Negative Affect Reduces Team Awareness

Abstract

Objective:

This article presents research on the effects of varying mood and stress states on within-team communication in a simulated crisis management environment, with a focus on the relationship between communication behaviors and team awareness.

Background:

Communication plays a critical role in team cognition along with cognitive factors such as attention, memory, and decision-making speed. Mood and stress are known to have interrelated effects on cognition at the individual level, but there is relatively little joint exploration of these factors in team communication in technologically complex environments.

Method:

Dyadic communication behaviors in a distributed six-person crisis management simulation were analyzed in a factorial design for effects of two levels of mood (happy, sad) and the presence or absence of a time pressure stressor.

Results:

Time pressure and mood showed several specific impacts on communication behaviors. Communication quantity and efficiency increased under time pressure, though frequent requests for information were associated with poor performance. Teams in happy moods showed enhanced team awareness, as revealed by more anticipatory communication patterns and more detailed verbal responses to teammates than those in sad moods.

Conclusion:

Results show that the attention-narrowing effects of mood and stress associated with individual cognitive functions demonstrate analogous impacts on team awareness and information-sharing behaviors and reveal a richer understanding of how team dynamics change under adverse conditions.

Application:

Disentangling stress from mood affords the opportunity to target more specific interventions that better support team awareness and task performance.

Keywords

stress mood communication team awareness team cognition computer-supported cooperative work

Introduction

Team communication analysis is an effective method for understanding the nuanced dynamics of team performance in technologically complex work environments. Among the research conducted now on team cognition in stressful environments, this work is unique in that it separates out the effects of mood from the effects of stress, resulting in deeper insights into how mood and stress individually and jointly affect the effectiveness of team communication. This study employed a simulated crisis management task environment to explore the relationship between communication behaviors and team awareness (Gutwin & Greenberg, 2001), a social aspect of teamwork that proves critical to successful team cognition.

Team cognition is more than the simple aggregation of individual cognitive activities of team members. As cognition describes the information processing activities of an individual, team cognition describes the information processing activities of a team. One critical distinction is that in team cognition, the interplay between cognitive processes happens socially, rather than solely within the mind of an individual (Kiekel, Cooke, Foltz, & Shope, 2001). Communication analysis is therefore an ideal approach to examine team cognition.

The high psychological and physical stress found in time-pressured and hazardous work environments is often blamed for disastrous operator errors (Perrow, 1984). Further complications arise when teams are distributed and using computer-mediated communication rather than face-to-face interactions (Fiore, Salas, Cuevas, & Bowers, 2003). Emerging research in human–computer interaction (HCI) suggests that these accidents may be attributed not just to human error during the incident but to errors in system design, particularly those features that fail to account for psychological states such as frustration, anxiety, fear, and boredom (Ahn & Picard, 2006; Hudlicka, 2003; Kapoor, Burleson, & Picard, 2007). Recent research in computer-supported cooperative work has explored social and emotional engagement in teams (e.g., Mentis, Reddy, & Rosson, 2010; Nardi, 2005), but generally focusing on the role of emotions in the social environment rather than their effects on team cognition.

Emotion is acknowledged to have a significant and supportive role in cognitive processes (Isen, 1993; Simon, 1990). It is increasingly explored in the field of HCI (e.g., affective computing; Picard, 1997), yet it remains relatively underexplored within team cognition in technologically complex environments. This dismissal of emotional factors can be attributed to a number of biases, including simply viewing emotions as a symptom or indicator of stress, rather than a precipitating influence on individual and team cognition. Furthermore, many organizational cultures, military ones in particular, which are often the subject of these team cognition studies, often suppress the expression or discussion of emotion among operators (Delahaij, Gaillard, & Soeters, 2006; Harper, 2006). However, an abundance of individual-level research into stress and emotion reveals evidence that we should not consider one without the other (Lazarus, 1999). It follows that there is value in exploring the unique role of team mood in stressful collaborative processes. Computer-mediated communication often compromises personal interaction via emotional cues (Riva, 2002), making this issue especially salient for distributed teams. The study presented here, rather than examining the content of messages for emotional cues, instead analyzes the effect of mood and stress manipulations on the quantity and proportion of specific types of chat messages among dyads working in a fast-paced crisis management simulation.

Background

Defining Stress

This study focuses on short-term or situational stress, as opposed to long-term or life stress. The definition of stress used here descends from Lazarus and Folkman’s (1984) transactional model, wherein stress is regarded as the relationship between the perceived demands of a given situation and the individual’s appraisal of his or her resources to respond. When those demands appear to overwhelm the individual’s ability to cope, that individual experiences stress.

Stress is not a simple high or low variable, but rather a construct composed of multiple dependent processes. Matthews et al. (2002) provided a three-factor framework to distinguish stressors in a manner consistent with the process of appraisal described by Lazarus. These factors are worry, a cognitive factor involving self-monitoring, self-esteem, and distraction; distress, a cognitive and affective dimension of tension and unhappiness; and task engagement, which is cognitive, affective, and motivational, describing the enthusiasm or apathy of the individual’s response. Their instrument, the Dundee Stress State Questionnaire (DSSQ, Matthews et al., 1999), identified unique stress profiles in tasks such as reading, card sorting, and visual and auditory vigilance (Matthews et al., 2002). This model surpasses vague attributions of errors to stress, as though it were a black box. Rather, we may specifically say that stressors bearing these particular characteristics cause these particular responses.

Though there are emotional facets of stress, stress is not a mood but a cognitive process. Too often stress is referred to interchangeably with mood states, such as anxiety and fear, so it is essential to further explicate the conceptual differences between the two.

Defining Mood

Discussions of emotion often suffer from terminological ambiguity with the terms affect, mood, emotion, and feeling used interchangeably. Russell (2003) proposed a taxonomy that distinguishes emotional episodes, which are oriented toward a specific event (such as feeling joyful at receiving a gift), from moods, which are defined as a prolonged affective state with no object, such as simply feeling sad without any cognitive or reflective activity surrounding that feeling. It is Russell’s definition of mood that is employed here.

However, rather than using Russell’s popular model of positive and negative affect as a single bipolar construct, the study presented here employed Watson and Tellegen’s (1985) two-factor model, in which positive and negative affect are orthogonal constructs. Contrary to Russell’s model, in which positive and negative affect are inversely correlated, researchers have found many instances of little to no correlation between the two (e.g., Cropanzano, Weiss, Hale, & Reb, 2003; Matthews et al., 2002; Watson & Clark, 1997).

Disentangling Stress and Mood

Distinguishing between the cognitive impacts of stress and mood is critical, conceptually as well as operationally. Stress is often conflated with negative emotional experiences, such as fear or anxiety, which are sometimes referred to ambiguously as “stressful moods.” However, it is misleading to focus entirely on the negative connotations of stress. Many team contexts, including emergency medicine, military operations, and professional sports, rely on the stimulating effect of stress to motivate participants to high levels of performance (Lazarus, 2000) and report positive emotional states such as excitement or eagerness (Morris, Hancock, & Shirkey, 2004), even within organizational cultures that discourage the expression of emotion (Delahaij et al., 2006). These findings provide motivation to pursue a factorial research design where stress and mood are independently controlled, to tease out their individual effects on team cognition and illuminate any mediating or moderating relationships existing between the two.

Overlapping findings in the stress and emotion literature suggest multiple possible interactions between mood and stress states in terms of their joint impact on individual cognitive performance and, by extension, team cognition. Of particular interest here are effects on communication behaviors that may change team members’ mutual awareness of each other.

One frequently cited cognitive impact of stress is attentional tunneling (Wickens, 1996), in which individuals under stress maintain their ability to attend to their primary task but lose the capacity to manage peripheral or less threatening activities. This occurs in extreme environments, such as with fighter pilots under stress, when one aspect of the situation engages their entire attention, causing them to lose overall situation awareness. For example, the stressed pilot may lose control of the aircraft by attending solely to the target while neglecting dangerous changes in altitude.

A similar narrowing effect is also observed in response to negative mood states such as anxiety, concentrating an individual’s focus and biasing attention toward the most threatening features of the environment (Hudlicka, 2003). Numerous studies on mood and visual processing reveal how different mood states cause an individual’s awareness of the environment to become broad or narrow. According to the affect-as-information perspective (Schwarz & Clore, 1996), affect will influence task performance when it seems to provide task-relevant information. A happy mood, for example, provides the information that the task is proceeding well, whereas a sad mood indicates there are problems and caution is recommended. That message of caution can result in more rigid thinking, fixation on local rather than global features, and a more cautious approach in decision making. Those experiencing a happy mood, however, are more likely to employ novel approaches to problem solving and be aware of global features (Gasper & Clore, 2002).

Team Awareness

The narrowing of individual attention could lead to a loss of team awareness, whether by failing to attend to the needs of other team members or neglecting opportunities to ask for information or assistance. In the context of collaborative systems, successful team cognition depends on high levels of awareness of fellow team members, the status of their tasks, and the events happening within the shared workspace. These factors combine to create workspace awareness (Gutwin & Greenberg, 2001). The team awareness component under investigation in the present study relates to simplifying communication and managing how tightly or loosely partners are working together from moment to moment.

In fast-paced, stressful, and complex work environments, executing the current task and planning the next one require rapid and smooth transitions in team interactions—knowing who you should be helping or who should be helping you. Peripheral monitoring of teammate activities supports team cognition by identifying appropriate opportunities to collaborate and reducing inappropriate and frustrating interruptions. Indicators of these monitoring behaviors supporting team awareness should be evident in the communications between team members.

Driskell, Salas, and Johnston (1999) defined a similar concept, team perspective, as the combination of perceived team interdependency and the shared mental model of the team’s present task. They noted that broader team perspective was found among teams that had to complete a task collaboratively, compared to teams that worked independently. However, high stress conditions diminished team perspective among even those working collaboratively, echoing the tunnel vision effect described above, but at a teamwork level—changing the individual’s perspective from a group-focus to a self-focus. Furthermore, their analysis revealed the impact of stress on performance was in fact being mediated by team perspective. This serves as a valuable example that in stressful situations, it is not stress alone that affects outcomes.

Study Rationale

The following dyadic communication data are a subset of a larger study examining multiple dimensions of team collaboration and performance in NeoCITIES, a computer-mediated simulated task environment based on crisis management (Pfaff, 2008). In that study, hypotheses were developed regarding the effects of mood and stress on team task performance, but no specific a priori hypotheses were developed for the communication measures. Their role in the larger study was primarily to reveal underlying reasons for changes in team performance resulting from mood and stress. There were no direct mappings between specific message categories in the chosen coding scheme and the trends related to team awareness described previously. Rather, the intention of the chat analysis was to see what aggregate picture appeared from the relative changes and trade-offs in communication behaviors.

Method

Participants

In exchange for course credit, 42 students (N = 21 dyads) in a junior-level undergraduate HCI course took part in this study. Participants ranged from 19 to 30 years old (M = 21.1 years, SD = 1.98). In all, 38 were men and 4 were women.

Procedure

Participants were quasi-randomly assigned to teams of six for the duration of two experiments conducted in five sessions over a period of six weeks (only the first experiment is reported here). Within each team, participants were randomly assigned to one of three dyads—Police, Fire/EMS, or Hazardous Materials (Hazmat)—and within each dyad were assigned the role of either information manager (IM) or resource manager (RM). Each dyad faced an equal workload.

The first session of the experiment was dedicated to training on the NeoCITIES simulation. NeoCITIES is a six-person online video game based on a 911 dispatch center. Participants were isolated at individual computer terminals with an interface including a map of the town with icons showing the location and status of each event, a text-based chat window, an event tracker showing the description and status of ongoing events, and additional tabbed pages for activity logs and other information. For complete details of the construction and game play of NeoCITIES, see McNeese et al. (2005) and Jones (2006).

Over the course of each scenario, emergencies were presented at predetermined intervals simultaneously to the three IMs, who could freely communicate with each other. When an IM determined that his or her dyad should take action, the IM passed that event along to the RM, who had exclusive control of the dyad’s emergency resources. RMs could communicate only with their own IMs. Communication was via a text-based chat module in the game interface. Only the within-dyad communications (between IM and RM) are presented here.

Following the training session, participants returned each week at the same time for two more sessions. The experiment used a 2 (stress: none, high) × 2 (mood: sad, happy) within-subjects full-factorial design. Groups completed two conditions per session with a 10-min break in between conditions (the mood and stress conditions were counter-balanced to avoid ordering effects). The stress level remained constant in each session, with the mood level alternating from the first trial to the second. For each trial, participants would watch a movie clip (the mood manipulation), complete a survey that included both the mood manipulation check and the stress pretest, play the NeoCITIES simulation for 10 min with their team, and finally complete a final questionnaire, including the stress posttest.

Manipulations and Measures

The time pressure manipulation had participants addressing either 18 or 30 events over a 10-min session in the NeoCITIES simulation. The mood manipulations in both experiments used pretested happy and sad clips from popular movies, a mood induction procedure that has worked effectively in several studies (e.g., Cavallo & Pinto, 2001; Wang, LaBar, & McCarthy, 2006).

The 24-item Short Stress State Questionnaire (SSSQ; Helton & Garland, 2006), based on the 90-question DSSQ, provided a rapid and reliable assessment of the three primary stress dimensions: task engagement (Cronbach’s α = .77), distress (Cronbach’s α = .75), and worry (Cronbach’s α = .72). The mood manipulations were checked immediately after watching the video clips using items from the Positive and Negative Affectivity Scales (PANAS; Watson, Clark, & Tellegen, 1988) and Short-Form State-Trait Anxiety Inventory (STAI; Marteau & Bekker, 1992). As the SSSQ, PANAS, and STAI measure related concepts and even share some questions, factor analysis provided a two-factor solution for the mood measures: PA/Engaged (“I feel active,” “I feel enthusiastic,” “I feel determined”; Cronbach’s α = .79) and NA/Anxious (“I feel upset,” “I am worried,” “I feel dissatisfied,” “I feel distressed”; Cronbach’s α = .72).

Quantitative task performance measures alone (e.g., a team’s score) are insufficient for a holistic assessment of team performance (Cooke, Salas, Kiekel, & Bell, 2004; MacMillan, Entin, & Serfaty, 2004). Team communication has been identified as an important process variable in teamwork at the input, throughput, and output stages (Dickinson & McIntyre, 1997). Consequently, effective metrics have been developed according to various taxonomies that identify message types and link them to specific components of collaborative team functions.

Approaches available for coding and analyzing communications each have their advantages and constraints. These range from fast but simple quantitative methods to complex and rich qualitative methods. Striking a balance in the middle is the approach of developing a well-defined and comprehensive set of coding categories that captures both the meaning and quantity of relevant types of messages (Entin & Entin, 2001). This approach is a good fit when the primary interest is to observe proportional shifts between various types of statements under different conditions, as was the case in this study, rather than a more sophisticated conversation analysis that evaluates and models group conversation behaviors, as might be done using sequential analysis (see Jeong, 2005).

Several criteria led to the selection of a balanced method as the most appropriate for this study. The nature of verbal exchanges in the NeoCITIES task is limited in its depth and complexity (typically 5–15 statements total per 10-min trial). Statements were simple and to the point. For example, a typical exchange between an IM-RM pair would be, “IM: Send some fire trucks to #4; RM: OK.” The next exchange a few moments later concerns another event entirely. An exchange on one issue rarely exceeded three lines; hence, it is a stretch to consider these as “conversations.” However, there is still much value in analyzing individual statements. Orasanu and Fischer (2008) note how stress may decrease explicitness in communication, such as the choice between saying “OK” and “I sent two fire trucks.” This can be investigated for mood as well. In addition, the ratios between requests and responses can quantify how well team members anticipated each other’s needs.

The chat data in this study were coded using a scheme adapted from Entin and Entin (2001) developed for the Adaptive Architectures for Command and Control. This coding scheme identifies both transfers and requests in three categories—information (regarding the status of events), action (referring to ground-level activities), and coordination (delegating or accepting responsibility to act)—in addition to acknowledgments like “OK.” The definitions of these messages were tailored for the activities within the NeoCITIES task, and an additional category was added to account for off-task messages, for a total of eight unique message types (see the appendix). The coding scheme also included ratios of anticipation, computed by dividing the number of transfers by the number of requests in a given category. For example, teams that transfer information frequently without first being prompted with many requests will have an information anticipation ratio greater than 1.0, whereas teams that respond only to frequent and repeated prompting will have a lower ratio. Team members anticipating the needs of others and taking the initiative to transfer information demonstrate higher team awareness than members requiring one or more prompts to communicate. Serfaty, Entin, and Johnston (1998) utilized this approach to validate the efficacy of a team adaptation and coordination intervention, finding increases in the anticipation ratios that indicated significantly greater communication efficiency. Applied to the NeoCITIES task, however, the focus was not on efficiency, but rather on how mood and stress influenced the team’s communication style, as revealed by the relative proportions of these different types of messages.

Results

Manipulation checks verified that the mood and stress conditions were effective. The measures of self-reported stress and mood states were averaged across both members of each dyad and analyzed using within-subjects ANOVAs (because of the small sample, the level of significance was relaxed to α = .10). Participants reported higher positive affect in the happy condition (M = 1.80, SE = 0.12) than in the sad condition (M = 1.14, SE = 0.12), F(1, 56.98) = 45.53, p < .01. Participants also reported higher negative affect in the sad condition (M = 0.81, SE = 0.07) than in the happy condition (M = 0.19, SE = 0.07), F(1, 57.14) = 91.30, p < .01. The differences between the post- and pretask measurements of the three dimensions of the SSSQ were tested for the effect of the stress manipulation. Stressed participants reported lower engagement (M = 0.83, SE = 0.13) than did nonstressed participants (M = 1.10, SE = 0.13), F(1, 58.77) = 3.31, p < .10. Stressed participants reported more worry (M = 0.19, SE = 0.08) than did nonstressed participants (M = −0.03, SE = 0.08), F(1, 58.92) = 4.20, p < .05. No significant effect was found for distress.

Chat logs for each dyad were analyzed by two coders. The coder’s ratings were highly correlated, r(894) = .97, p < .01. Each line of chat was treated as a single statement and assigned the most appropriate code. Compound statements, though rare, were split onto two lines before coding. Fortunately, the brevity and focus of the chat statements led to little ambiguity in coding (as evidenced by a very high interrater reliability).

All message proportions and anticipation ratios were tested using 2 (stress: none, high) × 2 (mood: sad, happy) factorial repeated-measures ANOVAs. One dyad was excluded from all conditions for not following the experimenter’s directions, and one dyad missed a session. Coordination requests and transfers are common between IMs but were very rare in the chat between IM and RM and were not analyzed. Table 1 summarizes the ANOVAs for all of the measures, and Table 2 reports the means and standard errors for the significant results.

Table 1:

Factorial Repeated-Measures ANOVA of Messages and Anticipation Ratios

	Mood		Stress		Mood × Stress
Measure	F(1, 18)	r	F(1, 18)	r	F(1, 18)	r
Total messages	0.05		5.91**	.50	0.80
Overall anticipation	0.20		4.72**	.46	0.52
Information anticipation	0.02		2.96		0.36
Action anticipation	4.53**	.45	2.44		0.41
Information requests	1.44		13.23***	.65	0.00
Information transfers	0.27		7.59**	.54	0.63
Action requests	0.00		0.07		0.83
Action transfers	0.19		2.79		0.27
Acknowledgments	3.39*	.40	0.74		0.12
Off topic	0.41		0.10		2.07

Note. Effect sizes are reported only for significant results.

p < .10. **p < .05. ***p < .01.

Table 2:

Effects of the Time Pressure Stressor and Mood Manipulation

	No Stress		Stress
Measure	M	SE	M	SE
Total messages	2.53 (6.39)	0.40	3.28 (10.77)	0.34
% Information request	0.16 (2.53%)	0.03	0.30 (9.06%)	0.05
% Information transfer	0.43 (18.15%)	0.06	0.59 (34.57%)	0.05
Overall anticipation	1.31 (1.70)	0.07	1.46 (2.12)	0.08

	Sad Mood		Happy Mood
	M	SE	M	SE

Action anticipation	1.04 (1.08)	0.04	1.14 (1.30)	0.05
Acknowledgments	0.30 (9.18%)	0.03	0.21 (4.24%)	0.03

Note. A square-root transformation of all measures was made for normal distributions. Back-transformed means appear in parentheses after the results. An anticipation value of 1.0 indicates a 1:1 ratio of transfers to requests. A value greater than 1.0 indicates more transfers than requests, whereas a lower value means fewer transfers than requests.

Total messages per trial increased under the time pressure stressor compared to the normal condition. Specifically, the percentages of information transfers and information requests significantly increased under time pressure. A full discussion of task performance is beyond the scope of this article, but it is worth noting that the percentage of information requests was negatively correlated with task score, r_s = −.28, p < .05. The overall anticipation ratio increased under time pressure as well.

Action anticipation was higher in the happy mood than in the sad mood. Dyads sent more action transfer messages (“I sent two fire trucks”) than requests (“Will you send more fire trucks?”) in the happy condition than the sad condition. Last, the percentage of acknowledgment messages was higher in the sad condition than in the happy condition.

Discussion

Several findings revealed different reactions to stress in light of the mood state. Time pressure could be considered a standard exemplar from Lazarus’s transactional model of stress: In the high-stress condition, participants perceived the amount of work to exceed their ability to respond. Participants perceived this manipulation as stressful according to the motivational (less engaged) and cognitive (more worried) dimensions of the SSSQ instrument but not the affective dimension (distress). For the type of tasks involved in the NeoCITIES simulation, the increased worry is expected. Participants would be increasingly concerned about their ability to keep up as events accumulated faster than they could resolve. The loss of engagement under the time pressure condition may result from the fact that compensation (course credit) was not tied to task performance. Once the task got too hard, participants may have lost interest in investing energy into a task when no reward was at stake.

In the flurry of activity created by time pressure, it is not surprising to see increased communication within dyads. It is especially encouraging to see dyads engage preemptively under stress, as evidenced by the increased overall anticipation ratio. The increase in action anticipation in the happy condition suggests that positive mood enhanced team awareness, whereas the sad mood reduced it: “He’ll do his job, and I’ll do mine.” This result parallels the cognitive impacts of negative moods focusing attention on local details rather than the big picture (Gasper & Clore, 2002). The trend of increased acknowledgments in the sad mood condition, in conjunction with the previous finding on action anticipation, suggests a trade-off in which participants in the sad mood tended to opt for simple acknowledgments of action requests (e.g., “OK”) rather than making the effort for an explicit action transfer statement (e.g., “I’m sending a fire truck”).

The increase in the proportion of information requests under time pressure helps identify cognitive processes affected by the stress manipulation, specifically memory and attention. The negative correlation between information requests and task score reveals that dyads that found themselves making information requests more frequently ultimately did worse on the task. Information requests fall into two general categories. Many are retrospective and analytical: “What did you do about event number 5?” or “Do you know why we failed on that car crash event?” Others seek information necessary to make the correct choice: “How many people can we transport in an ambulance?” A high proportion of information requests suggests that the dyad was engaging in higher amounts of sense-making behavior, indicating difficulty in attending to the details of multiple events and holding them in memory. In the stress condition, participants had to repeatedly shift their attention from the current task to incoming events, changes in status for ongoing events, or chat messages from their partner. These interruptions interfere with memory formation by limiting the amount of attention allotted to each event and interrupting the processes of consolidating and storing information after an event (Mendl, 1999). Interruptions also lead toward frustrated emotional states (Mentis, 2007), providing a partial explanation for how the effect of stress on performance was mediated by mood (Pfaff & McNeese, 2010).

Practical Implications

Mitigating the task conditions that make work stressful and emotionally charged is difficult, yet systems can be designed to better support communication between operators under such conditions. The communication structure in NeoCITIES is deliberately restricted to closely observe information moving between players, but this problem affects any situation where multiple teams must develop and sustain a common operational picture (McNeese et al., 2006). Decentralization and lack of communication system standards affect the interoperability of local emergency services (e.g., police and fire) as much as multinational forces in a peacekeeping operation. Solomon (1997), analyzing the social elements of information behavior in sense making, stresses the criticality of integrating information and communication systems into unified and seamless task processes. Designers are advised to recognize the effect of mood on how teammates interact with each other as sources of information (Dervin & Reinhard, 2007). Special attention must be taken to avoid design flaws leading to negative emotional experiences, such as frustration resulting from repeated interruptions.

Hellar (2010) implemented one practical intervention in the NeoCITIES interface to address some of the impacts of mood and stress on team awareness. Specifically, he added an interface element to monitor teammates’ resources and task loads. This successfully reduced the need for explicit requests for assistance, an action known to be inhibited under time pressure or, more specifically, by the negative mood induced by that stressor (Pfaff & McNeese, 2010). Implicitly sharing that information in the interface subsequently improved team task performance and enhanced team awareness.

A frequent problem with text-based communication is the unclear deictic reference. Although “this one” and “here” are efficient shorthand for referring to an object or place, distributed team members can easily misunderstand each other without accurate shared reference points (Gutwin & Greenberg, 2001). Therefore, systems supporting an over-the-shoulder view of a partner’s workspace, or a virtual “pointing finger” taking the place of a real finger in face-to-face collaboration, will help overcome the omission of details in the terse and abbreviated communication patterns observed in sad moods.

Last, automated stylometric text analysis techniques can identify emotional cues in blogs and chat data, in some cases as well as humans (Mishne & De Rijke, 2006). Real-time monitoring of communications using these methods could automatically initiate appropriate interventions on behalf of a team member in distress.

Limitations and Future Work

An obvious limitation of the work was the convenience sample of college students as participants, particularly with the overwhelming number of men in the sample. Of the 21 dyads, four were mixed gender, whereas the rest were all male. Though the data are insufficient for a valid analysis, the mixed-gender dyads performed noticeably worse than the all-male dyads. Possible explanations may be drawn from other studies showing how college-age men value and trust other men more than women in technological contexts (Joshi & Schmidt, 2006; Williams, Ogletree, Woodburn, & Raffeld, 1993) and act more competitively than cooperatively in mixed-gender pairs (Valenzuela & Raghubir, 2007).

One of the many advantages of using simulated task environments like NeoCITIES is the capacity to capture volumes of quantitative and qualitative data in real time for multiple simultaneous research objectives. Future plans include analyzing the between-dyad communication data and correlating communication behaviors with performance. The latter work intends to identify whether certain communication patterns act as coping mechanisms under stressful conditions. Further research plans to extend this work with additional stressors salient to the domain of team cognition, including task ambiguity and dynamic team structures.

Key Points

The purpose of this study was to identify how stress and mood affect information sharing and team communication behaviors in a technologically complex work environment.

A specific focus was the effects on team awareness, an essential component of successful team cognition.

An experiment was designed to assess the main effects and interactions of mood and stress on these aspects of team collaboration, using time pressure as the stress manipulation, and happy or sad movie clips as the mood manipulation.

Teams in happy moods demonstrated more anticipatory communication patterns and more detailed verbal responses to teammates than those in sad moods, indicating enhanced team awareness.

In addition to providing openings to further detailed research on the relationship of specific stressors to specific mood states in the domain of team cognition, these results provide opportunities for adaptations of systems and procedures to help mitigate these effects in computer-mediated collaborative environments.

Footnotes

Appendix

Coding Scheme

Measure	Description	Example
Total communications	Total number of messages in the session
Information request	Request for information	“What happened on number 5?”
Information transfer	Transmission of information	“I don’t think we got there in time.”
Action request	Request for a teammate to take specific ground action	“Go ahead and send more squad cars.” (Police IM to RM)
Action transfer	Statement of a specific ground action taken or to be taken	“Sent two cars.” (Police RM to IM)
Coordination request	Request to coordinate or take responsibility for an action	“Number 7 isn’t ours, it’s for the fire team.” (Police RM to IM)
Coordination transfer	Agreement to coordinate or take responsibility for an action	“We’ll take care of 7.” (Fire IM to police IM)
Acknowledgment	Acknowledgment without any task-specific message content	“OK” or “Got it”
Insignificant utterances	Task-irrelevant statements or chatter	“This is boring . . . ha ha”
Overall anticipation	Sum of transfers divided by sum of requests
Information anticipation	Information transfers divided by information requests
Action anticipation	Action transfers divided by action requests
Coordination anticipation	Coordination transfers divided by coordination requests

Mark S. Pfaff is an assistant professor of informatics at Indiana University in Indianapolis, Indiana, where he specializes in team decision making in stressful and technologically complex environments. He received his PhD in information sciences and technology from the Pennsylvania State University in 2008.

References

Ahn

Picard

R. W.

(2006). Affective cognitive learning and decision making: The role of emotions. In Trappl

(Ed.), Cybernetics and Systems 2006 (pp. 1-6). Vienna: Austrian Society for Cybernetic Studies.

Cavallo

D. A.

Pinto

(2001). Effects of mood induction on eating behavior and cigarette craving in dietary restrainers. Eating Behaviors, 2, 113–127.

Cooke

N. J.

Salas

Kiekel

P. A.

Bell

(2004). Advances in measuring team cognition. In Salas

Fiore

S. M.

(Eds.), Team cognition: Understanding the factors that drive process and performance (pp. 83–106). Washington, DC: American Psychological Association.

Cropanzano

Weiss

H. M.

Hale

J. M. S.

Reb

(2003). The structure of affect: Reconsidering the relationship between negative and positive affectivity. Journal of Management, 29, 831–857.

Delahaij

Gaillard

Soeters

(2006). Stress training and the new military environment. In Human dimensions in military operations—Military leaders’ strategies for addressing stress and psychological support (pp. 17A-11–17A-10). Neuilly-sur-Seine, France: RTO.

Dervin

Reinhard

C. D.

(2007). How emotional dimensions of situated information seeking relate to user evaluations of help from sources: An exemplar study informed by sense-making methodology. In Nahl

Bilal

(Eds.), Information and emotion (pp. 51–84). Medford, NJ: Information Today.

Dickinson

T. L.

McIntyre

R. M.

(1997). A conceptual framework for teamwork measurement. In Brannick

M. T.

Salas

Prince

(Eds.), Team performance assessment and measurement: Theory, methods, and applications (pp. 19–44). Mahwah, NJ: Lawrence Erlbaum.

Driskell

J. E.

Salas

Johnston

J. H.

(1999). Does stress lead to a loss of team perspective? Group Dynamics: Theory, Research, and Practice, 3, 291–302.

Entin

E. E.

Entin

E. B.

(2001, June). Measures for evaluation of team processes and performance in experiments and exercises. Paper presented at the 6th International Command and Control Research and Technology Symposium, Annapolis, MD.

10.

Fiore

S. M.

Salas

Cuevas

H. M.

Bowers

C. A.

(2003). Distributed coordination space: Toward a theory of distributed team process and performance. Theoretical Issues in Ergonomics Science, 4, 340–364.

11.

Gasper

Clore

G. L.

(2002). Attending to the big picture: Mood and global versus local processing of visual information. Psychological Science, 13, 34–40.

12.

Gutwin

Greenberg

(2001). The importance of awareness for team cognition in distributed collaboration (Rep. No. 2001–696–19). Calgary, Alberta, Canada: University of Calgary, Department of Computer Science. Retrieved from www.cpsc.ucalgary.ca/grouplab/papers/

13.

Harper

(2006). No pain, no gain: Pain behaviour in the armed forces. British Journal of Nursing, 15, 548–551.

14.

Hellar

D. B.

(2010). NeoCITIES: A simulated command and control task environment for experimental research. In Proceedings of the 54th Annual Meeting of the Human Factors and Ergonomics Society (pp. 1027–1031). Santa Monica, CA: Human Factors and Ergonomics Society.

15.

Helton

W. S.

Garland

(2006). Short Stress State Questionnaire: Relationships with reading comprehension and land navigation. In Proceedings of the 50th Annual Meeting of the Human Factors and Ergonomics Society (pp. 1731–1735). Santa Monica, CA: Human Factors and Ergonomics Society.

16.

Hudlicka

(2003). To feel or not to feel: The role of affect in human–computer interaction. International Journal of Human-Computer Studies, 59, 1–32.

17.

Isen

(1993). Positive affect and decision making. In Lewis

Haviland-Jones

J. M.

(Eds.), Handbook of emotions (2nd ed., pp. 261–277). New York, NY: Guilford.

18.

Jeong

(2005). A guide to analyzing message-response sequences and group interaction patterns in computer-mediated communication. Distance Education, 26, 367–383.

19.

Jones

R. E. T.

(2006). The development of an emergency crisis management simulation to assess the impact a fuzzy cognitive map decision-aid has on team cognition and team decision-making (Doctoral dissertation, Pennsylvania State University). Available from ProQuest Digital Dissertations database. (AAT 3231841)

20.

Joshi

K. D.

Schmidt

N. L.

(2006). Is the information systems profession gendered? Characterization of IS professionals and IS careers. ACM SIGMIS Database, 37(4), 26–41.

21.

Kapoor

Burleson

Picard

R. W.

(2007). Automatic prediction of frustration. International Journal of Human-Computer Studies, 65, 724–736.

22.

Kiekel

P. A.

Cooke

N. J.

Foltz

P. W.

Shope

S. M.

(2001). Automating measurement of team cognition through analysis of communication data. In Smith

M. J.

Salvendy

Harris

Koubek

R. J.

(Eds.), Usability evaluation and interface design (pp. 1382–1386). Mahwah, NJ: Lawrence Erlbaum.

23.

Lazarus

R. S.

(1999). Stress and emotion: A new synthesis. New York, NY: Springer.

24.

Lazarus

R. S.

(2000). How emotions influence performance in competitive sports. Sport Psychologist, 14, 229–252.

25.

Lazarus

R. S.

Folkman

(1984). Stress, appraisal, and coping. New York, NY: Springer.

26.

MacMillan

Entin

E. E.

Serfaty

(2004). Communication overhead: The hidden cost of team cognition. In Salas

Fiore

S. M.

(Eds.), Team cognition: Understanding the factors that drive process and performance (pp. 61–83). Washington, DC: American Psychological Association.

27.

Marteau

T. M.

Bekker

(1992). The development of a six-item short-form of the state scale of the Spielberger State-Trait Anxiety Inventory (STAI). British Journal of Clinical Psychology, 31, 301–306.

28.

Matthews

Campbell

S. E.

Falconer

Joyner

L. A.

Huggins

Gilliland

. . . Warm

J. S.

(2002). Fundamental dimensions of subjective state in performance settings: Task engagement, distress, and worry. Emotion, 2, 315–340.

29.

Matthews

Joyner

Gilliland

Campbell

S. E.

Falconer

Huggins

(1999). Validation of a comprehensive stress state questionnaire: Towards a state “big three.” In Mervielde

Deary

I. J.

De Fruyt

Ostendorf

(Eds.), Personality psychology in Europe (Vol. 7, pp. 335–350). Tilburg, Netherlands: Tilburg University Press.

30.

McNeese

M. D.

Bains

Brewer

Brown

Connors

E. S.

Jefferson

. . . Terrell

I. S.

(2005). The NeoCITIES simulation: Understanding the design and experimental methodology used to develop a team emergency management simulation. In Proceedings of the 49th Annual Meeting of the Human Factors and Ergonomics Society (pp. 591-594). Santa Monica, CA: Human Factors and Ergonomics Society.

31.

McNeese

M. D.

Pfaff

M. S.

Connors

E. S.

Obieta

J. F.

Terrell

I. S.

Friedenberg

M. A.

(2006). Multiple vantage points of the common operational picture: Supporting complex teamwork. In Proceedings of the 50th Annual Meeting of the Human Factors and Ergonomics Society (pp. 467-471). Santa Monica, CA: Human Factors and Ergonomics Society.

32.

Mendl

(1999). Performing under pressure: Stress and cognitive function. Applied Animal Behaviour Science, 65, 221–244.

33.

Mentis

H. M.

(2007). Memory of frustrating experiences. In Nahl

Bilal

(Eds.), Information and emotion (pp. 197–210). Medford, NJ: Information Today.

34.

Mentis

H. M.

Reddy

Rosson

M. B.

(2010). Invisible emotion: Information and interaction in an emergency room. In Proceedings of the 2010 ACM Conference on Computer Supported Cooperative Work (pp. 311–320). New York: ACM. doi:10.1145/1718918.171897510.1145/1718918.1718975

35.

Mishne

De Rijke

(2006). Capturing global mood levels using blog posts. In AAAI 2006 spring symposium on computational approaches to analysing weblogs (pp. 145–152). Menlo Park, CA: AAAI Press.

36.

Morris

C. S.

Hancock

P. A.

Shirkey

E. C.

(2004). Motivational effects of adding context relevant stress in PC-based game training. Military Psychology, 16, 135–147.

37.

Nardi

B. A.

(2005). Beyond bandwidth: Dimensions of connection in interpersonal communication. Computer Supported Cooperative Work, 14, 91–130.

38.

Orasanu

Fischer

(2008). Improving healthcare communication: Lessons from the flightdeck. In Nemeth

C. P.

(Ed.), Improving healthcare team communication: Lessons from aviation and aerospace (pp. 23–45). Aldershot, UK: Ashgate.

39.

Perrow

(1984). Normal accidents: Living with high-risk technologies. New York, NY: Basic Books.

40.

Pfaff

M. S.

(2008). Effects of mood and stress on group communication and performance in a simulated task environment (Doctoral dissertation, Pennsylvania State University). Available from ProQuest Digital Dissertations database. (AAT 3325967).

41.

Pfaff

M. S.

McNeese

M. D.

(2010). Effects of mood and stress on distributed team cognition. Theoretical Issues in Ergonomics Science, 11, 321–339. doi:10.1080/1463922100372918510.1080/14639221003729185

42.

Picard

R. W.

(1997). Affective computing. Cambridge, MA: MIT Press.

43.

Riva

(2002). The sociocognitive psychology of computer-mediated communication: The present and future of technology-based interactions. CyberPsychology & Behavior, 5, 581–598.

44.

Russell

J. A.

(2003). Core affect and the psychological construction of emotion. Psychological Review, 110, 145–172. doi:10.1037/0033–295X.110.1.14510.1037/0033–295X.110.1.145

45.

Schwarz

Clore

G. L.

(1996). Feelings and phenomenal experiences. In Higgins

E. T.

(Ed.), Social psychology: A handbook of basic principles (pp. 433–465). New York, NY: Guilford.

46.

Serfaty

Entin

E. E.

Johnston

J. H.

(1998). Team coordination training. In Cannon-Bowers

J. A.

Salas

(Eds.), Making decisions under stress (pp. 409–413). Washington, DC: American Psychological Association.

47.

Simon

H. A.

(1990). Alternative visions of rationality. In Arkes

H. R.

Hammond

K. R.

(Eds.), Judgment and decision making (pp. 97–113). Cambridge, UK: Cambridge University Press.

48.

Solomon

(1997). Discovering information behavior in sense making. II. The social. Journal of the American Society for Information Science, 48, 1109–1126.

49.

Valenzuela

Raghubir

(2007). The role of strategy in mixed-gender group interactions: A study of the television show “The Weakest Link.” Sex Roles, 57, 293–303.

50.

Wang

LaBar

K. S.

McCarthy

(2006). Mood alters amygdala activation to sad distractors during an attentional task. Biological Psychiatry, 60, 1139–1146.

51.

Watson

Clark

L. A.

(1997). Measurement and mismeasurement of mood: Recurrent and emergent issues. Journal of Personality Assessment, 68, 267–296.

52.

Watson

Clark

L. A.

Tellegen

(1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063–1070.

53.

Watson

Tellegen

(1985). Toward a consensual structure of mood. Psychological Bulletin, 98, 219–235.

54.

Wickens

C. D.

(1996). Designing for stress. In Driskell

J. E.

Salas

(Eds.), Stress and human performance (pp. 279–295). Mahwah, NJ: Lawrence Erlbaum.

55.

Williams

S. W.

Ogletree

S. M.

Woodburn

Raffeld

(1993). Gender roles, computer attitudes, and dyadic computer interaction performance in college students. Sex Roles, 29, 515–525.