I Hope You're Happy Now: Emotional Contagion in CMC Groups

ABSTRACT

Though previous work has demonstrated transfer of sadness between CMC dyads, this study sought to better understand transfer of high-energy, negative emotion in groups. Affect was induced in one of three group members via a film clip after which participants engaged in a group activity to facilitate angry affect transfer. Emotional transfer occurs in this setting, evidenced by higher negative affect, more disagreement, higher verbosity, and use of more complex language in groups transferring angry affect compared to groups sharing no specific emotions.

Author Keywords

Computer-mediated communication, affect, group, emotion.

ACM Classification Keywords

J4: Social and behavioral systems: Psychology.

INTRODUCTION

Through emotional contagion, members of groups can influence the emotion or behavior of other members via the conscious or unconscious transfer of emotions, behaviors and attitudes [1] While emotional contagion has been examined in face-to-face (Ftf) dyads and groups, it is a little studied phenomenon in computer-mediated communication (CMC) [1]. Though early work suggested that nonverbal cues were essential for emotional expression and perception [2], recent work shows that emotional contagion occurs in CMC [3] [4].

Since ostentatious displays of emotion are discouraged in social settings, emotions must be conveyed in more subtle forms via physical cues [2]. As CMC lacks these cues (e.g., facial expressions, gestures, and vocal inflection), it is a reasonable assumption that emotional understanding would be impugned in CMC.

Work on Social Information Processing Theory (SIP) demonstrates that despite early claims that CMC relationships inhibit emotional expression, relationships can reach the same level of closeness in CMC as in Ftf over time. Emotion is detectable in CMC based on word choice, punctuation, and timing [5]. Participants in an experiment acted out either happy or sad emotions and linguistic analysis revealed that “sad” participants used fewer words, were more disagreeable, and responded slower than “happy” participants [4].

Additional research demonstrates the ability to convey and transfer emotions through linguistic cues. In this study, when one subject was induced to feel sad before a CMC interaction, the other subject was able to detect negative emotion in the subject, and reported feeling sadder after conversing with the partner. Sad participants used fewer words, disagreed more, used more negative-affect words, and responded more slowly time. Despite the absence of nonverbal cues in CMC, people can perceive and transfer negative emotion.

Because emotional detection in CMC relies on attention to linguistic cues, when cognitive resources are distributed among language of several individuals, will emotional detection be deterred? Our research explores this question, and determines whether contagion can occur in group CMC. Unlike those who multitask to a lower degree, heavy media multitaskers are less able to effectively switch tasks [6]. This suggests that when attention is dispersed, the ability to filter out irrelevant interference is weakened.

To address these issues, our study employs high-energy, negative emotion in the form of anger. Though previous contagion studies in Ftf contexts did not demonstrate differences in the degree of emotional contagion based on emotional energy, given the lack of physical cues in CMC, we predict that energy level will be an important factor in emotional contagion in CMC due to the higher visibility of high-energy emotion in text. We thus propose the following hypotheses:

H1: Groups sharing high-energy, negative emotion in CMC will report feeling more negative emotion, compared with group members sharing no specific emotions.

Previous research also suggests that linguistic differences should emerge between groups sharing emotions and emotionally neutral groups, thus we propose the following hypotheses:

H2: Groups sharing high-energy, negative emotion will use more negative affect words compared to groups sharing no specific emotions.

H3: Groups sharing high-energy, negative emotion will disagree more often compared to members in groups sharing no specific emotions.

Though previous research showed that inducing sadness led to participants using fewer words and less punctuation in CMC [3], we believe that using a high-energy, negative emotion (e.g., anger) may lead to the use of more words and punctuation due to its arousing nature. Thus:

H4: Groups sharing high-energy, negative emotion will use more words than groups sharing no specific emotion.

H5: Groups sharing high-energy, negative emotion will use more punctuation than groups sharing no specific emotion.

Methods

Participants

The participants (N=84) were students who either received class credit or $10 for their participation. There were 25 male and 47 female participants (ranging in age from 18 to 26 years old); two participants did not provide their gender. Participants signed up in groups of three friends with 27 same-sex and mixed gender groups participating in the study. Sixteen groups were randomly assigned to the angry affect group and twelve groups were randomly assigned to the neutral affect group.

Procedure

Phase I: When the participants arrived, groups were randomly assigned to either the angry or neutral affect condition. All participants were randomly assigned to one of three rooms and were told that they would be participating in two unrelated studies, the first being a “perceptions of film” study. Participants then watched a 7-minute video clip. In the angry affect condition, one participant (the angry emotion experiencer) was randomly assigned to view a clip from the film My Bodyguard depicting instances of bullying and injustice, which previous research has shown to induce anger. The other two partners of the group watched a clip from Before Sunset, a boring film clip that was chosen to have no effect on affect [3]. In the neutral affect condition, all three participants watched the Before Sunset clip.

After watching the clips, all participants filled out a questionnaire that measured emotion using a PANAS scale. This emotional scale was used to ensure the success of the anger manipulation and asked participants to assess their level of feeling on a 7-point Likert scale for twenty affect words (e.g., hostile, guilty, inspired, etc.). A high-energy, negative affect factor was created to assess the effectiveness of the anger manipulation on the experiencers in the present study and was reliable (Cronbach’s µ = .88).

Phase II: All participants were then informed that they would begin the second study, a “group multi-tasking in CMC” experiment. Their primary task was to chat with the other two members of their group via AOL Instant Messenger to generate a list of “Three Tips for Surviving Freshmen Year.” To motivate participants to talk about themselves (a highly emotional subject), every tip the group came up with was to be accompanied by an example explaining the tip’s personal relevance.

In the angry affect condition, the experiencer was asked to simultaneously listen to heavy metal music and solve word scrambles that all solved to words with an anger-related prime (e.g., limaec > malice) to maintain negative affect. In the neutral affect condition, participants listened to light jazz music and completed word scrambles that solved to neutral primes (e.g., ldbna > bland) to maintain neutral affect. The manipulation was presented as a study on how multiple stimuli affect group productivity. As added incentive, participants were told that the group with the best tips would be awarded $20 gift certificates.

Following the activity, all participants filled out a second questionnaire measuring their emotional state using the Circumplex Affect Scale [7]. Participants were additionally questioned about their perceptions of their group members’ performance and emotional states.

Participants were then debriefed, with particular focus on ensuring that participants in the angry affect condition did not feel anxious or angry after the clip and activity. Participants were given cookies upon exit to dispel residual negative affect.

Manipulation Check

Induction of anger via the My Bodyguard video clip was successful, with no gender effects. Experiencers who watched the anger-inducing clip reported having significantly more high-energy, negative affect on the PANAS scale (M = 2.57, SD = 1.06) than participants who watched the neutral clip (M = 1.23, SD = .28), t(80)= 9.21, p < .001.

A second manipulation check assessed experiencers’ mood after the group activity. The induction was partially maintained as participants who watched the angry clip were marginally more annoyed (M = 1.81, SD = .83) and frustrated (M = 1.81, SD = 1.05), than participants who watched the neutral clip (M = 1.38, SD = .80; M = 1.42, SD = .75), t(80) = 1.93, p = .06; t(80) = 1.72, p = .09. The reason for marginal significance after the group activity may have been due to the emotional contagion occurring between participants in the angry affect condition. An analysis comparing experiencers to the neutral affect condition participants revealed that experiencers were significantly more annoyed (M = 1.81, SD = .83) and frustrated (M = 1.81, SD = 1.05), than neutral affect participants (M = 1.29, SD = .69; M = 1.32, SD = .70, t(52) = 2.38, p < .05, t(52) = 2.04, p < .05, indicating that our manipulation was successfully maintained.

Linguistic Analysis

The language of group chats was analyzed using the Linguistic Inquiry and Word Count (LIWC) data analysis program [8]. LIWC explores word count, frequency of word use and punctuation based on a variety of different psychological dimensions. In the present study the following dimensions were explored: negative emotion words, discrepancies, word count, and punctuation.

Results and Discussion

Perception of Group Members

Initial analyses were conducted to determine whether participants perceived differences in emotionality of their partners, both within the angry affect and neutral affect conditions. These analyses revealed no significant differences within or between groups in perceived emotion of group members, suggesting that none of the participants were aware of emotional differences, despite the success of the angry affect manipulation on experiencers. There were also no differences in participants’ ratings of closeness with their group members or in participants’ perceptions of group members contribution to the group’s performance on the activity.

Emotional Contagion

Our first hypothesis suggested that groups sharing negative emotion would report feeling more negative emotion, compared with group sharing no specific emotions. If this was the case, partners who interacted with experiencers in the angry affect condition should report more negative affect than participants in the neutral condition. Analysis comparing only partners to the neutral condition revealed that partners reported being significantly more tense (M = 1.64, SE = .13), than participants in the neutral condition (M = 1.19, SE = .12), F(1, 21)= 6.96, p < .05. Experiencers also felt significantly more tense (M = 1.79, SE = .19) than neutral participants (M = 1.20, SE = .19), F(1, 21)= 5.78, p < .05. Means are in Table 1. These findings in combination suggest that the same negative emotion, tension, was transferred within groups sharing angry affect, thus supporting H1. No other significant differences between partners and neutral participants were observed.

Additional analyses were conducted to determine differences between emotions felt by anger experiencers, partners and the neutral affect condition to determine whether the group activity affected experiencers, partners and neutral affect participants differently. Table 1 indicates that partners felt significantly more glad and content than experiencers. Additionally, experiencers felt more alarmed than neutral participants.

	Angry		Neutral
Emotion	Experiencer	Partner	All
Glad	2.31a (.30)	3.25b (.21)	2.81ab (.18)
Content	2.75a (.23)	3.54b (.19)	3.39ab (.16)
Alarmed	1.63a (.30)	1.25ab (.13)	1.05b (.04)
Tense	1.81a (.26)	1.64a (.17)	1.18b (.06)

Table 1. Post-activity emotion means with standard deviations for experiencers, partners and neutral condition.

Note. Comparisons between means, specified by lowercase superscripts, are horizontal only. Cell means that do not share a letter in their superscripts differ at p < .05 according to Tukey’s HSD.

Linguistic Differences

The remaining hypotheses explored whether there would be linguistic differences between the angry affect and neutral affect conditions. Results for linguistic categories are displayed in Table 2. The second hypothesis (H2), which predicted that participants in the angry affect condition would use more negative affect words than neutral affect participants as there were no differences in the use of negative emotion words between these groups, F(2, 59) = 1.03, p = .364.

Next, we predicted that participants in the angry affect condition would disagree more than neutral affect participants (H3). There were significant differences in the amount of discrepancy words (e.g., couldn’t, should’ve, wouldn’t) used by participants, such that angry affect participants used significantly more discrepancy words than neutral participants, F(2, 59) = 6.39, p < .01, supporting H3.

The fourth hypothesis explored whether participants in the angry affect condition used more words in general than those in the neutral affect condition. This hypothesis was confirmed as partners and experiencers used significantly more words than neutral affect participants, F(2, 59) = 3.79, p < .05.

Our final hypothesis predicted that participants in the angry affect condition would use more punctuation than participants in the neutral condition. H5 was not supported as there were no differences between groups in punctuation use, F(2, 59) = .64, p = .533.

In addition to our hypothesized findings, some interesting differences emerged between participants. Patterns are displayed in Table 2.

Category	Angry		Neutral
	Experiencer	Partner	All	p
Word Count	227.09ab (28.49)	231.10a (19.32)	171.03b (13.98)	.03
Function	52.32a (1.81)	51.96a (1.32)	45.70b (1.20)	.00
Verb	16.81ab (.70)	16.85a (.82)	14.11b (.58)	.01
Present	11.78a (.80)	11.71a (.74)	9.94a (.41)	.05
Quantifiers	3.88a (.43)	3.84a (.35)	2.90a (.25)	.05
CogMech	19.21a (.68)	16.87a (.65)	14.10b (.67)	.00
Discrepancy	2.19a (.33)	1.87a (.22)	1.13b (.16)	.00
Tentative	3.24a (.42)	2.66ab (.31)	2.06b (.23)	.04
Preps	10.27a (.43)	9.95a (.57)	8.49a (.43)	.03

Table 2. Lingusitic profiles of emotion across angry and neutral groups.

Note. Comparisons between means, specified by lowercase superscripts, are horizontal only. Cell means that do not share a letter in their superscripts differ at p < .05 according to Tukey’s HSD.

Discussion

Emotional contagion in group CMC operates differently than with dyads [3]. Unlike previous studies [3] [4], participants were unable to recognize emotional differences in partners, suggesting that the process occurs via more subtle means in groups. This may be due to the noise of having to process multiple group members’ messages. Additionally, it seems that though angry affect expressed by experiencers does result in more negative affect in partners in the form of tension, partners are not experiencing anger. Though experiencers may feel more high-energy, negative emotion (e.g., frustration, annoyance) they are not expressing these frustrations outwardly to their partners, which may be socially inappropriate [2]. The finding that negative emotion words were not used more frequently by angry affect than neutral affect groups is reflective of this type of restraint on the part of experiencers. Indeed, no linguistic differences were found between partners and experiencers language patterns.

The finding that partners were significantly more content and glad than experiencers was curious and may be indicative of relief from mental exhaustion that they experienced by interacting with experiencers. Higher within group disagreement via linguistic discrepancies suggests that the interaction was more mentally taxing, which could explain these differences.

Though the finding that people disagreed significantly more between groups supports linguistic findings [3] [4], analyses in this study advance work in this area by demonstrating differential effects on word count. The induction of high-energy, negative emotion led both experiencers and partners to use significantly more words than neutral affect participants, providing further evidence of contagion of emotional energy within the angry affect group. Though early work in emotional contagion in CMC [3] [4] focused primarily on manipulating the valence of emotion in these contexts, our findings highlight the importance of considering emotional energy in understanding contagion processes in CMC specifically.

Additional linguistic findings in this study suggest that sharing high-energy, negative emotion affects complexity of language use. Participants in the angry affect condition used significantly more cognitive mechanism words and prepositions in their chats than neutral affect participants. These types of words are indicative of more complex language and are used more frequently in complex sections of texts (e.g., the discussion section of research articles) [8]. Given that the task instructions explicitly encouraged participants to share personal stories as examples for their tips, it seems that angry affect participants were more focused on completing the task based on the instructions than neutral affect participants. Though further work needs to be conducted to determine the specific relationship between emotion and performance, these findings have interesting parallels with earlier research on the relationship between stress and performance. Too much or too little stress can negatively affect performance, but people perform better at moderate stress levels [9]. Though we do not have data that speaks to actual performance per se, the differences in cognitive complexity between groups suggest that angry affect participants’ are engaging in more task-oriented behaviors, which should lead to better performance.

Conclusion

In addition to demonstrating emotional contagion in CMC contexts, the present study makes several important advances in understanding this process in CMC. First, this work demonstrates the importance of emotional energy in understanding the way that emotions affect CMC interactions.

Second, this study suggests that the process of emotional contagion in CMC is more complex than previously thought. Though participants in the angry affect condition shared tension, they differed in their experiences of more positive emotions (e.g., content, happy), suggesting that expressions of affect occur via more subtle mechanisms and beyond partners’ conscious awareness. Future work should explore whether participants are aware of their partners’ emotions on an implicit level.

Finally, this study suggests that emotional contagion may be a useful tool for encouraging more complex thought processes and deeper discussion in virtual groups. The tension felt by members of angry affect groups was actually an efficient means of encouraging task focus. Though our current data don’t speak specifically to performance, task behaviors should be explored in greater detail using emotions of different valence and energy to determine, which emotions may encourage optimal performance without causing negative impact.

REFERENCES

1. Barsade, S. G. (2002). The ripple effect: Emotional contagion and its influence on group behavior. Administrative Science Quarterly, 47, 644 – 675.bat/.

2. Mehrabian, A. (1972). Nonverbal communication. Chicago: Aldine-Atherton.

3. Hancock, J. T., Gee, K., Ciaccio, K., & Lin, J. M. (2008). I’m sad you’re sad: Emotional contagion in CMC. Proceedings of the ACM 2008 Conference on Computer Supported Cooperative Work. San Diego, CA.

4. Walther, J. B., Loh, T., & Granka, L. (2005). Let me count the ways: The interchange of verbal and nonverbal cues in computer-mediated and face-to-face affinity. Journal of Language and Social Psychology, 24, 36-65.

5. Walther, J.B. (1992). Interpersonal effects in computer-mediated interaction: A relational perspective. Communication Research, 19, 52-90.

6. Ophir, E., Nass, C., & Wagner, A.D. (2009). Cognitive control in media multitaskers. PNAS, 106, 15583-15587.

7. Russell, J.A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39, 1161-1178.

8. Tausczik, Y.R., & Pennebaker, J.W. (in press). The psychological meaning of words: LIWC and computerized text analysis methods. Journal of Language and Social Psychology, 1-52.

9. Anderson, C.R. (1976). Coping behaviors as intervening mechanisms in the inverted-U stress-performance relationship. Journal of Applied Psychology, 61,30-34.

Emotional Contagion Presentation

Here it is. Enjoy!
Sincerely,
Team 1980 U.S. Olympic Hockey Team

Team1980.5.3 3

View more presentations from guest21933e.

Part 1: Language Use and Status

For this part of the assignment, we collected emails members of Team 1980 U.S. Olympic Hockey Team had sent to high status people, which included emails to people such as professors, employers, and the like. We also assembled a collection of 25 emails to low status people, which included our peers or anyone we oversee for projects, etc. The definition of high status was different for the undergraduate students and our graduate partner.

A semantic analysis of the words used in each of the different conditions indicated that no specific kind of word was particularly prevalent in the emails. Although some types of words were used far more frequently in one condition in another, no category exceeded a 0.46% frequency. In this particular case, a fraction of a percent of words about politics or science, for example, meant that the words would be insignificant to the overall discourse. This overall infrequency indicates that there is no specific subcategory of the Semantic Tagset that is use more often in high or low status conversation. Instead, the words used probably vary from situation to situation, not from status to status. It makes sense to think that people use the same semantics to discuss, for example, politics, whether talking about politics with a friend or a professor. Where, then, do people vary the words they use?

Interestingly, the use of preceding nouns of title, with a frequency of 0.35% (less than the percentage of the semantic words we deemed insignificant) is possibly one of the more striking results of the analysis of the parts of speech. These terms, which (in our case) included “Ms.” and “Professor” were used six times in emails to people of high status, but were not used at all in emails to low status people. The use of these nouns indicates that people take a stance of deference when communicating with people they consider to be high status.

Two other parts of speech used in the high status emails more frequently than in the low status emails were “for”, as a preposition, and the past participle of a verb (in our case, words such as “given” were used). These words seem to be words that are used for explanation or excuse - perhaps people feel that they need to explain themselves more when speaking to a person of high status.

In the emails to people of low status, one part of speech that was used more often was the third person singular subjective pronoun, such as "he" or "she". In general, these words are used with a sense of informality or familiarity. In an email to a person of high status, on the other hand, someone might be more likely to repeatedly refer to people by name or title (e.g. “Professor Hancock was..., and the professor said...” as opposed to “Jeff was..., and he said...”), which could explain the infrequency of these pronouns in the high status emails.

Overall, then, our analysis indicates that when people email people of high status, the semantics do not vary between statuses. However, the parts of speech do vary: people use words that refer to a person’s title and that give explanations and excuses for actions when they talk to a person of high status. The words used in emails to low status people are not exactly the converse of this; instead, they are simply different. In particular, the words indicate a sense of familiarity and informality. These differences make sense when we consider the social norms when communicating with someone of high verses low status.

Euphemism Assignment: Part 2 (Making Love)

For this assignment, two individuals went through the ten most recent e-mails of all five members of Team 1980 US Olympic Hockey Team, coding for both ideas and euphemisms in the text. In order to successfully code, both coders individually looked at every email, and broke them down into ideas. Once both coders were in agreement about how many ideas an e-mail contained, the two then separately rated each idea with either a “0” (meaning the idea contained no euphemism) or a “1” (meaning the idea contained at least one euphemism).

Upon completion of this exercise, both coders then constructed a “confusion matrix” that illustrated any areas of dissent (for example if one coded an idea as a “1” while the other coded the idea “0”). The “percent agreement” was then calculated by determining how many times the coders agreed (meaning both coded an idea as “0” or “1”) and contrasting that number with the total number of ideas. Accordingly, out of the 158 ideas the coders identified from the e-mail sample, the coders were in agreement 140 times, presenting a reliability of 140/158 or roughly 89%.

Once the reliability was determined, we then went back to our data to examine some examples of euphemism both coders agreed upon. One obvious example was when a group member e-mailed someone that he would “eat” the $20 he was owed, meaning the person would forgo a debt owed to him. Another member referred to “crunch time,” in the context of staying ahead of a group project rather than leaving all the work undone until right before the assignment’s due date. Yet another member referred to a co-worker being “swamped,” meaning excessively busy, while another member used the phrase “plowing through the next two days” to imply nobly working through the academic challenges she faced that week. While many more examples were found, these were some of the most readily apparent.

Research Proposal

Introduction

When working in groups it is inevitable that group members will have some effect on each other’s cognitions, attitudes and behaviors. In fact, much research in groups focuses not only on how others affect our attitudes and behaviors (Bateman, Griffin & Rubenstein, 1987; Shetzer, 1993), but also on how their presence alters our moods or emotions (Totterdell, Kellet, Teuchmann & Briner, 1998; Totterdell, 2000; Barsade, 2002; Kessler & Hollbach, 2004). Emotional contagion is a type of social influence and refers to the transfer of moods, or the sharing of emotions, between people in a group (Barsade, 2002). According to Barsade (2002), emotional contagion is the process through which “a person or group influences the emotions or behavior of another person or group through the conscious or unconscious induction of emotional states and behavioral attitudes” (p. 646). A key factor in this theory is the notion that people do not live on “emotional islands,” but rather affect those around them as their moods “ripple” outwards in what Barsade calls the ripple effect.

Many studies have demonstrated emotional contagion in face-to-face contexts, with both dyads and groups (Barsade, 2002; Sullins, 1991), but understanding emotional expression and contagion from a computer-mediated communication (CMC) context is a little studied phenomenon. Though early work suggested that emotional contagion required others to perceive emotions via nonverbal cues, rather than words (Mehrabian, 1972), recent research in CMC, which lack nonverbal cues, demonstrates that perception of emotions and emotional contagion are possible in CMC (Hancock, Gee, Ciaccio, & Lin, 2008; Walther, Granka & Loh, 2005).

Conveying and Perceiving Emotions in CMC

Mehrabian (1972) argues that ostentatious verbal displays of emotion are generally discouraged in social settings. Thus emotions must be conveyed in “less consensual and less easily recognizable forms.” (pp. vii). These forms manifest in gestures, facial expressions, body movements and other nonverbal cues. More specifically, Mehrabian suggests that the reason nonverbal cues are more important in conveying emotion is that they provide a subtle, implicit means of communicating (e.g., slumping one’s shoulders to indicate sadness), rather than explicitly sharing emotions through words (e.g., telling a friend, “I am feeling sad.”).

Early theories of CMC shared Mehrabian’s (1972) view that communication would suffer without nonverbal cues (Kiesler, Siegal, & McGuire, 1984; Sproul & Kiesler, 1986). This perspective, dubbed the Cues Filtered Out model by Culnan and Markus (1987), suggests that functions of nonverbal cues are unaddressed in CMC due to their absence. This perspective is problematic not only because it draws from theories developed to study other media, but also because more sophisticated studies of online communication suggest the opposite (Walther & Parks, 2002).

The social information processing (SIP) theory of CMC suggests that relationships between communicators in CMC can reach the same level of closeness as face-to-face (ftf) relationships, but that this takes longer due to the slower rate at which information is shared (Walther, 1992). Nonverbal cues in this case are substituted by “content, style and timing of verbal messages online” (pp. 535, Walther & Parks, 2002). One particular study demonstrated that users are able to express emotion in CMC via emoticons and other verbal expressions and that communication partners were in fact able to discriminate between emotions that their partners shared (Walther, Loh, Granka, 2005). Recent research of dyadic CMC conversations also suggests that emotional contagion can occur, regardless of the presence of nonverbal cues (Hancock, Gee, Ciaccio, & Lin, 2008).

Emotional Contagion in CMC

Hancock et al. (2008) make several crucial advances in demonstrating the potential to convey and transfer emotions in the absence of nonverbal cues. Specifically, their research sought to determine whether others’ emotions could be perceived in CMC and whether these emotions would transfer to the communication partner. Because nonverbal cues are absent in CMC, they predicted that emotions would be transferred via linguistic cues. These predictions are not necessarily in opposition to Mehrabian’s (1972) claims about communication as the linguistic cues identified by Hancock et al. (2008) may also be viewed as subtle, implicit means of conveying emotion. Hancock et al. (2008) demonstrate that without nonverbal cues, communication partners can perceive negative emotion and experience negative moods as a result of interacting with a negative emotion experiencer (i.e., a participant induced to feel negative emotion).

Though Hancock et al. (2008) demonstrate emotional contagion in dyadic interactions, will the same be true for groups? Small group researchers in many contexts have demonstrated detriments to group satisfaction, participation and cooperation that often occur as groups become larger (Kerr, 1989; Markham, Dansereau & Alutto, 1982; Pinto & Crow, 1982). It seems that larger group size should also be detrimental to emotional contagion, which should occur with lesser intensity as group size increases. This is because cognitive resources become more distributed amongst group members. A major component of the process of emotional contagion is the recognition of emotion in other individuals (Barsade, 2002). If group members’ attention is distributed among many individuals simultaneously, emotions in other individuals will be more difficult to recognize, and thus will be less contagious (Hatfield et al., 1992; 1994). Thus we pose the following research question:

R1: Will emotional contagion occur in a group CMC setting?

Linguistic cues to emotion. As previous research has demonstrated linguistic cues are important when conveying and “catching” emotions in CMC (Hancock et al., 2007; Hancock et al., 2008; Walther, Loh, & Granka, 2005). In general, linguistic cues to emotion refer to both the sounds and the displays produced in language. Though previous research demonstrates the importance of gestures and facial expressions in conveying emotion (Mehrabian, 1972; D’Mello, Craig, & Grasser, 2008), linguistic cues to emotion exist in CMC and differ between positive and negative emotions (Hancock et al., 2008; Hancock, Landrigan, & Silver, 2007; Walther, Loh, & Granka, 2005).

One experiment found differences in verbosity, punctuation, agreement, and the use of affective words between individuals experiencing positive versus negative emotions (Hancock et al., 2007). Specifically, users displaying happiness in CMC tended to type more, use more punctuation, agree more with their communication partners and use more positive affect words compared to negative emotion experiencers who did the opposite. Hancock et al. (2008) provide similar findings suggesting that individuals experiencing negative emotion use fewer words overall and use more negative affect words. Walther et al. (2005) also demonstrated linguistic cues to express affinity (i.e., liking), which included sharing overt statements of positive affection, expressions of joy, offers of encouragement, and personal information. We use these findings to pose the following hypotheses for expression of negative emotion CMC groups:

H1: Group members sharing negative emotion in CMC will produce fewer words compared to members in groups sharing no specific emotions.

H2: Group members sharing negative emotion in CMC will use less punctuation compared to members in groups sharing no specific emotions.

H3: Group members sharing negative emotion in CMC will use more negative affect words compared to members in groups sharing no specific emotions.

H4: Group members sharing negative emotion in CMC will disagree more often with group members compared to members in groups sharing no specific emotions.

Additional Cues to Emotion in CMC. Although CMC changes how people convey emotion to one another, it does not limit emotional cues to being shared through text. Instead, it shifts the portrayal of emotion to a variety of nonverbal cues that can be expressed online. Emoticons, for example, are used to convey emotion in CMC. People use emoticons most often to convey emotions with a strong positive valence, similar to how they would use facial expressions in face-to-face communication (Derks, Bos, & Von Grumbkow, 2008). Additionally, in instant-messaging situations, people reported that the amount of time it takes for someone to respond to a message can be used as an indicator of emotion, with people with positive emotions responding more quickly than people with negative emotions (Hancock et al., 2007). However, empirical data suggested that this indicator may not be significant. Based on this previous research we propose the following research questions:

RQ2: Will group members sharing negative emotion in CMC share more negative emoticons compared to members in groups sharing no specific emotions?

RQ3: Will group members sharing negative emotion in CMC respond more slowly compared to members in groups sharing no specific emotions?

Group member characteristics. Communication in CMC is often carried out by partners who have prior information about each other. Sender characteristics (e.g., gender) and sender-receiver relationships (e.g., strength of ties) aid the perception of emotional content in CMC (Byron, 2008). For example, with respect to gender, men are more likely to express negative emotions (Witmer & Katzman, 1997), and receivers are more likely to identify negative emotions when they are expressed by men (Rotter & Rotter, 1988). Also, the strength of the tie between the communicators will affect how accurately emotions are perceived in CMC. Communicating partners with little mutual history have been found to inaccurately assess the other's emotions (Walther, Anderson, & Park, 1994). Whereas when two communicators have a longer relationship, their perceptions of each other will be more accurately as their communication will tend to be more relational and sociable (Walther et al., 1994). We thus propose the final research questions:

RQ4: Will negative emotion be identified more easily in all male groups compared to all female groups?

RQ5: Will negative emotion be identified more easily in acquainted groups compared to unacquainted groups?

Proposed Method

To study the effects that CMC in groups has on emotional contagion, we will put volunteers in a situation where they have to interact through computers. We want the subjects to get to know one another before the study begins, so we have three main options: we may have them interact in a face-to-face setting, we may give them time to interact in a mediated setting before we assign them a task, or we may simply use subjects who already know one another. For our study, we need one person in the CMC setting to display negative affect. We can either induce emotion in one of the participants or we can use a confederate to outwardly display negative emotion while interacting with the study participants.

The participants will be given a task that encourages open conversation between the participants, because this kind of creative discourse will allow one person to overtly express negative emotions, and will allow for extensive dialogue to occur between the participants. After the conversation has been completed, the students will be debriefed on the study and its purpose. There will also be a control group, wherein all three participants will be volunteers and they will be given the same task.

To measure whether emotional contagion occurred in conversation, the words used in conversation will be measured by LIWC (Pennebaker, Chung, Ireland, Gonzales, & Booth, 2007), a computer program used to determine the valence and strength of emotions expressed by the volunteers as the conversation progressed. We also plan to use the dialogue produced by participants to conduct discourse analyses that will allow us to determine differences between experimental and control conditions in emotions shared, perceived and "caught."

References

Barsade, S. G. (2002). The ripple effect: Emotional contagion and its influence on group behavior. Administrative Science Quarterly, 47, 644 – 675.

Bateman, T. S., Griffin, R. W. & Rubinstein, D. (1987). Social information processing and group-induced shifts in responses to task design. Group and Organization Studies, 12: 88–108.

Byron, K. (2008). Carrying too heavy a load? The communication and miscommunication of emotion by email. Academy of Management Review, 33, 309-327.

Culnan, M. J., & Markus, M. L. (1987). Information technologies. In F.M. Jablin, L.L. Putnam, K.H. Roberts, & L.W. Porter (Eds.), Handbook of organizational communication: An interdisciplinary perspective (pp. 420-443). Newbury Park, CA: Sage.

D’Mello, S. K., Craig, S. D., & Graesser, A. C. (2009). Multimedia assessment of affective experience and expression during deep learning. International Journal of Learning Technology, 4, 165-187.

Hancock, J. T., Gee, K., Ciaccio, K., & Lin, J. M. (2008). I’m sad you’re sad: Emotional contagion in CMC. Proceedings of the ACM 2008 conference on Computer Supported Cooperative Work. San Diego, CA.

Hancock, J. T., Landrigan, C., & Silver, C. (2007). Expressing emotion in text-based communication. Proceedings of CHI 2007. San Jose, CA.

Hatfield, E., Cacioppo, J. & Rapson. R. L. (1992). Primitive emotional contagion. In M.S. Clark (ed.), Review of Personality and Social Psychology: Emotion and Social Behavior, 14: 151–177. Newbury Park, CA: Sage.

Hatfield, E., Cacioppo, J. & Rapson. R. L. (1994). Emotional Contagion. New York: Cambridge University Press.

Kessler, T. & Hollbach, S. (2005). Group-based emotions as determinants of ingroup identification. Journal of Experimental Social Psychology, 41, 677 – 685.

Kiesler, S., Siegel, J., & McGuire, T. W. (1984). Social psychological aspects of computer-mediated communication. American Psychologist, 39, 58-60.

Mehrabian, A. (1972). Nonverbal communication. Chicago: Aldine-Atherton.

Pennebaker, J. W., Chung, C. K., Ireland, M., Gonzales, A. & Booth, R. J. (2007). The development and psychometric properties of LIWC 2007. http://www.liwc.net.

Rotter, N. G., & Rotter, G. S. (1988). Sex differences in the encoding and decoding of negative facial emotions. Journal of Nonverbal Behavior, 12, 139-148.

Shetzer, L. (1993). A social information processing model of employee participation. Organization Science, 4: 252–268.

Sproul, L., & Kiesler, S. (1986). Reducing social context cues: Electronic mail in organizational communication. Management Science, 32, 1492-1512.

Sullins, E. S. (1989). Perceptual salience as a function of nonverbal expressiveness. Personality and Social Psychology Bulletin, 15: 584–595.

Totterdell, P. (2000). Catching moods and hitting runs: Mood linkage and subjective performance in professional sport teams. Journal of Applied Psychology, 85: 848-859.

Totterdell, P., Kellet, S., Teuchmann, K. & Briner, R. B. (1998). Evidence of mood linkage in work groups. Journal of Personality and Social Psychology, 74: 1504–1515.

Walther, J.B. (1992). Interpersonal effects in computer-mediated interaction: A relational perspective. Communication Research, 19, 52-90.

Walther, J. B., Anderson, J. F., & Park, D. W. (1994). Interpersonal effects in computer-mediated interaction: A meta-analysis of social and antisocial communication. Communication Research, 21, 460-487.

Walther, J. B., Loh, T., & Granka, L. (2005). Let me count the ways: The interchange of verbal and nonverbal cues in computer-mediated and face-to-face affinity. Journal of Language and Social Psychology, 24, 36-65.

Walther, J. B. & Parks, M. R. (2002). Cues filtered out, cues filtered in: Computer-mediated communication and relationships. In I.M. Knapp, & J.A. Daly (Eds.) Handbook of Interpersonal Communication (3rd. ed., p.529-563). Thousand Oaks, CA: Sage.

Witmer, D. F., & Katzman, S. L. (1997). On-line smiles: Does gender make a difference in the use of graphic accents. Journal of Computer-Mediated Communication, 2, http://jcmc.indiana.edu/vol2/issue4/witmer1.html.

Emotional Contagion Proposal

As a group, we are interested in studying whether emotional contagion takes place when groups participate in computer-mediated communication. Emotional contagion refers to the ability to transfer moods between individuals in a group. Specifically, we are looking to study whether one person being sad can make other people sad when they communicate in groups online. Research has shown that emotions can be transferred when people are together in groups (Barsade, 2002). Additionally, there is evidence that emotional contagion also occurs between two people when they communicate in a computer-mediated setting (Hancock, Gee & Lin, 2008). Therefore, we want to determine whether emotional contagion takes place when both of these conditions (i.e. groups of people and computer setting) are applied. Thus we propose the following research question:

RQ1: Will emotional contagion occur when groups participate in computer-mediated communication?

To study the effects that computer-mediated communication in groups has on emotional contagion, we will put volunteers in a situation where they have to interact through computers. So that these subjects can get to know one another before the study begins, we will have them introduce themselves over the computer. Two student volunteers will participate, and one confederate will also engage in the computer-mediated communication. The confederate will display negative emotions in an online chat setting. The participants, two volunteers and one confederate, will be given the task of coming up with a list of twenty things to do at Cornell, because this kind of creative conversation will allow the confederate to express negative emotions, and will allow for extensive dialogue to occur between the participants. After the conversation has been completed, the students will be debriefed on the study and its purpose. There will also be a control group, wherein all three participants will be volunteers and they will be given the same task.

To measure whether emotional contagion occurred in conversation, the words used in conversation will be measured by a computer to determine the valence and strength of the emotions expressed by the volunteers as the conversation progressed. We also plan to use the dialogue produced by participants to conduct discourse analyses that will allow us to determine differences between experimental and control conditions in emotions shared, perceived and "caught".

Currently, there are quite a few issues that we need to resolve before finalizing our proposal. First, we are still unsure about what linguistic differences we should focus on specifically in studying emotional contagion in computer-mediated groups. Second, we need to find an appropriate setting in which the participants can interact online. One suggestion is that we use Google Wave, although as of right now we don’t exactly know how that would work. Third, to measure the results of the study, we will also need a computer program that will help us to code the words used by the participants and what they indicate about the participants’ emotions. As of right now, we are also not sure what program we can use for linguistic analysis or how to use it. Fourth, we need to determine whether inducing emotion in a participant, or using an experimental confederate to share emotions would be most appropriate for this study. Finally, there is also the possibility of comparing the effects of positive versus negative emotions in these settings to determine if one set of emotions is more easily "caught" than the other.

References

Barsade, S. G. (2002). The ripple effect: Emotional contagion and its influence on group behavior. Administrative Science Quarterly, 47, 644 – 675.

Hancock, J.T., Gee, K., Ciaccio, K., & Lin, J.M. (2008). I’m sad you’re sad: Emotional contagion in CMC. Proceedings of the ACM 2008 Conference on Computer Supported Cooperative Work. San Diego, CA, USA.

Assignment Three

Hello Professor Hancock and 4500 students! Welcome to the blog of Team 1980 U.S. Olympic Hockey Team. In class last week, our team took an important pledge. As we held our right hand on Clark’s Using Language and our left hand in the air, each of us solemnly swore to “ pull my own weight, to work hard for the benefit of the group, to contact any and all group members should I be unable to complete a task given to me, and to always go for the gold (defined as working to the best of our abilities whenever our abilities allow).”

While meeting to discuss the readings, multiple members of our group expressed interest in the politeness section at the end of chapter 12. In a face-to-face conversation prompting a joint commitment to a task, Clark found that (1) the less threatening the joint task person A proposes, the more polite person A is judged to be by person B; if the task is accepted (2) the less threatening the response is to person B’s self-worth, the more polite person A is judged by person B; and finally if the task is declined, (3) the more legitimately person B accounts for declining the task, the more polite person B is judged by person A. In other words, Clark declares that the politeness of an inquiry (using words like “please”) will indicate how polite the response will be (using words like “certainly I can help” or “I’m sorry I can’t). Our team thought it might be interesting to research whether or not Clark’s hypotheses of politeness holds equally true in a computer mediated setting where providing information with the greatest concision and economy is placed at a premium. Based on Susan Herring’s article on Computer-Mediated Discourse Analysis, we understood that a good research question is both open-ended and motivated by a hypothesis. Therefore, we present the following experimental design based on the idea that politeness in the CMC setting will be based on speed and accuracy of response rather than a reflection of self worth.

To test the hypothesis we have designed an experiment in which a subject will be asked to find out the time, place, and location of a concert at Cornell University. The events information will not be posted online, and the subject will be told that the only way to receive the information is by texting the event coordinator whose phone number is provided. The events coordinator will, of course, be a confederate, who is either told to answer the subjects questions with responses that project an equal equity (“yes of course I can provide you with the time, it’s at 6:00”), or alternatively, with responses that are short and simply answer the question with no additional fluff (“it’s at 6:00”). Additionally, the confederate will either answer quickly (within 60 seconds) or delayed (~3 minutes).

After the subject has engaged in text-conversation with the confederate, they will be given a survey asking a number of questions, with the goal in mind of analyzing how polite they deemed to confederate to be throughout the conversation. Based on our hypothesis, we assume that the fast response condition will be rated more polite than the delayed response condition, even though none of the polite jargon that Clark describes had been used.

While this experimental design is rather simple, we needed to make sure the data we collected was “not trivial” as Herring dictates. Therefore, looking at our design in a macro sense, this experiment will help us understand how individuals gauge politeness online; a venue where more and more important business transactions are taking place. By understanding how people evaluate politeness online (for example valuing quick responses over more elaborate, but delayed responses) we can understand how to best economize our language use online for the most efficient and mutually agreeable computer mediated experience.