What motivates twitter users to engage with libraries?

Introduction

Social media plays an important role in libraries’ communication strategies. Due to the ubiquity of social media, an active online presence is commonly positioned as an essential requirement for libraries. Currently, one of the most popular social media applications is the microblogging service Twitter. Twitter users can communicate with their followers via tweets of 280 characters or less. In addition, users can retweet, or share, another account’s tweet with followers, like a tweet to demonstrate appreciation, and connect to conversations across Twitter by highlight keywords and phrases using hashtags. In the professional literature, libraries are encouraged to use Twitter for two primary reasons: to engage their communities and to disseminate information (Vassilakaki and Garoufallou, 2015). Social media applications like Twitter provide libraries with a new place to engage directly with clients and communities. Typically, library engagement is focused on connecting patrons, and potential patrons, with library services; however, Twitter and other social media platforms offer libraries a new form of sociality by enabling and mediating "new forms of online user participation, engagement, access and interactivity" (Cavanagh, 2016, p. 247). We are arguing that these new forms of participation, engagement and interactivity form the foundation for libraries to develop and nurture a new kind of relationship with clients and communities. Traditional understandings of the kinds of relationships library have with their members rely on measures such as membership registration and library use statistics; however, Cavanagh (2015) argues these understandings overlook the active nature of public libraries’ relationships with their membership. For instance, a person who does not hold a library card but follows the library on Twitter creates a relationship with the library through their choice to follow the library’s Twitter account. This descriptive study explores the relationship public libraries have with their communities via Twitter – exploring their interactions, how these interactions lead to relationships, and how these relationships can be characterized.

The questions guiding this study are:

• What content motivates a Twitter user to engage with the library?
• What kinds of interactions are public libraries initiating with Twitter users?
• What kind of interactions are Twitter users liking and retweeting to their own followers?

Literature review

Research on libraries and Twitter largely focuses on how libraries are using Twitter, whether through case studies examining an individual library (Cahill, 2011) or an examination of the content libraries tweet (Aharony, 2010; Shiri and Rathi, 2013). Much of this research focuses on how academic libraries are using Twitter. In a content analysis of the tweets from 17 academic libraries, Al-Daihani and Alawadhi (2015) found that the top categories of tweets were news and announcements (15%), library collections (8%), library services (4.5%), and technology (2.7%). Communication and interaction between the library and its followers largely consisted of tweets that fell into the suggestions/satisfaction and interaction. Subcategories under suggestions/satisfaction included gratitude and greetings which, according to the authors, indicates that "libraries listen to and appreciate their followers" (p. 1011). Using a text mining approach, Al-Daihani and Abrahams (2016) examined the Twitter feeds of 10 academic libraries. The authors found that all the libraries in their study most frequently tweeted about library resources and community building; however, there was much more variety between the libraries when it came to other tweet categories. For example, some libraries had questions and answers as their third ranked category while others offered study support. Al-Daihani and Abrahams suggested that the prevalence of categories such as questions and answers indicated that libraries were interested in interacting with users.

What these two studies had in common was the suggestion by their authors that academic libraries were interested in engaging and interacting with their followers on Twitter. However, in a comparative study of user-librarian interaction on both Twitter and Weibo, the leading microblogging application in China, Huang, Chu, Liu and Zheng (2017) found that Twitter was most often used for information dissemination of library news (51%), followed by information sharing (25%), and communication (23%). In contrast Weibo was largely used for communication (49%), then information dissemination (39%), and information sharing (13%). The authors suggest that these differences indicate that Weibo interactions were more reciprocal in nature compared to the use of Twitter as a broadcasting tool. As a result, there were more one-on-one interactions between libraries and followers on Weibo than on Twitter. And, regardless of the microblogging application used, posts that sought out an interaction generated more retweets. Other studies support Huang et al.’s finding that there are low engagement rates on Twitter for academic libraries. In their study of four academic libraries in Montreal, Winn, Rivosecchi, Bjerke and Groenendyk (2017) found that generally there was low engagement with library tweets with an average of 3.15 likes or shares per tweet over a nine-month period. Similarly, in an analysis of the Twitter networks of two academic libraries, Yep, Brown, Fagliarone and Shulman (2017) found that there were fewer accounts and interactions amongst the replies and mentions Twitter networks of the library than the follower/followee networks and that the strongest relationships were between the libraries’ Twitter accounts and other institutional accounts, i.e., accounts related to the broader university community.

There is comparatively little research examining public libraries’ use of Twitter. In a survey of Canadian public libraries with Twitter accounts, Cavanagh (2016) found that while respondents indicated some concerns with using Twitter, namely management issues relating to responding to follower tweets and dealing controversies, Twitter was regarded as a way for public libraries to communicate local and library events with Twitter followers and to participate in local community-building by "bringing together local business, community groups and individual library followers into a single conversation space" (p. 257). These two perceived uses of Twitter are also reflected in content analysis of public libraries’ Twitter feeds. In an examination of both public and academic library tweets, Aharony (2010) found that public library tweets fell into four categories: Library in general, information about, general recommendations, and technology. The majority of tweets (50%) fell under library in general and was comprised of tweets about library events, the library’s collection, services, and libraries generally as well as book recommendations and references. This category was follow by information about (33% of tweets) which included tweets about blogs, lectures, events, classes and other topics of potential interest for the library’s Twitter followers. In contrast, academic libraries tweeted about the library collection (28%), library services (24%), library events (18%), libraries in general (12%), book recommendations (10%), and references (9%). In addition to differences in the content public and academic libraries tweet, Aharony noted that there were some differences between how public and academic libraries tweet, specifically around the formality of the language used and the number of tweets.

Shiri and Rathi (2013) expanded on Aharony’s (2010) categorization scheme using tweets from the Edmonton Public Library’s Twitter account. Like Aharony (2010), Shiri and Rathi (2013) found that the Edmonton Public Library mostly used their Twitter account to communicate traditional library interests, such as announcements about library events or activities, information sharing, making recommendations that lead the user to the library’s catalogue; however, they also found that interactions with users, such as responding to Twitter users about library-related services (advisory services) and discussions with users about films and other kinds of popular culture (informal conversation) were amongst the top five categories of tweets.

Methods

Data were collected from Social-Biblio.ca, an archive of Canadian public library Twitter data. Social-Biblio.ca is a "real-time feed and open data archive" (http://social-biblio.ca/) of public library tweets. The archive collects the tweets, retweets, modified tweets (i.e., retweets that have been altered in some way) of 189 Canadian public libraries. We downloaded tweets posted from September 10-23, 2017 from the five libraries with the highest number of followers: Toronto, Edmonton, Halifax, Vancouver, and Calgary municipal public libraries. These libraries were selected because it was assumed that a high number of followers would result in more instances of engagement and interaction. The timeframe was randomly selected in order to represent a typical period of library tweets; however, given that libraries often run special events and promote community events, we acknowledge that there is likely no typical two-week period for many public libraries. In total, there were 432 original tweets in the data set.

Several coding schemes have been used to categorize library tweets (see Shiri and Rathi, 2013). None, however, reflected this study’s concern with library-user relationships. Therefore, we chose an inductive approach to coding. Coding consisted of two phases. First, the research team each separately coded a sample of approximately 100 tweets from the overall data set to identify content categories. Following this, the research team met to review and discuss the initial coding categories. Differences in interpretation were reconciled through team discussion until consensus was reached. After this first phase of coding, 12 content categories were identified. The second phase of coding, completed by a research assistant, applied these 12 categories to the entire data set. In consultation with the research team, the initial 12 coding categories were then further grouped into five more comprehensive categories. This classification allowed us to analyse the data at both a global and more granular level. A full description of each coding category and sub-category can be found in the Appendix.

In order to measure user engagement, the likes and retweets of each tweet in the data set were collected by a research assistant and one of the researchers. Retweets and likes were selected as a way to measure user engagement because they are two of Twitter’s most commonly used affordances. Originally, Twitter’s like affordance was called favouriting. This affordance changed to likes in 2015; however, the function of the like/favouriting affordance remained the same. Although retweeting often resembles information diffusion or sharing, Boyd, Golder and Lotan (2010) argue that retweeting does "not simply get messages out to new audiences, [they also] validate and engage with others" (p. 1). Or, as Naveed, Gottron, Kunegis and Alhadi (2011) argued, retweets act as a follower’s expression of "interestingness". Likes function in a similar manner. Stvilia and Gibradze (2014) interpreted them as an indication of a tweet’s usefulness and value for a follower.

Results

What did public libraries tweet about?

The five libraries posted 432 original tweets during the two-week study period. Inductive coding resulted in five categories that we grouped into 12 sub-categories (see Table 1).

The libraries tweeted most frequently about programs (29.86%, N=129), in response to a user query (19.91%, N=86), or about collections (15.28%, N=66). (See figure 1). The categories with the fewest tweets were spaces (0.46%, N=2), humour (0.69%, N=3), and user feedback (0.69%, N=3).

What types of tweets solicited user interaction or engagement?

Overall, a large proportion of the tweets posted by libraries were liked (79.4%, N=343). The number of likes per tweet ranged from zero likes to 115. Twitter users liked library tweets 2,497 times. The average number of likes per tweet was 5.78.

When analysed by sub-category, Twitter users liked every tweet related to the physical presence of the library: building (100%; N=15) and spaces (100%; N=2). (See Table 2). They liked a large proportion of the services (93%, N=28) and programs (91%, N=118) tweets. Twitter users liked tweets relating to interaction less frequently: user feedback (33%, N=1), responses (50%, N=43), and participation (56%, N=10).

As previously discussed, there was substantial variation in the number of likes per tweet. The average number of likes per tweet was calculated to illustrate which categories were most liked my users. Categories with the highest average number of likes were partnerships (25.2%), spaces (16.5%), and humour (14.67%). Categories with the lowest average rate of like to tweet were the three categories grouped as interaction: responses (0.62%), user feedback (0.67%), and participation (1.72%).

What were Twitter users motivated to retweet?

More than half tweets were in the data set were retweeted (62.73%, N=271). The number of retweets per tweet ranged from zero retweets to 106. Overall, Twitter users retweeted library tweets 1,541 times during the study period. The average number of retweets per tweet was 3.57.

When analysed by category, Twitter users were mostly likely to retweet tweets relating to the physical library: spaces (100%, N=2) and building (80%, N=12). They also retweeted a large proportion of tweets about resources: collections (88%, N=58), services (80%, N=24), and programs (79%, N=102) (See Table 3). Twitter users retweeted tweets relating to interaction less frequently: user feedback (33%, N=1), responses (50%, N=43), and participation (56%, N=10). The category with the highest average retweet per tweet was partnerships (24%), followed by services (6.93%) and collections (6.65%). The categories with the lowest average retweet per tweet were response (0.05%), participation (0.94%), and general interest (2.40%).

Discussion

Although our coding scheme differs to those previously used to analyse libraries’ tweets, there are some similarities between the findings of this study and those of previous studies. Our finding that libraries tweeted about traditional library interests and activities, such as services, programming, and collections, is supported by Aharony (2010) and Shiri and Rathi (2013) as well as studies of academic libraries tweeting behaviour, including Al-Daihani and Abrahams (2016). Similarly, our finding that interaction was a frequently tweeted category supports the data collected by Shiri and Rathi (2013).

Overall, when public libraries tweet, Twitter users tend to respond in some way, with three- quarters of tweets receiving a positive acknowledgement using the like affordance and over half being retweeted or shared. Winn et al. (2017) argued that academic library tweets had a low engagement from Twitter users with an average of 3.15 likes or retweets per tweet. Stvilia and Gibradze (2014) found a significantly lower rate of engagement in their study of the Twitter accounts of six academic libraries. In their study, tweets were retweeted an average of 0.67 times and favourited, or liked, an average of 0.23. Our study found that public libraries have, on average, more engagement with their clients with an average of 3.57 retweets per tweet and 5.78 likes per tweet.

Traditional cornerstones of public library offerings – collections, service and programs – seem to play a big role in library-user interaction on Twitter. Programs and resources are among the most frequently tweeted by libraries, between 89 and 93% of tweets in these categories are liked, and they also have a high frequency of retweets (between 79 and 88%). These findings are similar to Stvilia and Gibradze’s (2014) findings where traditional academic library services, such as study support, received the highest average number of both retweets and favorites, followed by the community building category; however, due to differences in the classification scheme used in their study, they found that resources was the fifth highest category retweeted, but the third highest category for favourites. Stvilia and Gibradze suggest that this discrepancy may be the result of Twitter users using the favourite function to bookmark tweets for later use. Although a similar discrepancy between the number of retweets and likes does not occur in this study, the high portion of likes in the resources category and the community events sub- category suggests that the Twitter users in this study were using likes in a similar manner.

Our results indicate that Twitter is good medium for communicating with and generating interest among users about traditional library roles and services. This supports Huang et al.’s (2017) finding that Twitter was primarily used as a broadcasting tool by academic libraries. However, the findings of this study also suggest that by only using Twitter to communicate about events and resources public libraries are overlooking other opportunities to engage users. For instance, while tweets about the physical library were rare in the data set, they generated a disproportional reaction from Twitter users. They were uniformly liked and had a high percentage of retweets, which indicates an interest in the library as a physical presence -- perhaps a pride in a building that represents the community or shared excitement in the growth of the community. Librarians may want to capitalize on this interest by tweeting more library renovations or redesign of interior spaces.

Similarly, the community relations category had a high rate of user engagement. The partnership sub-category had 80% of tweets liked and 60% retweeted and the community event category had 78.79% liked and 69.7% retweeted. These community-oriented categories appear to be unique to this study. We believe tweets of this nature are an example of public libraries continuing their traditional role as a community hub. This is an example of the active nature of library membership described by Cavanagh (2015). Membership can be defined as "an intentional form of interaction in a social group" (p. 410). Twitter users, who may or not be followers the library’s Twitter account and may or may not be registered users, but who engage with the library by liking or retweeting library tweets are interacting intentionally with the library making library members. Using Cavanagh’s typology of public library membership, we are arguing that Twitter users who interact with the library fall into two membership categories: opinion holders and social media followers and subscribers. Twitter users, both followers and non-followers, can become opinion holders by sharing their opinion about the library online. By expressing an opinion, via likes and retweets, Twitter users are indicating their interest and support of the library’s role as a community hub.

Tweets in sharing category demonstrated a different trend. The retweet rate for both sub- categories was low (humour 3.00 and general interest 2.40). This suggests that Twitter users are not inclined to share tweets from the library that are amusing or interesting and not focused on library events. However, humour tweets did have a fairly high like rate (14.67) This suggests that while the "interestingness" (Naveed et al., 2011) of these tweets was minimal for Twitter users, they did find the tweets to be useful or valuable. It also supports the finding by André, Bernstein and Luther (2012) that humorous tweets are worth reading. The category of sharing is unique to this study; however, Aharony (2010) and Shiri and Rathi (2013) did identify tweets that feel into general information and news categories, respectively; however, like with this current study where only 10.41% of fell into the sharing category, tweets that feel into these categories were very small.

One category of tweets, interaction, was focused on directly encouraging Twitter users to engage with the libraries’ Twitter accounts. The interaction category was responsible for nearly a quarter (24.77%) of all tweets in the data set. Of those tweets, 86% were responses to Twitter users enquires; however, these tweets were among the least liked and retweeted. Requests for user feedback and participation were also seldom liked or retweeted. This category illustrates the limitations of using affordances such as retweeting and liking as a measure for engagement. Although liking and retweeting can be considering measures of interest or value, the absence of likes or retweets does not indicate that Twitter users are not engaged with the library. For instance, the majority of tweets in the response sub-category were replies to user enquires. The initial query from the Twitter user is itself an indication of engagement; however, given that Twitter enables tweets to be broadcast to a large audience and for conversations between small groups and individual accounts a response from the library may simply be the end of a conversation. As a result, the response may not warrant being shared with a wider audience or be of future value. In addition, tweets soliciting user feedback or participation might not receive a reaction on Twitter itself, but they may result in direct engagement with an online survey or encourage an in-person visit to the library.

Conclusion

This study contributes to research on public library Twitter use. Past research has largely focused on content classification of tweets; however, how libraries and Twitter users engage with each other has received less attention. Our analysis identified five content categories for tweets and determined that tweets in the physical library and resources categories garnered the most user engagement as measured by likes and retweets, while tweets in the interaction category had the least user engagement. These patterns provide insights into what content Twitter users respond to or share with their network. In addition, our findings can be used by public libraries looking to engage people on Twitter. Suggestions include giving the physical library a strong online presence by highlighting renovations and lesser known spaces, reminding users of upcoming programs and events, and highlighting partnerships with other community organizations.

The use of likes and retweets as a measure of user engagement create limitations for this study. As discussed above in relation to the interaction category, an absence of likes of retweets does not necessarily mean that users are not engaging with the library, only that their engagement is not visible using this measure. Patterns that appear in the data, namely that tweets that directly sought out interaction had the least user engagement, raised more questions than they answered. In addition, this approach does not account for the influence of Twitter bots. Twitter bots are a software that are designed to independently use Twitter’s affordances, such as tweeting, retweeting, and liking, to interact with other Twitter accounts. Although we conjecture that the influence of bots is not significant in our data set, we must accept that some proportion of user engagement observed in this study may be skewed as a result of bots.

The finding of this study also suggests areas for future research. In designing the study, the intention was to determine which tweets motivated users to engage with the library; however, as coding progressed, we noted differences between the libraries’ tweeting behaviours which we perceived as stylistic differences. For example, some libraries mostly responded to user queries, while others mostly tweeted about program offerings. Although Aharony (2010) did note differences in the formality of the language used between public and academic library Twitter accounts and Huang et al. (2017) noted differences between how different microblogging services are used, no other study has explored these stylistics differences.

In addition, during our analysis, we relied on two assumptions: 1) The Twitter users who engaged with the libraries’ account were representative of library users, an assumption shared by other studies of retweets and likes; and 2) Following Cavanagh (2016), that 4% of followers of library Twitter accounts were other libraries or librarians (p. 255). However, as our analysis progressed we began to question the effect of these assumptions on our results. For example, is the finding that 100% of the tweets in the space sub-category were liked and retweeted valuable if it is Twitter users living in other jurisdictions or other libraries responsible for the likes and retweets? Analysing who was engaging with the libraries was outside the scope of this study.

However, we are currently pursuing a social network analysis of Twitter users who liked and retweeted the library tweets in this study (VanScoy, Hicks and Cavanagh, 2018). We believe that exploring the identities of the larger nodes and individual outlier followers will clarify who is engaging in conversations with public libraries and contribute to interpretation of this study’s findings. This study made a first step in understanding how Twitter users want to engage with libraries. We hope that our future work will provide a more detailed examination of this relationship.

About the authors

Amy VanScoy is an Associate Professor in the Department of Library & Information Studies at the University at Buffalo, Buffalo, NY, USA. She received her PhD from the University of North Carolina at Chapel Hill, and her research interests are professional work and practitioner thinking in the information professions. She can be contacted at vanscoy@buffalo.edu.
Deborah Hicks is an Assistant Professor at the School of Information at San Jose State University, San Jose, CA, USA. She received her PhD from the University of Alberta and her research interests focus on the professional identity of information professionals and how this informs the work they do, the organizations they run, and the relationships and communities they build. She can be contacted at deborah.hicks@sjsu.edu.
Mary Cavanagh is an Associate Professor at the School of Information Studies at the University of Ottawa at Ottawa, Ontario, Canada. She received her PhD from the University of Toronto and her research interests are contemporary online information organizations (e.g., public libraries) and their information relationships and information interactions serving community building and engagement. She can be contacted at mary.cavanagh@uottawa.ca.

How to cite this paper

Appendix. Codebook