted various individual measurement methodologies to teams. The exact nature of the measurement problem dictated specific types of methods over others. Initial forays into a team domain are usually accompanied by a need for techniques that elicit and analyze team cognition. These methods have been used to circumscribe the task or domain in terms of the knowledge, skills, and abilities required. Information gleaned from these activities is used to construct tests to assess a team with respect to a desired cognitive state. The general idea is that through assessment we can determine if a team’s cognition is improving with a particular intervention and rank order teams with respect to team cognition. Finally, with a deeper understanding of team cognition in a domain, the researcher can begin to identify diagnostic information for targeted interventions. This stage is analogous to understanding whether observed memory deficits in an individual are due to a biological cause or loss of motivation.
In this section we describe some of the measurement methods that have been applied to team cognition, some limitations of those measures, and some approaches to measuring team cognition that address those limits.