Exploring social presence within an online course using Twitter

COI and social presence

One of the popular frameworks for designing effective online learning is the COI model that operationalizes critical and reflective discourse as an essential component of meaningful online learning. COI emphasizes that knowledge acquisition is not necessarily done through a fixed reality, rather it is “shaped by purposeful, open and disciplined critical discourse and reflection” (Garrison and Vaughan, 2008: 14). Further, “ideal educational transaction is a collaborative constructivist process that has inquiry at its core” (Garrison and Vaughan, 2008: 14). COI was selected for this research because it meaningfully incorporates social presence as a focus of the conceptual model, which is the emphasis of the present study.

The three core elements in the framework are described as social presence, cognitive presence, and teacher presence as shown in Figure 1. All of these elements are necessary for the COI to be effective and these elements are interdependent and support each other. While cognitive presence emphasizes student interaction with the content and teacher presence denotes the extent of student interactions with the instructor, social presence, examined in this article, is described as the ability to cast one’s self and establish personal and purposeful relationships with the community (Garrison, 2007). Social presence relies on three essential components: open communication, group cohesion, and personal connection (Garrison and Vaughan, 2008). Students need to feel safe and free to openly communicate with other students in the course. They also must feel as if they belong to a group that encourages them to think critically and share a common purpose. With these elements in place, they will develop trust and emotional bonding to others in the group. The term social presence, originally defined by Short et al. (1976), is the “salience of the other in a mediated communication and the consequent salience of their interpersonal interactions” (Rourke et al., 1999: 4).

OPEN IN VIEWER

Figure 1.

COI model.

Mehrabian (1968) noted that interpersonal actions, including facial expressions, posturing, and eye contact, make up the part of the nonverbal communication process that enhances closeness within a community. While these elements are lacking in a computer-mediated text-based discussion, they can nonetheless be achieved through rich messages, such as expressions of feeling, use of humor, personal greetings, and thoughtful self-disclosures (Hara, Bonk, and Angeli, 2000). Rourke et al. (1999) found that these and other research results indicate that computer-mediated communication can support both the cognitive and affective dimensions of higher education. In order to establish social presence in class and create a robust COI as defined by Garrison et al. (1999), it is important to establish a climate that allows for open discussion. Clear guidelines for all discussions need to be provided, reviewed, and posted. The teacher must maintain his or her presence to facilitate the conversation and nurture cohesion, which is what sustains a COI (Garrison et al., 1999).

Micro-blogging and Twitter

Social media use continues to increase among the general population, with 74% of online adults in the U.S. now reporting using a social media tool and 52% using more than two platforms, according to a 2015 Pew Research report (Duggan et al., 2015). As of 2016, Twitter was used by 24% of online adults, the majority of whom are college educated (Greenwood et al., 2016). A 2015 Pew report found 33% of all teens are on Twitter (Lenhart et al., 2015) and this demographic is soon to be college aged. Twitter is one of the world’s most popular micro-blogging sites, with an average of 313 million active monthly users, and 82% of those users accessing the network in a mobile environment (Twitter, Inc., 2016). While not fully adopted by millennials or other generations (use is less than 50% overall), Twitter does offer an online social networking platform that can potentially connect and engage users.

Micro-blogging is limited in the amount of information shared for any given post, which in turn requires that the content generator be more selective with content (Thoms and Eryilmaz, 2015). Instead of inhibiting the reader’s understanding of these limited communiqués, sentiment analysis research by Bermingham and Smeaton (2010) determined that this shortened form of writing can make it easier to understand the writer’s opinion and outcomes. Ebner et al. (2010) state that the exchange of information within 140 characters requires competency and the ability to focus and express oneself clearly. Twitter allows registered users to read and post tweets, which are short messages, limited to 140 characters, or about 25 words, as well as post videos, emojis, memes, and photographs. Tweets, which are public by default, are labeled with a user-generated hashtag and are archived and searchable. This allows users to follow conversations related to their interest and then add comments of their own. Users can easily retweet, or quote, someone else’s comment and extend the thought with an additional comment, or they can reply directly to the comment with their own thoughts on the subject.

Research of the use of Twitter in formal education

A number of studies have explored incorporating micro-blogging into formal educational settings (e.g. Carpenter and Krutka, 2014; Ebner et al., 2010; Junco et al., 2011, 2013; Lin et al., 2013; Lomicka and Lord, 2012; Lord and Lomicka, 2014; Thoms and Eryilmaz, 2015). In fact, some scholarship has even reviewed studies on the use of Twitter over several years in language learning (Hattem and Lomicka, 2016). Thoms and Eryilmaz (2015) examined enhancing the traditional LMS by embedding Twitter into the LMS platform to see if it would further engage students in meaningful learning. Their study measured students’ perceptions and found that the majority of students agreed that the use of Twitter increased their learning, social interaction, and sense of community in the course. In a study of graduate students asked to micro-blog as part of a class, Ebner et al. (2010) showed that their average daily posts increased through the time of the course thereby creating a constant flow of information. Ebner et al.’s (2010) research analyzed the content of the tweets and found evidence of a high number of activities with great potential for process-oriented learning. The study also found a high percentage of private messaging among the students and concluded that micro-blogging supports informal and formal process-oriented learning.

Twitter has also been shown to enhance social presence in addition to establishing informal, just-in-time communication between student and faculty (Dunlap and Lowenthal, 2009). Results of this study showed that social interactions occurred more naturally and immediately, and with a “persistent presence” (Dunlap and Lowenthal, 2009: 132) and that engaging in Twitter discussions unexpectedly helped them “attend to the other two components of the COI framework – cognitive and teaching presence” (p. 133). Lord and Lomicka (2014) conducted a study with preservice teachers in a graduate seminar on language teaching methodology using Twitter. Using discourse analysis and survey data, their findings showed that the micro-blogging tasks enabled participants to form a virtual Community of Practice in which they were able to learn, share, and reflect; in essence, enhancing the social presence of the course (Lord and Lomicka, 2014).

Evans (2014) study found that increased Twitter usage improved the students’ feeling of curricular engagement; however, the study failed to show whether increasing Twitter usage engages students, or if more engaged students tend to increase Twitter usage. It may be that increasing Twitter usage causes an increase in engagement, or that students who are more engaged will make greater use of Twitter. Other studies have examined students’ perception of the use of Twitter and found that the majority of them felt more connected, involved in the class, and they had positive reactions to using it as tool (Domizi, 2013; Rinaldo et al., 2011). While Munoz et al. (2014) study contradicts these findings, it did find the content of the online students’ Twitter conversation to be more friendly and engaging than students using a traditional LMS embedded micro-blog.

Online learning context

This study drew data from one asynchronous online undergraduate course offered at a major public research university in the southeast United States. The course took place over a 12-week summer semester and had been set up in Canvas, the LMS of choice at the university. The course was in journalism and looked at historical and current political trends and the challenges facing today’s media, especially with the advent of digital platforms. The course was divided into eight topic-specific modules. In addition to weekly readings, writings, and lectures, students needed to respond to a weekly discussion prompt initiated by the instructor in Twitter. Each of the eight modules’ discussion prompts were independent of each other and once posted the students had seven days to respond. Students were required to participate in the Twitter discussion and interact with their peers in the digital environment for part of their course grade. At the end of the week, the next prompt was posted and a new discussion would ensue. Students could start different threads, but all the conversations were held together with a common hashtag, which was the course number. Example of a prompt is provided below: “What steps should the media be taking to make sure market pressures don’t get in the way of us fulfilling our vital role of fostering democracy? #MMC3614”.

Students were required to follow the conversation and post at least twice during the week. Weekly Twitter discussions accounted for 25% of the students’ total course grade. In addition to the weekly discussions, each student had to choose a current political news story from an online news source and then take one point of interest of the article and create a prompt to be used as an additional discussion topic that would then be tweeted to the class. That student was then required to act as discussion leader for their prompt that week. Garrison and Anderson (2003) offered that having students moderate the online forum may lead to freer discussions. The instructor required that each student has a personal Twitter account and that they follow him/her on that account. The students were not required to follow each other but they were required to use the course-specific hashtag with all of their posts. This would keep the conversation threaded and enable students to track when someone had added a post.

Participants

Twenty-five students enrolled in the course participated in the study; 76% of the students were female. All of the students were undergraduates, the majority were seniors, and all were under the age of 25. Of the 25 students, their majors included Journalism (13), Telecommunication (6), Public Relations (3), International Studies (2), and Advertising (1). The students were geographically dispersed throughout the U.S. and two were based internationally. All students had taken at least one online course before, and 15 of them reporting having taken five or more.

All but one of the students had made use of online discussion boards before in their online classes and their overall impressions of the forums were mixed. Survey data revealed that 44% agreed that they enjoyed using discussion boards (M = 2.96). Almost half (48%) agreed that the online discussion boards helped with understanding the course subject (M = 3.16) and more than half (60%) agreed that discussion boards were an important part of the class (M = 3.32). Social relationships area also established, as 56% responded that they agreed that they helped in getting to know their classmates (M = 3.20). Data revealed that 100% of the students already had personal Twitter accounts and 68% of them had previously used Twitter within the realm of a classroom. Activity on the already established Twitter accounts was varied from having less than 100 tweets to some having more than 500.

Measures

Social presence density

At the end of the course, a content analysis of all of the tweets was performed to assess social presence density (see Table 1). Rourke et al. (1999) measure was employed. The tool measures three broad categories that contribute to social presence: interactive, affective, and cohesive responses. The first category, interactive responses, reflects the open communication described in the COI. It refers to indicators of threaded exchanges combined with messages of a socially supportive nature. Interactive responses can be identified by the presence of six indicators as shown in Table 1. The second category used to code the tweets was affective responses, which correlate to emotional presence. “Affect is expressed in computer conferencing in a number of ways, including use of emoticons (Falman, 1981), humor and self-disclosure” (Rourke et al., 1999: 6). The third category that a tweet could be designated into was a cohesive response, which includes three indicators.

OPEN IN VIEWER

Table 1.

Social presence density instrument example (per Rourke et al., 1999).

Category	No.	Indicators	Definition	Rourke et al.’s example	Examples from current study
Affective	1	Expressions of Emotion	Conventional expressions of emotion, or unconventional expressions of emotion, includes repetitious punctuation, conspicuous capitalization, emoticons	“I just can’t stand it when..!!” “ANYBODY OUT THERE??”	Emojis, or Hashtags, such as #Beachtime, #SummerFun
	2	Use of humor	Teasing, cajoling, irony, understatements, sarcasm	The banana crop in Edmonton is looking good this year	Hashtag with Humor #ToupeTrump
	3	Self-disclosure	Presents details of life outside of class or expresses vulnerability	“Where I work, this is what we do..”, “I just don’t understand the question.”
Interactive	4	Continuing a thread	Using reply feature of a software, rather than starting new thread	Software dependent, e.g. “Subject:RE” or “Branch from”
	5	Quoting from others’ messages	Using software features to quote others’ entire message or cutting and pasting selections of others’ messages	Software dependent, e.g. “Martha rites:” or text prefaced by less-than symbols	Retweet or Quote Retweet
	6	Referring explicitly to others’ messages	Direct reference to contents of others’ posts	“In your message, you talked about Moore’s distinction.”	Referencing or continuing the subject of the others’ post
	7	Asking questions	Student asks questions of other students or moderator	“Anyone else had experience with WEBCT?”
	8	Complimenting, expressing appreciation	Complimenting others or contents of others’ messages	“I really like your interpretation of the reading”
	9	Expressing agreement	Expressing agreement with others or content of others’ messages	“I was thinking the same thing.” “You hit the nail on the head”
Cohesive	10	Vocatives	Addressing or referring to participants by name	“I think John made a good point.” John, what do you think?”	Using the @TwitterHandle feature, @smithb; @natshaw
	11	Addresses or refers to the group using inclusive pronouns	Address the group as we, us, or our	“Our textbook refer to…” “I think we are offtrack”
	12	Phatic, salutations	Communication that serves a purely social function; greetings, closures	“Hi all”, “That’s it for now” “Beautiful weather here today”	Can also be hastags, #ThatsAllFolks

Coding and inter-rater reliability

The researchers used the online software tools Tweet Archivist and TwDocs to download the students’ Twitter™ conversations. Transcripts from the class’s Twitter conversations were subjected to a content analysis using the social presence density tool (Rourke et al., 1999). Two researchers used a combination of the thematic unit and syntactical unit to code for the presence of the 12 indicators within the tweets. It is not always appropriate to identify syntactical units such as grammatically correct phrases, clauses, and complete sentences. Thematic units, such as single thoughts or ideas that can be interpreted from a conversation should be included in the measurement, as online conversations “combined the telegraphic style of email with the informality of oral conversation” (Rourke et al., 1999: 9). By combining the flexibility of the thematic unit, with the ruled-based structure of a syntactical unit, coding consistency increased.

The simplest method to compute inter-rater reliability, and the method used by Rourke et al. (1999), is the percent agreement statistic, or the Holsti method. Before coding, the two researchers separately analyzed the first 100 tweets from the course: coding for the presence of the 12 indicators. The researchers then came together in a one-on-one discussion and compared how they categorized the 100 tweets. After the discussion, the researchers separately analyzed tweets number 101–200 and came together again to compare results. This time, sufficient agreement was reached, and aggregate inter-rater reliability was .989, so the researcher then independently coded the remaining 896 tweets (see Table 2).

OPEN IN VIEWER

Table 2.

Overall social presence density with inter-rater reliability.

Category	Indicator	Inter-rater reliability	Examples from current study	Total instances
Affective	Expresses Emotion	1	@TWITTERHANDLE! I love this example, I almost used it myself!  It scares me how my social media feed basically regurgitates the information I want to see right back to me	43
	Humor	1	If your mother says she loves you, check it out. I was hoping someone would post this lol	8
	Self-disclosure	0.98	I think I fit in the Issue public theory. I tend to read mostly about issues I care about. I am trying to get better at that	99
Interactive	Continues thread	0.98	RT @TWITTER HANDLE: “If your mother says she loves you, check your sources.” Good advice for good journalism.	1
	Quoting	0.99	RT @TWITTER HANDLE: Rick Perry’s ad does a good job relating to the “American Dream” and veterans.	8
	Refer to message	0.96	@TWITTER HANDLE “Rock the boat!” I like that lol. Esp. when opposing views are hard to find, it’s important to keep asking questions #mmc3614	496
	Asking?	0.99	@TWITTER HANDLE: Donald Trump has been very vocal through his campaign. Do you think this will go over well with voters?	92
	Complimenting	0.99	You make a good point that those who see social media (etc.) Great little @RandPaul plug @tyschildts	60
	Agreement	0.99	I agree, voters will be more attuned to the issues that directly effect them	124
Cohesive	Vocative	1	He is the poster child for, “No publicity is bad publicity,” @TWITTER HANDLE	550
	Group pronouns	0.99	Just how everyone else has mentioned, priming at its core	17
	Phatics	1	#Go's	1

Data analysis

The study used content analysis to code the social presence density in the students’ Twitter™ conversations. We used descriptive analysis of the background survey and COI survey by interpreting means, standard deviations, and response frequencies. We also performed internal consistency reliability (Cronbach’s alpha) and correlation analysis (Pearson correlations) to identify the relationships between the social presence density scores and the COI survey results. An alpha of .05 was used for all statistical tests.

Content analysis of tweets using social density tool

While inter-rater reliability for each indicator varied, this variation was within acceptable ranges (>.96). Manifest indicators such as “Use of Vocatives” had a reliability of 1.0; other indicators, with presumably manifest characteristics, such as “Asking a Question” had a .99. Indicator “Uses Humor” had an inter-rater reliability of 1.0 as shown in Table 2. Rourke et al. (1999) designed the social presence density measurement tool to equally weight each of the 12 indicators. In this study the most prevalent social presence indicator was “Vocative” or the use of personal pronouns in a post. There were 550 incidences of vocatives, which were 50% of all tweets. Examples of this include “@TwitterHandle, journalists need to come together to raise the standard of information they discuss,” and “He is the poster child for “No publicity is bad publicity”, @TwitterHandle.”

The second most prevalent social presence indicator was “Refer to message,” with 496 occurrences, which was 45.26% of all tweets. For example “I hear you with exhausting, but it’s worth it” and “@TwitterHandle. Well said. Journalists do a public disservice when commentary and $get in the way of sharing news.” It is notable that the indicators “Use of Vocatives” and “Refers to Message” usually appeared together. There were 113 tweets that showed no social presence indicator. There were 269 tweets (25% of total tweets) representing two indicators and 228 tweets (21% of total tweets) reflecting three indicators, and 134 tweets with more than three indicators. There was one tweet with six indicators, which was the highest recorded.

The final calculated aggregate social presence density for all 12 indicators was 11.40%. The indicators fall into three categories, and the aggregate density for each category was as follows: Affective: 4.56%; Interactive: 11.88%; Cohesive: 17.27%. Figure 2 illustrates the exact density score of each indicator. The lowest incidences were “Phatics” and “Continues Thread,” both with only one occurrence. The indicator “Continues Thread” was coded when retweets were used within the class. If a student’s comment was retweeted by someone else outside of their class, it was not counted as part of the class’s social presence.

OPEN IN VIEWER

Figure 2.

Social presence density per indicator.

COI survey results and relationships

The COI survey developed by Arbaugh et al. (2008) was administered through an online survey on the last week of the class and yielded 13 responses, which is a 52% response rate. Extra credit was given to those taking the survey. Appendix 1 includes all of the descriptive statistics for the individual items. Mean responses for teaching presence were 4.37 (SD = 0.94), social presence was 4.42 (SD = 0.77), and cognitive presence was 4.63 (SD = 0.53).

Two questions had the lowest means. One was within the Teaching Presence construct, Item 10: “Instructor actions reinforced the development of a sense of community among course participants” was at M = 3.69. The other response with M = 3.69 was Item 16 within the Social Presence domain: “Online or web-based communication is an excellent medium for social interaction.” While these were the lowest means, it is notable that both are above the central point. The highest mean score was M = 4.85 in Teaching Presence (Item 4: “The instructor clearly communicated important due dates/time frames for learning activities”). Per our guiding research question, we performed Pearson correlations using the results of the COI survey and the findings from the social presence density content analysis. Fourteen of the correlations were statistically significant and ranged from r = +.985 to r = −.630, p < .05 (see Table 3).

OPEN IN VIEWER

Table 3.

Pearson correlation matrix between social presence density scores and COI survey scores.

r	Teaching	Social	Cognitive	Expresses Emotion	Humor	Self-disclosure	Continues Thread	Quoting	Refer to message	Asking question	Compliment	Agreement	Vocative	Group Pronouns	Phatics	Total tweets
Teaching average	1	0.431	0.417	−0.144	−0.438	−0.458	0.17	.a	−0.143	−.630*	−0.4	0.134	−0.096	−0.453	−0.199	−0.193
Social average		1	.834**	−0.153	−0.467	−0.244	0.288	.a	0.035	−0.072	−0.134	−0.003	0.008	−0.232	−0.153	0.021
Cognitive average			1	−0.332	−.619*	−0.462	0.161	.a	−0.037	0.096	−0.363	−0.078	−0.063	−0.141	−0.231	0.023
Expresses emotion				1	.728**	0.47	−0.079	.a	0.004	−0.041	0.541	0.016	0.071	.588*	−0.079	−0.076
Humor					1	0.303	−0.108	.a	0.006	−0.088	0.364	0.036	−0.007	.656*	0.243	−0.115
Self-disclosure						1	−0.075	.a	0.549	0.185	.885**	0.215	.591*	0.318	0.102	.608*
Continues thread							1	.a	0.38	−0.29	−0.135	0.259	0.382	−0.198	−0.083	0.376
Quoting								.a	.a	.a	.a	.a	.a	.a	.a	.a
Refer to message								.a	1	−0.093	.554*	.555*	.985**	0.004	0.091	.928**
Asking?										1	0.21	−0.436	−0.076	0.253	−0.068	0.084
Compliment											1	0.037	.591*	0.398	−0.135	0.537
Agreement												1	0.538	−0.098	0.127	0.451
Vocative													1	0.021	0.013	.931**
Group pronouns														1	−0.198	0.103
Phatics															1	0.033
Total tweets																1

^*Correlation is significant at the 0.05 level (two-tailed).**Correlation is significant at the 0.01 level (two-tailed).^aCannot be computed because at least one of the variables is constant.

Using the guide for strength of effect size as suggested by Evans (1996), a significant and very strong positive relationship (i.e. r = .80–1.0) was found between the indicators “uses vocatives” and “referring to others’ message” (r =.985, p < .01), indicating that students were inclined to call a person by their name when they were referring to something that person said. This shows a relationship between the interactive and cohesive categories of social presence density tool. A significant and strong positive relationship between the indicators “self-disclosure” and “complimenting” (r =.885, p < .01) indicated that students would likely reveal something about themselves when they complimented another student on their tweet, showing a relationship between affective and interactive social presence density categories.

Those students that had a high number of total tweets tended to be the same ones that referred to others’ messages and called others by their name, and there was a significant and strong positive relationship between “number of tweets” and “refers to others’ message” (r=.928, p < .01), as well as “number of tweets” and “uses vocatives” (r = .931, p < 01).

Significant and strong positive relationships were found between the reported COI categories of Cognitive Presence and Social Presence (r = .834, p < .01); however, there was no statistically significant relationship between Teaching Presence and Cognitive Presence (r = .417, p=.156) or Social Presence (r = .431, p = .141). Significant and strong positive relationships (r = .60–.79) were found between the indicators “uses humor” and “expresses emotions” (r =.728, p <.005), both of which are indicators in the affective category of social presence density tool. Students who had a high total number of tweets also had a tendency to participate in self-disclosure and a significant and strong positive relationship was detected (r = .608, p = .027). A significant and strong positive relationship was also found between students who used group pronouns and used humor (r = .656, p = .015).

Significant and moderate relationships (r = .40–.59) were found between “both refers to message” and “complimenting” (r = .554, p = .049) and “refers to message” and “agreement” (r = .555, p = .049). There was a significant and negative relationship between the COI construct “teaching presence” and “asking questions” (r = −.630, p = .021) and a significant and negative relationship between the construct Cognitive Presence and “humor” (r = .619, p = .024).

Recommendations for practice

Tools shape our practice. Today’s students will be using social networking tools when they enter the workplace, whether they are in marketing, engineering, business, education, medicine, or even sports. As educators and researchers, we should try to incorporate social media and social networking tools in to our teaching practice so students have an opportunity to master the skills in their educational experience. By using a micro-blogging tool, such as Twitter, online instructors can engage these always-connected students and establish student-to-student social presence and community. In turn, this can contribute to positive learning outcomes.

Instructors need to be actively involved in the online discussion forum, whichever platform they decide to use for implementation. Twitter provides a simple, just-in-time, persistent presence for instructors to connect with students. Instructors can implement it in teaching in other ways as well, such as imparting important class information (e.g. “Read Chapter 2 tonight, not Chapter 3”), asking questions (e.g. scheduling appointments), and research (e.g. pose questions to Twitter-verse to find answers).

Instructors should leverage opportunities to connect with these always-connected and mobile students. In this study, almost half of the students accessed their Twitter account from mobile devices when participating in the class discussions. Smartphones were the main device for mobile, but iPads were also used. This finding indicates that if the online discussion is easily accessible on a mobile device, students will take advantage of the opportunity, and possibly participate more frequently, further enhancing the idea of Rourke et al. (1999) that computer-mediated discussions need to allow for high levels of freedom of time and place to engage in interactivity.

While students in this research seemed comfortable tweeting to each other (1096 tweets over the 12-week course), further research should take a look at learner outcomes, both perceived and real. Some research has examined student engagement and learning outcomes using Twitter as an intervention in higher education settings (Junco et al., 2011, 2013). However, the importance of collaboration and community in online classes has been established. Tying the use of these just-in-time discussions to learning outcomes would further sanction the use of micro-blogs in the online class. More research is necessary to fully comprehend the use of Twitter in formal educational settings, particularly in online learning spaces.