Exploring the Dimensions of Affinity Space, Public Sphere, and Opinion Leadership through Educational Twitter

Exploring the Dimensions of Affinity Space, Public Sphere, and Opinion Leadership through Educational Twitter
Abstract
This study explores the dynamics of affinity space, public sphere, and opinion leadership in the context of education-related conversation on Twitter. Twitter tweet and user data was collected between the months of August 2020 and March 2021 by querying five hashtags related to education. A content analysis of tweets was conducted to define the prevalence of affinity space or public sphere on Twitter. Topic modeling was done to identify possible or latent topics discussed each month. A social network analysis of users was run to identify opinion leaders according to in-degree centrality, followed by a user role content analysis to identify the offline social status of leaders. The preliminary results of this study provide potential contributions to the fields of both education and communication.
Introduction 
The start of the pandemic in March 2020 has significantly impacted thousands of educational institutions across the United States that transitioned to offering completely remote online courses and programs. Anecdotally, it has not been easy for students and educators to transition to an all-digital setting. While the development of communication technologies and online social networking sites has allowed for the expansion of online learning opportunities, the radical shift to online education has also yielded new challenges. Both opportunities and challenges can cause educators to turn to social media for sharing and seeking solutions. In this study, we seek to characterize the space that educators have created for these discussions online. In doing so, we hope to deepen our understanding of interest-based conversations and their ramifications for wider political debates. Using the theoretical framework of affinity spaces, from the field of education, and the public sphere, we aim to quantitatively assess the extent to which political and interest-based conversations are integrated with one another.
As discussed by McGee, affinity spaces are "the kinds of learning communities to be found on interest-based websites, which he presents as a model for learning that seriously challenges current school practices" (Nygard & Skaftun, 2017). This theoretical model differs from the learning environment of traditional face-to-face classes for various reasons such as the medium of communication, modes of learning, and even the stakeholders present, among others. Thus, we begin with the theoretical frameworks of affinity space and public sphere to investigate the topical arrangement of these conversations, making a connection between the two theories and approaching the broader theoretical question of whether political and non-political discussions can be fully separated, particularly on the public forum known as Twitter. Finally, we consider leaders' position in the conversations of educators on Twitter, investigating user characteristics associated with typical leadership in a given topic area based on affinity space and public sphere literature.
Theoretical Overview and Review of the Literature
Educators on Twitter
The literature on Twitter use in education describes various educational applications at different educational levels, from first graders to graduate students. Studies show that educators use Twitter for three primary reasons; (1) to improve communication, (2) to enhance classroom activities, and (3) to promote professional development (Carpenter & Krutka, 2014). Carpenter and Krutka (2014) demonstrated that Twitter could boost communication between learners and educators. Through participation in these conversations, educators become more proactive and engage in teaching their students through alternative modes of learning, which is found to increase their efforts (Chawinga, 2017). Although research on communication use of Twitter is limited, numerous practitioner-oriented articles detail its use in communication. According to Carpenter and Krutka (2014), educators can use micro-blogging to inform or update school communities on events, activities, and policy changes.
 Studies also show that the effective use of Twitter could influence the students' perception of the professor's credibility. In one study done by Johnson (2011), the authors found out that professors who use both "social and scholarly tweets" are perceived by students to be more credible than professors who only post either 'social' or 'scholarly' tweets in their profile. Perhaps this is related to the advantage found in a more recent study, which shows that professors who are posting social tweets increase their "engageability" on the part of the student compared to professors who are not (Kunka, 2020). 
Twitter can also connect educators, learners, administrators, and other educational stakeholders through a common hashtag or interaction between accounts. In widespread usage, hashtags help social media users categorize their social posts and join popular conversation topics on almost any form of discussion. Accordingly, Carpenter et al. (2020) have shown that one of the most effective ways of using hashtags is "facilitating professional connections and interactions between geographically dispersed educators with common interests and needs." Additionally, in higher education, micro-blogging can facilitate greater communication between students and instructors (Carpenter & Krutka, 2014). An instructor can use Twitter to remind students about assignment deadlines and specific activities. Twitter can also help students interact socially and professionally. Most importantly, Twitter enables students to speak directly to the educator (Ricoy & Feliz, 2016).  In this case, instructors use Twitter to encourage introverted students to participate in learning activities.
Another area where educators use Twitter is in classroom activities. Examples of classroom activities where Twitter is applicable include back-channeling, resource sharing, historical perspective-taking and reenactment, and media studies (Carpenter & Krutka, 2014).  Back-channeling through twitter refers to an action that enables learners to continue with a conversation or ask questions regarding a lecture, a movie, or any concurrent activity (Carpenter & Krutka, 2014). University students have linked Twitter use in class activities to increased course satisfaction. According to Rehm and Notten (2016), Twitter can engage students with emerging technology, strengthen their interactions with instructors, and increase access to information related to course material. Therefore, the use of Twitter in class activities fosters a community feeling and enhances students' relationships with the educators. 
Studies also show that it is essential to have educators on Twitter to facilitate learning through this social media platform (Rehm & Notten, 2016). As noted, Twitter strengthens communication between students and instructors. It generates communication and interaction, which lead to the creation and dissemination of content at tremendous speed. A study indicates that educators can use Twitter to increase the learning process. The number of tweets is associated with students' involvement in class activities (Ricoy & Feliz, 2016). It is important to note that the purpose of introducing Twitter was to improve writing, develop reflection, and expand the class community.  Therefore, having educators on Twitter improves the learning process.
Educators use Twitter as a learning tool and a teaching tool because it increases communication and enhances participation in-class activities. As noted, Twitter helps inhibit students from communicating with their educators effectively. Twitter also enables educators to share and acquire resources by tweeting links to education-related articles, wikis, blogs, and other websites (Carpenter & Krutka, 2014). In a more recent study done by Carpenter et al. (2020), of around sixteen popular hashtags that professors were using, the authors have shown that this method of sharing and acquiring resources is increased, not only of the increased traffic towards these resources but also because of the increase sharing and resharing of links, thereby increasing the chances of educators being aware of their existence. Ultimately, these methods improve the learning process as they contribute to the overall benefit of the student. Given the sharing and acquiring of resources within particular education-related hashtags, these hashtag communities have been defined as affinity spaces by education scholars. 
Affinity Spaces 
Affinity Spaces are effective learning environments where individuals with shared interests gather and engage with each other to gain expertise on a particular subject (Gee, 2004). These spaces have no "limitations" and can be either virtual, physical spaces, or hybrid. Involving digital media in an affinity space can provide a rich area for participation (King, 2010). An encouraging part of affinity spaces stems from the multimodal possibilities for engagement and communication with others in this shared space. Educators can use hashtags, directly engage with others using the @ symbol, utilize videos, podcasts, and many forms of alternative digital expression to demonstrate their expertise and facilitate learning (Curwood, 2013).
Original affinity spaces refer to those affinity spaces where digital media are not involved. Churches provide a classic example of an original affinity space (Gee, 2004). People go to churches to share their religious thoughts and values and participate in social events to connect with others, and they meet other members face-to-face at a physically existing place (Gee, 2004). Another example of an original affinity space is the high school newsroom (Gee, 2004). The high school newsroom offers participants varying degrees of participation. Whoever comes into space is considered part of the affinity space, no matter how long or how often they stay in the space (Gee & Hayes, 2012). 
According to Gee's article, digital media transformed affinity spaces into a new form that we can call digital affinity spaces. Across the literature on digital affinity spaces, the defining characteristic is that these spaces are all structured by digital media, in other words, entirely online. They can be created by online games (i.e., MMORPG), social media, or any other digital content. For instance, in The Hunger Games affinity space, the involvement of players created the affinity space for them to share their common goals and thoughts that are relevant to the game and have discussions in various other platforms such as Subreddit forums, discord, and even in social media groups that focus on the matter. 
While they can be either virtual, physical spaces, or hybrid, there are 11 features of affinity space (Gee, 2004), including the importance of common endeavor (rather than race, class, disability, gender) and that people with differing levels of knowledge all share common space. Affinity spaces do not necessarily have to incorporate all the elements to be considered successful – a few elements essentially fulfill the 'space' needed for collective intelligence. Most of the elements allow for enormous flexibility for a group of people to share resources and offer a "new analytic lens" in specific topic areas (Gee, 2004). Further, particularly with adult learners, where the roles and statuses may range more extensively than other groups (i.e., experienced principal vs. recently graduated teacher), affinity spaces allow them to share resources and knowledge through high levels of interconnectivity, flexibility (especially with leadership and status), and complexity that be otherwise difficult to achieve. Affinity spaces transform the notion of distance education as it is no longer a passive activity but requires active engagement and interaction within the space (Gee, 2004). However, due to the low barrier for entry in online affinity spaces, especially on Twitter, educators may need to filter out spam and irrelevant content (Carpenter et al., 2020). 
Many researchers argue that social media brings new opportunities for informal learning through participatory digital cultures (Greenhow & Lewin, 2016). The presence of affinity spaces on Twitter is crucial, particularly for educators, since many educators use social networking sites as valuable resources for professional development and further learning purposes (Carpenter & Krutka, 2014). Twitter can serve as an educator's professional "learning network" that extends beyond the confinements of typical geographical and physical barriers that may otherwise be limiting. Exchanging new ideas, knowledge, and experiences in these learning networks can create a heightened sense of belonging and mutual affiliation, mainly through "collective intelligence." 
We are particularly interested in the relationship between affinity spaces and the public sphere, as affinity spaces where politics may be salient, as in education, may intersect with discussions traditionally considered the purview of the public sphere.
Affinity Spaces and the Public Sphere
According to Gee (2004), the affinity space literature developed out of the concept of communities of practice. Both concepts suggest that people can learn better when they are engaged in highly motivated social practices that they value. Nevertheless, the term "community" connotes labeling a group of people, which raises vexatious issues like identifying membership. Therefore, the notion of affinity space adopts a “space” orientation instead of emphasizing the sense of “group”(Sharma & Land, 2019) like the public sphere. The public sphere also downplays specific identities of participants within the space. Membership of the public sphere can be regarded as coterminous with citizenship (Webster et al., 2004 ).
However, when we consider the conceptual background, there is a fundamental difference between affinity space and the public sphere. Affinity space was initially applied to critique traditional schooling (Gee, 2004). Compared with the notion of an affinity space, the public sphere is a classical macro concept rooted in civil society (Friedland, Hove & Rojas, 2006). The original context of this idea is the transition of public spaces from coffee houses to mass media (Habermas & Habermas, 1991). According to Habermas, a core construct of this concept is the formation of public opinion (Habermas, Lennox & Lennox, 1974). The public sphere has significant societal and political functions that can sufficiently integrate public communication (Maier, Waldherr, Miltner, Jähnichen, & Pfetsch, 2018). Therefore, researchers mainly use this theory to examine political discourse in the media, which diverges from the educational focus of studies related to affinity space.
Considering the above, the coexistence of the public sphere and affinity space on digital platforms is not contradictory. Previous studies generally acknowledge the internet or social network as a new public sphere (Papacharissi, 2002; Yang, Quan-Haase & Rannenberg, 2017), while education-related hashtags have been regarded as affinity spaces (Greenhalgh et al., 2020). It is reasonable to expect some crossover between these theoretical spaces. The idea of 'the public' is closely linked to democratic ideals that call for citizen participation in public affairs (Papacharissi, 2002). Because individuals in affinity spaces by definition share an interest and engage with each other, public issues related to these interests are likely to emerge as discussion topics alongside the education-focused discussion characterizing affinity spaces. 
In the context of our study, education can be regarded as both the subject of an affinity space as well as a significant public issue. For example, teachers engaged in specific discussions about education techniques can be highly concerned with other education-related political news and social activities on Twitter. 
Hashtag Communities on Twitter
A hashtag is a metatag tag that begins with a # symbol, frequently used to enable cross-referencing of content based on a particular topic or theme (Messina, 2007). Hashtags allow users to quickly find trending or important topics, with the advantage of also being continuously updated in real-time. 
Hashtag Communities as Affinity Spaces and/or Public Spheres
Among educators, some of the largest affinity spaces stem from discussing a single hashtag on social media. Education-related hashtag communities on Twitter fulfill multiple elements of affinity spaces, as they allow users to learn about educational topics and facilitate new conversations based on shared interests and goals. As discussed by Carpenter et al. (2020), education-related hashtags help increase traffic and bridge geographically diverse educators with similar factors or experiences. In turn, sharing experiences, knowledge, and even educational resources based on a popular category becomes an avenue for their learning. For example, in education-related hashtag communities, educators use the hashtag to engage with others using video, podcasts, replies, and user mentions to determine their experience and enhance learning (Curwood 2013). Knowledge, ideas, and involvement expertise are exchanged within the learning network.
Education-related hashtags can be general, focused on specific subject topics, or issues that are helpful for certain professional positions (e.g., assistant principals). Other researchers have investigated the general landscape of education-related hashtags as affinity spaces. Greenhalgh (2020) analyzed teachers' participation within the #michED hashtags and suggested that diverse affinity spaces exist within the same hashtag for different aims. Other hashtags, such as #edchat, cater to more general educational topics, while #apchat are affinity spaces that act as valuable resources for assistant principals (Mazza, 2012). Rosenberg et al. (2016) found that users who actively participated in weekly chats associated with the #edchat hashtag on Twitter had positive growth in their professional learning and development. Other studies (Carpenter et al., 2020; Rosenberg et al., 2016) have collected Twitter data from multiple hashtags, considering each hashtag a unique affinity space for educators.
Given the previous characterization of hashtag communities as affinity spaces, along with our proposition that affinity spaces and the public sphere are likely to overlap, we posit the following research questions:
RQ1: Can educator hashtag communities on Twitter be primarily defined as affinity spaces or a public sphere?
RQ2: What topics are being discussed in educator hashtag communities on Twitter? 
Further, with the flexibility of discussion across both affinity spaces and the public sphere, we are interested in the relationship between topics of discussion and time, leading to the following research question:
RQ3: Will the primary topic of discussion in educator hashtag communities on Twitter change over time?
Leadership and Social Media (REMOVE)
Leadership in general is the ability to accomplish a goal by directing human action and intention. To be a successful leader, one must be able to have followers who will collaborate with the leader’s guidance to achieve a desired end. A good leader stands by his or her principles and exhibits great character amidst adversity. Leadership stems from innate social belongingness and conducive relationships with others. Leaders follow through with their intentions with their followers and their character, which qualify them to lead others. Their character traits include charisma and the ability to inspire others, which contributes to the success of a movement. (Kruse, 2021).
In modern times, leadership has evolved thanks to social media. In the online setting, there is faster transaction and exchange of ideas. Leaders can use this platform in order to express effective leadership. In the traditional setting for leadership, there needs to be an organizational hierarchy between leaders and other members. In the online setting, leaders are empowered through their followers who are able to share and influence movements. In Twitter, leadership works by trends and number of followers. People perceive more power and influence from people who are liked and subscribed to by the majority. In the online setting, followers are also sources of credible information, as leaders are no longer the sole source of opinion and knowledge. (Hind, 2021) 
The dynamics between leaders and followers follow social trends, and are dependent on the movement that they are subscribed to. Most netizens engage with leaders who are outspoken and active in social media. Twitter and other social media sites have become avenues for reaching a wider audience, as it has become critical for influence and dissemination of information. 
Opinion Leadership 
Finally, we draw on the literature surrounding the concept of opinion leaders to understand the proposed structure of the public sphere and affinity spaces. The literature surrounding opinion leadership suggests that leadership is flexible and disconnected from social status or role outside of the space (Park, 2013). Opinion leadership, traditionally explored through mass media forms such as television, radio, and print, can now also be found in online communities such as Twitter, where individuals are given more agency to reach large audiences without having formal authority (Park, 2013; Park & Kaye, 2017). Initially conceptualized in the theory of the two-step flow of information, opinion leaders can be defined by different characteristics that a particular individual or group of individuals have, which allow them to hold more decisive influence over non-opinion leaders in the diffusion of information from an original, usually official, source (Rogers & Cartano, 1962; Park, 2013). With the role of mediator between source and general public, opinion leaders, as the name suggests, have an outsized influence on public opinion. More recently, research has been done on the prevalence of opinion leadership in computer-mediated e-commerce environments or social media platforms such as Twitter (Lyons & Henderson, 2005; Park, 2013; Park & Kaye, 2017). 
While opinion leadership was traditionally explored in those with higher social status and more extensive social networks, today the phenomenon is relevant in digital settings among those with perceived expertise regardless of socioeconomic status (Park, 2013). Moreover, opinion leadership is inclusive of characteristics found in different individuals depending on the social topic. In a cross-section survey study on housewives in the 1970s, Myers & Robertson (1972) concluded that opinion leadership traits might not necessarily be transferable to all subject matters. The theoretical framing of leadership in affinity space literature follows a similar vein, where shifts in discussion topics may lead to shifts in leadership. However, where leadership in the public sphere drives the formation of public opinion, leadership in affinity spaces is primarily focused on information sharing for educational purposes. Affinity space leadership in hashtag communities has not been a primary area of research for previous scholars, who have focused on laying the foundation of identifying hashtag communities as affinity spaces. Given the shared characterization of leadership in both public sphere and affinity space literature, we propose the following research questions:
RQ4: To what extent will users acting as leaders in the hashtag community on Twitter change over time?
RQ5: Will users acting as leaders change based on topic discussion in educator hashtag communities on Twitter?
RQ6: How will the offline social status of users acting as leaders in educator hashtag communities on Twitter change over time?
Justification & Contribution (MOVED TO DISCUSSION)
This study will contribute to the existing literature by making a new theoretical connection between the education theory of affinity space and the macro concept of the public sphere. While previous literature has studied the two individually, we propose that the dynamics of an affinity space create a prime opportunity for the deliberative discussions present in an ideal public sphere. In addition, we will be building on past research in the activity of educators in affinity spaces on Twitter with the addition of network analysis and new measures of leadership through this method.
Method
To address our listed research questions, we conducted a content analysis of tweets, topic modeling of tweets, network analysis, and a user content analysis. We collected all tweets containing four key hashtags created during the 2020-2021 academic year period under the impact of the pandemic (September 1-March 31) using the Brandwatch service. These hashtags include #EdChat, #Edu, #Education, and #Educhat. We also included lowercase versions of the same hashtags (#edchat, #edu, #education, and #educhat). Brandwatch is a paid service that allows users to access the full history of publicly available tweets on Twitter. The total number of tweets containing at least one of the hashtags listed is 709,903. Using the date information from the Twitter dataset, tweets were binned by month to operationally define time (Table 1).
Tweet Content Analysis
To answer and examine RQ1, we conducted a content analysis of a sample of the tweets in our dataset. We used a stratified random sample where 0.03% tweets were taken from each month, and the final sample included 2,090 tweets. Two coders coded a total of 209 tweets (10% of the sample) at a time, consensus coding any disagreements until intercoder reliability was established. For our preliminary results, half of the sample, 1,045 tweets, were coded, and ICR of 0.84 was established after the third round of 209 tweets.
Tweet Topic Modeling
To address RQ2 and RQ3, we used topic modeling to categorize the topics of discussion for each month. Topic modeling is a method that abstractly statistically summarizes the dataset by identifying patterns of word co-occurrences in documents. Through the identification of such patterns, topic modeling algorithms can highlight latent topics within the data. Further, with large datasets where manual content analysis of all tweets is infeasible, topic modeling provides a reasonable alternative to gain a general understanding of topical trends within the data. Our dataset included over 700,000 tweets, so in order to present such a massive dataset clearly, we chose to use topic modeling to examine how frequently each topic had been discussed.
We implemented the Latent Dirichlet allocation (LDA) algorithm with Gibbs sampling from the topicmodels package in R, running a unique topic model for each month. Before running topic models, we cleaned the text data for stop words, links, digits, and punctuation other than hashtags (#) and mentions (@).We identified the optimal number of topics (k) using the Cao Juan (2009), Arun (2010), and Deveaud (2014) goodness of fit statistics for selecting topic number, checking values of 5, 10, 15, 20, and 25 for each month. We then compared coherence and exclusivity measures across topics for each value of k and each month. Based on these assessments, we directed the LDA algorithm to identify 10 topics when run on each month.
Topics in LDA are defined by probability of occurrence of topics across words and documents. Because the tweets in each month may contain slightly different topic keywords, we used the top ten words most likely to occur within a given topic to derive topic labels, which we can compare across months. Two coders individually assessed and then collaboratively labeled the topics.
Social Network Analysis
To address if users acting as leaders in the hashtag community on Twitter will change over time, we conducted a social network analysis on NodeXL. Using the tweet id numbers provided by Brandwatch, we created directed networks for each month from August 2020 through March 2021. The networks were composed of Twitter users as vertices and retweets, mentions, and replies as edges. A high in-degree indicated that a user had been frequently retweeted, mentioned, or replied to by others. Therefore, in-degree can be considered to be the extent to which a user is the central focus of discussion by other users. Following the example of previous literature (Guo et al., 2020),...