Abstract
Police departments across North America and Europe are using Twitter for many different reasons, one of the most important being public relations (i.e., image work). This article examines the relationship between Twitter use, police image work, and public engagement in the Canadian context. On the basis of 8,174 police-related tweets sent by the Toronto Police Service (TPS) and citizens, we advance the argument that despite its dialogical potential, the TPS use Twitter first and foremost as a means to legitimize the organization and that the organizational precepts of police image appear to preclude meaningful forms of engagement with citizens via Twitter.
Keywords
Introduction
Research indicates police organizations across North America, Europe, and Australia have made Twitter part of their corporate communications platforms (see Lee & McGovern, 2014; LexisNexis, 2012). Insofar as Twitter enables officers to disseminate pertinent information, solicit intelligence, and engage citizens, police have heralded its ability to meaningfully connect officers to the communities they serve (Meijer & Thaens, 2013; Meijer & Torenvlied, 2016; O’Connor, 2015). But is Twitter helping to meaningfully connect officers to their communities? To help answer this question, we examine the Twitter practices of the Toronto Police Service (TPS). As Canada’s largest municipal police organization, the TPS claims to be “taking an active role in participating in Social Networks as a means of extending [its] reach to all members of the community [italics added].” 1 Taken at face value, one might assume that the TPS’s use of Twitter is facilitating the kinds of civic engagement that ultimately produce transparency, accountability, and trust—the very qualities that, according to Mawby (2002b, pp. 71-72) and Bullock (2017), legitimize the institution of policing in a democratic society. But if we look beyond police talking points, what does this engagement actually look like? What kinds of issues are being addressed and with whom are police engaging? In addition, what do answers to these questions tell us about, not just social media’s role in promoting civic dialogue, but also its concomitant role in the development of police legitimacy?
Scholars have begun to shed light on the aforementioned questions while drawing attention to Twitter’s varied contribution to the kinds of “image work” that typically legitimize the organization (see Lee & McGovern, 2014; O’Connor, 2015; Schneider, 2016a, 2016b). However, because the existing literature has focussed primarily on tweets sent by police, it has been unable to fully appreciate the analytical significance of the police-related tweets comprising the broader communicative context. We believe a deeper understanding of the extent to which Twitter facilitates civic engagement as well as what that engagement (or lack thereof) means for police image work requires an examination of all police-related tweets because it is the only way one can truly assess how police organizations strategically navigate the Twitter sphere.
To explore the aforementioned issues, we used Computer-Assisted Qualitative Data Analysis Software (CAQDAS) and a text-mining module to examine 8,174 police-related tweets sent by the TPS and citizens. Our data suggest that while the TPS believes Twitter has allowed officers to engage the “wider community,” actual patterns of usage suggest they rarely engage users at all and that, on those rare occasions when they do, they typically engage users who seem to take the police organization’s legitimacy for granted while ignoring users who voice critical points of view. Our argument is, therefore, twofold. First, we argue the organizational precepts of image work continue to undermine any prospect of meaningful police/citizen engagement via Twitter. Second, we make the often overlooked point that police image work in the Twitter sphere involves selective engagement with some Twitter users and not others; the latter observation being tenable because we included all police-related tweets in our analysis.
We begin with a brief overview of what Twitter is, how it works, and the broader issues scholars have been discussing since the technology’s inception. Next, we clarify what is meant by police image work before moving on to examine the literature on how police organizations are using Twitter. We then highlight how our research fits into the existing literature before delving into a detailed description of our research methods, a careful examination of our data, and a discussion regarding why our findings are important.
Literature Review
Emerging in 2006, Twitter is an online micro-blogging service that allows users to post short messages (i.e., “tweets”) in 140 characters or less using a computer or mobile device. With approximately 320 million active users, 2 Twitter users can contribute to and/or monitor ongoing exchanges about specific subjects by searching for unique identifiers called “hashtags” (e.g., #policing). In addition, one can communicate publicly with a particular user by incorporating his or her Twitter handle into a tweet (e.g., “hello @Torontopolice”) or, if users are following one another, privately by using the direct message function.
There is now a vast literature on Twitter’s social, political, economic, and cultural significance (see Murthy, 2013; Weller, Burns, Burgess, Mahrt, & Puschmann, 2014). That social scientists are eager to capitalize on what Twitter has to offer is not the least bit surprising given its accessibility and the extent to which it provides a glimpse into people’s everyday social and psychological experiences (Reips & Garaizar, 2011; Williams et al., 2013). However, with widespread scientific interest comes widespread debate; for example, the extent to which Twitter has altered patterns of civic engagement is a point of contention in current research with some scholars arguing Twitter represents a type of “democratic turn” insofar as it allows users to contest the authenticity of “official” information while generating solidarity among like-minded people (Post, McGinnis, & Moody, 2014); as Maireder and Ausserhofer (2014) note, “the platform facilitates the integration of very different actors into a common conversation” (p. 313). Some have argued this democratic turn is especially evident when it comes to how news is being produced and consumed: the argument generally holds that Twitter has created a “virtual army of citizen journalists” (Murthy, 2011, p. 783; see also Ahmad, 2010; Post et al., 2014) who, while often ahead of the mainstream media, record and disseminate information about significant events (Ahmad, 2010). This was the case when, for example, bombs detonated in Mumbai (Dolnick, 2008), US Airways flight 1549 made an emergency landing in the Hudson River (Murthy, 2011), and Typhoon Haiyan struck the Philippines (Takahashi, Tandoc, & Carmichael, 2015).
However, other scholars are more cautious about overstating Twitter’s democratizing potential and/or its role in facilitating new forms of civic engagement (Mihailidis & Thevenin, 2013; Morozov, 2009; Rebillard & Touboul, 2010). For example, although some were quick to celebrate Twitter’s galvanizing effects during the 2010 and 2011 “Arab Spring,” more sober assessments suggest only a tiny fraction of the population in countries where resistance emerged had access to Twitter and that its effects were more diffuse than reported initially (Murthy, 2013). In addition, and as Gaffney and Puschmann (2014) note, although millions of users read other people’s tweets, most do not post their own content on a regular basis and, if and/or when they do, their motives are often difficult to discern (Hermida, 2010; Murthy, 2013). In fact, if it is true that 75% of Twitter’s content comes from 5% of users (Greer & Ferguson, 2011), then not overstating its democratizing and/or civic potential is perhaps the most prudent course of action. Nevertheless, one thing is certain: Twitter has been embraced widely by individuals and organizations alike and is now an important part of our broader media ecology (Bruns & Burgess, 2014).
Policing, Image Work, and Twitter
Police organizations manage external communications carefully to portray their mandate, mission, and actions in the best light possible (Chermak, 1995; Manning, 1978). Although the content of these communication strategies has changed over the past century, the need to symbolically craft a positive organizational image for public consumption has remained constant. In fact, strategically using the media to legitimize the police organization—that is, to cultivate a shared belief that police conduct and the rules/statutes enforced via their authority are morally and ethically just and, therefore, worthy of voluntary compliance (Tyler, 2003)—has become fully institutionalized across North America to such an extent that police departments routinely hire civilians with public relations expertise to help “sell” the organization (Lee & McGovern, 2013). Guided by internal policy and managed by specialists, many police communications departments are now fully professionalized and civilianized (Mawby, 2002a; Surette, 2001). This has led some scholars to argue that since the mid-1980s, the corporate and managerialist approach to police communication has ultimately served the interests of police organizations at the expense of accountability (Doyle, 2003; Mawby, 2002b; Terpstra & Trommel, 2009). Whatever the case may be, and as Lee and McGovern (2011, 2014) argue, police organizations’ widespread use of corporate communications strategies and techniques (including social media) now means “the image of policing and operational policing are inseparable” (Lee & McGovern, 2011, p. 104): that is, the representation and promotion of an active, professional, accountable, and transparent police image helps legitimize the organization while facilitating the public compliance necessary for police to carry out their mandate (Lee & McGovern, 2013). In short, these strategic, communicative processes comprise a dramaturgical form of “image work” whereby all police communications are intended to underscore the authority and legitimacy of the organization and its members (Bullock, 2017; Chermak, 1995; Ericson, 1982; Manning, 1997; Mawby, 2002b).
With the relationship between media, image work, and police legitimacy established, we now take a more discerning look at the emerging literature on social media’s role in the aforementioned processes, focusing in particular on the role of Twitter. We then highlight an important methodological shortcoming in the literature which we believe has prevented a more accurate understanding of the relationship between police Twitter use and image work from emerging.
Although some police organizations have managed to incorporate social media technology into their organizational frameworks faster and more successfully than others (Goldsmith, 2015; LexisNexis, 2012; Schneider, 2016b), it is not the least bit surprising that Twitter has been embraced given its ubiquity and communicative efficiency. Of course, the social media platform has not replaced traditional modes of police/public communication (press releases, television, radio etc.); rather, it parallels and at times supplements existing communications platforms while allowing police to bypass traditional media outlets and send information directly to Twitter users (Heverin & Zach, 2010; Procter, Crump, Karstedt, Voss, & Cantijoch, 2013; van de Velde, Meijer, & Homburg, 2015). The Seattle Police Department (SPD), for example, began posting its emergency calls on Twitter to show citizens what a day in the life of an officer was like (Seelye, 2011, A17). Unlike traditional media, however, Twitter has dialogical potential which, according to police, allows them to connect meaningfully to the people they serve.
The academic literature on how police use Twitter is starting to emerge; for example, scholars have noted the technology is used extensively to circulate crime-related information, public service announcements, and community event details while also allowing officers to respond to simple questions from citizens (Crump, 2011; Haverin & Zach, 2010; Meijer & Torenvlied, 2016; O’Connor, 2015). Addressing the issue of image work more directly, Schneider (2016a, 2016b) demonstrates how Twitter is used by police to enact specific presentational strategies. His study of the TPS, for example, effectively demonstrates how officers use Twitter to promote the organization’s commitment to community policing and police professionalism. A similar argument is made by O’Connor (2015) who argues that police departments across Canada promote community building via Twitter as part of a broader image management strategy.
However, scholars seem to be at odds over whether Twitter is actually facilitating meaningful police/citizen engagement. Some scholars argue Twitter does not offer any meaningful communicative engagement with the public (Bullock, 2017; Crump, 2011; Procter et al., 2013). Bullock (2017), for example, argues that social media staff across England are skeptical of Twitter’s dialogical potential due to reputational risks associated with online interaction, officer beliefs that Twitter is better for broadcasting information rather than engaging with the community, and the fact that some social media staff don’t have the personality and/or skills necessary to use Twitter as a social relations tool. However, other scholars, including Schneider (2016b) and O’Connor, (2015), suggest even limited interaction is important because it reflects a noteworthy shift in the way police engage citizens and legitimize the organization. No doubt part of the problem lies in how scholars are conceptualizing “engagement.” O’Connor (2015), for example, suggests that inviting the public to attend events, requesting information pertinent to ongoing investigations, and merely mentioning other users constitutes engagement. Other scholars measure engagement on the basis of how many Twitter followers the police have (Crump, 2011), whether users send crisis-related information to police Twitter handles (Procter et al., 2013), and whether users retweet police-authored tweets (van de Velde et al., 2015). In contrast, we argue that engagement should be conceptualized as something far more meaningful. As Kent and Taylor (2002) argue, engagement involves a willingness to understand and appreciate the values, interests, and concerns of others by acknowledging their voice despite one’s ability to ignore it. More specific to policing, engagement should reflect a willingness to explain how the organization works and how decisions are made while acknowledging and being open to independent scrutiny (Mawby, 2002b, pp. 71-72); in short, even when limited to 140 characters, there should be substance to the exchange.
How exactly does one properly assess police claims of meaningful engagement with citizens in the Twitter sphere? Thus far the literature has focussed incorrectly on tweets sent by police personnel while largely ignoring the significance of police-related tweets sent by citizens (see O’Connor, 2015; Schneider, 2016b). For example, O’Connor (2015) readily admits that his analysis was “purposely one-sided” (O’Connor, 2015, p. 4). That being the case, and because his analysis did not include tweets sent by citizens, his work fails to demonstrate exactly how police Twitter use is reflective of “community building” (O’Connor, 2015, p. 10). Moreover, his conclusion that “there seemed to be a genuine attempt to engage with the public [italics added]” is purely speculative and would require an exploration of the actual Twitter interactions between the police and public (O’Connor, 2015, p. 10; see also Brainard & Edlins, 2015). In another Canadian study, Schneider (2016b) provides a detailed analysis of over 100,000 tweets sent by the TPS and concludes that “police calls to the public are managed and orchestrated for organisational purposes [italics in original]” (p. 143). He importantly highlights the way TPS members engage in apolitical and non-controversial discussions with the public and demonstrates how the TPS attempts to legitimize police–community relations by appearing personable and relatable during Twitter interactions. Unlike O’Connor (2015), Schneider (2016b) actually examines interactions between the police and the public, but the analysis of those interactions is somewhat limited as there is no systematic and methodologically rigorous attempt to examine tweets sent by the public. Instead, Schneider (2016b) focusses on a limited number of exchanges, including the TPS responding to a self-proclaimed “anarchist,” personal exchanges about sports, and humorous exchanges about stereotypical donut-eating police officers. While Schneider (2016b) overcomes O’Connor’s (2015) limitation by examining tweets sent by the public, we would argue it is imperative that police-related tweets sent by officers and citizens be examined more systematically and in relation to one another because what matters is not just what police say on Twitter, but also how often they engage other users, whom they choose to engage (or ignore), and which issues they appear willing to address.
TPS Social Media Strategy
The TPS was the first Canadian police department to integrate social media into their day-to-day operations (Meijer & Thaens, 2013). While Twitter was originally used to disseminate information about local traffic, the TPS sought to formalize its social media use in 2011. Their social media strategy has been described by senior officers as being a “bottom-up initiative” where certain officers are given the freedom to engage in “authentic communication” in an effort to expand their social networks and build long-term relations with citizens (Meijer & Thaens, 2013, pp. 347-348). Indeed, what sets the TPS apart from other police departments is their apparent decentralized and impersonal social media practices (Meijer & Thaens, 2013; Schneider, 2016b). Through social media, the TPS believes it can engage the public more often and in more meaningful ways; the outcome, proponents argue, will be an improved level of trust between the police and the public.
Methods
In early Twitter research, data sets were often small because data had to be archived manually (Rogers, 2014). More recent scholarship, however, has benefited from the availability of Application Programming Interfaces (APIs) capable of automatically extracting and archiving Twitter data over time. In fact, scholars can use APIs to generate datasets comprising hundreds of thousands of cases. 3 Moreover, because manually coding large data sets can be impractical, researchers now commonly use CAQDAS (Einspanner, Dang-Anh, & Thimm, 2014; Fielding, Fielding, & Hughes, 2013). For the purposes of this study, a commercially available API and SPSS’s Modeler (with text-mining) were used to collect and analyze the data.
Between 13 August and 13 October 2013, tweets containing the “@torontopolice” Twitter handle were collected and archived for analysis; therefore, tweets sent by the TPS social media officer (who manages the official TPS Twitter account), other TPS personnel, and citizens comprise the dataset. 4 At this point, it is important to note that incomplete data sets are not uncommon in Twitter-based research because streaming APIs only archive tweets containing the researcher’s chosen criteria (Einspanner et al., 2014; Gaffney & Puschmann, 2014). In the present context, for example, tweets that made reference to the “Toronto Police” would not have been recorded unless they also contained the @torontopolice handle. That said, our research questions should mitigate concerns about incomplete data to some extent. Users who deliberately incorporated the TPS handle into their tweets did so knowing others were likely to monitor exchanges about the TPS via the organization’s handle. Thus, our data comprise tweets sent by users who, perhaps more than anyone else, wanted to be part of the social media “conversation” about the TPS.
Data Preparation
Three important modifications were made to the data before our analysis began. First, we deleted all tweets sent by a particular user who routinely accused the TPS of racism. To be clear, his or her tweets were not deleted because of what was being said but, rather, because he or she posted the same tweet repeatedly. Second, using information obtained from the TPS website, we created a binary variable that differentiated TPS from Non-TPS users (coded 1/0, respectively). Third, we created an additional binary variable that differentiated initial messages from replies (again, coded 1/0). Identifying reply tweets was a relatively straightforward task because they usually start with the intended recipient’s Twitter handle (e.g. “@MLeone8768 @Torontopolice Thanks for the RT”); however, this wasn’t always the case insofar as some users started their new tweets in the same way. Thus, to ensure our reply codes were valid, every tweet starting with a Twitter handle was reviewed online where it was possible to determine whether it was, in fact, a reply. 5 After all three modifications were complete, the 14,145 remaining tweets were imported into SPSS Modeler to be analyzed using its text-mining module. Text-mining is an “automated processes for analyzing unstructured, text-based data from a wide range of sources including document transcriptions/repositories, open-ended surveys, websites, blogs, wikis, databases, and other electronic sources” (Michalski, 2014, pp. 814-815). Modeler’s text-mining module uses Natural Language Processing (NLP) and a series of linguistic resource dictionaries to extract recurring concepts from text-based data before creating conceptual categories against which it “scores” each unit of analysis. For example, the concepts homicide, theft, and assault were extracted and used to create the category “crime.” 6 The software then identified and counted the number of tweets containing one or more of the extracted concepts.
Although more efficient than manual coding, CAQDAS is prone to obscuring the context-specific meaning of data. Therefore, data analysis tends to be a semi-automatic and iterative process because researchers must periodically adjust how the software extracts and categorizes concepts (Einspanner et al., 2014; Fielding et al., 2013). For example, Modeler initially extracted the concepts “guns,” “bullets,” and “shooting” to create the categories “firearms,” “munitions,” and “sports” (respectively). Because the tweets were actually about gun crime, the original categories were deleted and the concepts were placed under “Crime.”
Sometimes preserving the data’s context-specific meaning required the use of more complex techniques. Boolean operators, for example, were used to create “linguistic rules” capable of identifying specific combinations of concepts and aligning them with an existing or newly created category. Table 1 provides three examples of how linguistic rules were used in this study to preserve the data’s context-specific meaning.
Examples of Linguistic Rules.
In addition, preserving context-specific meaning also required the creation of custom dictionaries that could be set to supersede the base dictionary when specific concepts were identified. For example, the text-mining software initially associated “copper” with the category “precious metals.” A review of the data confirmed, however, that citizens were using copper as a synonym for police officer. Thus, a custom dictionary entry was created to ensure the concept was categorized appropriately. The custom dictionary was also modified hundreds of times to accommodate Twitter abbreviations, uncommon spelling mistakes, acronyms, and the names of streets in the greater Toronto area. 7 Although developing a custom dictionary can be a long and sometimes arduous process, the dictionary files can be used for other projects with semantically similar data (Fielding et al., 2013; Michalski, 2014).
The iterative model building process described above led to the creation of 20 categories that accounted for 8,174/14,145 tweets (58%). Why 20 categories and not 75 or 100? Unless certain model building parameters are established before data analysis begins, the software will extract thousands of concepts and create hundreds of conceptual categories. Thus, to strike a balance between elucidation on one hand and sheer practicality on the other, a decision was made to only include categories that accounted for 25 tweets or more. 8
Results
We begin with a brief assessment of all 8,174 tweets categorized by the model to provide an overarching sense of who was tweeting and what subjects were being tweeted about. The data were generated by 3,432 different users, the vast majority of whom were non-TPS personnel (see Table 2). That said, the average number of tweets per user was significantly higher for members of the TPS. Whereas the non-TPS group generated 2.15 tweets per user, members of the TPS generated 9.06. In short, compared to civilians, TPS personnel tweet more often.
Tweets by Source (%).
TPS: Toronto Police Service.
Table 3 shows the combined distribution of tweets across the 20 categories. 9 Consistent with the literature (see Heverin & Zach, 2010; Lieberman, Koetzle, & Sakiyama, 2013), tweets about crime occurred most often (2,400 tweets), followed by tweets about missing persons (1,946 tweets), and tweets about community events and/or police outreach (1,274 tweets). For example,
Distribution of All Tweets by Category.
“@Torontopolice looking for at least one suspect in Queen Alexandra Middle School shooting” “@Torontopolice search for girl, 13, missing for 2 weeks. Click through to see photo” “Hey @JosephKhargie looking sharp! Thanks for supporting our @torontopolice #MusicNotMischief program @TPSCopsRock”
Conversely, tweets about police oversight (43 tweets), political scandal (41 tweets), and the G20 Summit (37 tweets) occurred less often. For example, “From the College: Above all else, an effective authority figure knows trust and accountability are paramount. @Torontopolice #TPStraining” “RT @blackbrownbeige: Hey @Torontopolice—where DOES [Toronto Mayor] Rob Ford go when he’s not at work, anyway?” “Breaking!!! @Torontopolice officer charged in G20 assault on Adam Nobody found guilty”
Table 3 reveals the extent to which narratives pertaining to crime fighting and order maintenance dominate the Twitter sphere, thereby reflecting what is typically generated by more conventional media outlets (see Ericson, 1982; Surette, 2007).
However, a more nuanced sense of what the tweets were about can be gleaned from the category web shown in Figure 1, which renders the number of co-occurring categories visible. Readers should note the size of each node corresponds to the number of tweets that made reference to a specific category, while the width of the connector lines between nodes represents the number of times the two categories co-occurred.
TPS and non-TPS category web.
Given our primary interest is in how TPS and non-TPS tweets relate to one another, a few general observations about the combined category web will suffice. First, the web reveals how often multiple categories appear in a single tweet; indeed, contrary to popular opinion which has, at times, criticized Twitter for its brevity (if not its banality), the web demonstrates clearly that users address multiple subjects of importance in a single 140 character post. Second, although tweets were usually informational or transactional (i.e., users were either sending or requesting information; see Brainard & McNutt, 2010), the co-occurrences involving “praise for police” and “critical of police” demonstrate users were willing to post normative comments about the police and/or police-related activity. Overall, the combined data show that when it comes to police-related communications, the Twitter sphere is teeming with 140 character posts suggesting some degree of conceptual and normative depth.
Examining TPS and Non-TPS Tweets
Our comparison of TPS and non-TPS tweets begins with a review of what we call their “category density”; that is, the number of categories and co-occurrences identified for each user group. When comparing the TPS and non-TPS category webs (see Figures 2 and 3, respectively), we are able to see clearly the important differences in how Twitter is being used by both groups.
Non-TPS category web.
TPS category web.
The category density of non-TPS users is visibly higher, suggesting non-TPS users who include @torontopolice in their tweets address a greater number of categories than police while also tending to connect those categories more frequently and in more diverse ways. Specifically, whereas Modeler scored non-TPS tweets into all 20 categories, it scored TPS tweets into 15; moreover, whereas the web for non-TPS users shows myriad co-occurrences, police appear to keep their tweets relatively simple by rarely addressing more than one issue at a time. Interestingly, although perhaps not surprising, the categories missing from the TPS web include “critical of police,” “G20 Summit,” “Sammy Yatim shooting,” “political scandal,” and “policing and race.” TPS users were, therefore, never self critical and did not tweet about the socially and politically sensitive subjects that were being routinely acknowledged by civilians. Thus, when the two category webs are compared, we get our first glimpse of how police strategically navigate the Twitter sphere: unlike non-TPS users, police keep their tweets simple and avoid socially and politically sensitive subjects prone to disagreement.
Normative Tweets
Although both user groups tweeted extensively about mundane issues (e.g., social media, transportation, community events etc.), the data revealed important differences with respect to normative tweets. For example, non-TPS users praised police for their efforts in relation to a number of different initiatives, including missing persons (172 co-occurrences), community events or outreach (26 co-occurrences), social media (22 co-occurrences), police recruitment (20 co-occurrences), and crime (20 co-occurrences). For example, “RT @G_Smitherman: Thanks to @Torontopolice, Christopher is alive + safely in the hands of TO Western Hospital. Love&Thank you to all” etc.” “Check me out on the cover of @InTorontomag next week. I’ll be discussing the great work of the @Torontopolice LGBT Internal Support Network!” “Local @Torontopolice now Twitter Celebrity, @TPSChrisBoody #BacktoSchool advice is RT’d by @katiecouric + 1000 ++” “RT @TPSFernandes: Kudos to @tps54 @youthinpolicing students who removed #graffiti from a senior’s home!. @Torontopolice” Congrats to DS Gary Giroux @Torontopolice on the arrest of suspect in the #Cabbagetown #murder.”
Likewise, TPS users praised fellow officers for their efforts in relation to police recruitment (14 co-occurrences), community events or outreach (16 co-occurrences), and social media (5 co-occurrences): “I’m at the @Torontopolice #YIPI (Youth In Policing Initiative) grad right now. So proud of them and our very own @TheStringPuppet, Malik!” “RT @OfficerRod8037: When A Community Claims An Officer As Their Own, its truly Amazing. Great Job #OfficerBradford @Torontopolice #33CRU ht” “@SgtVirji Thx Aly—this award is validation for all the good social media work by our @torontopolice members #TopCop #smilecon”
However, it is the difference in the extent to which each user group used Twitter to praise the police that is of interest here. Although non-TPS users tweeted more messages of praise overall, TPS-users were actually more active in terms of the number of tweets sent per user. Specifically, although 3,316 non-TPS users generated 508 tweets praising the TPS (a tweet/user ratio of 0.15), 116 TPS users generated 115 tweets (a tweet/user ratio of 0.99). Thus, although both user groups used Twitter to publicly praise the police, officers were far more active in that regard, taking advantage of the social media platform to bolster their organizational image.
Non-TPS users were also critical of the police, calling their conduct into question with respect to their use of tasers (64 co-occurrences), the shooting of Sammy Yatim (84 co-occurrences), policing and race (19 co-occurrences), and the G20 Summit (12 co-occurrences). For example, “SO disspointed [sic] in @Torontopolice decision to give all front line cops tasers. What happened to training, de-escalation tactics? #topoli” “@VigilantCanuck: After @torontopolice Murdered #SammyYatim W/ Nine Bullets & A Taser, The Answer Is … Drumroll … #Moartasers” “RT: @DurhamGal1: @Torontopolice @marksaunderstps Carding & racial profiling is a reminder of dirty tricks of apartheid South Africa, has no place in #Canada” “@Torontopolice Why 2 cmmndrs who ran #G20 detention cntr not charged despite OIPRD report saying should be. B/c they retired? Fair? #TOpoli”
In contrast, and as noted above, TPS users avoided tweeting about the sensitive issues being raised by non-TPS users. Interestingly, although “Use of Tasers” is present in the TPS web (see Figure 3), at no time did TPS users comment on the legitimacy of taser use. In fact, the category is present only because several officers tweeted information about an upcoming public meeting that was to involve a taser-related discussion. 10
Replying to Tweets
As mentioned previously, meaningful engagement involves a willingness to understand and acknowledge another person’s voice, despite one’s ability to ignore it (see Kent & Taylor, 2002; Mihailidis & Thevenin, 2013). In the present context, this begs the following questions: did TPS users reply (i.e., acknowledge) to other people’s tweets? And, if they did, what were the instigating tweets about? Of the 1051 TPS tweets, only 74 (7%) were direct replies, of which 38 (51%) were in response to tweets sent by other TPS users. Furthermore, our linguistic model was able to categorize 66 of the 74 instigating tweets. As Table 4 indicates, TPS users replied most often to tweets that praised the TPS (22 tweets), pertained to community events and/or outreach (15 tweets), or contained PSA or crime prevention information (10 tweets). For example,
Distribution of Instigating Tweets by Category.
Instigating tweet: “Good work by a former @Torontopolice officer documenting distracted driving” TPS Reply: “@chrisjamesdrew Thanks, I will take a look! @Torontopolice #walkTO #BikeTO” Instigating tweet: “Great meeting of @TPS14 CPLC. Robust agenda. Proud of our community partners & the important work we continue to do together @Torontopolice” TPS Reply: “@CHINTvCanada @tps14 @Torontopolice thank you for your commitment to community and on going support. It is much appreciated” Instigating tweet: “Proud of my friends @AnneMarieBatten @GraffitiBMXCop @Torontopolice for launching #RealTimeCrisis to prevent suicide‚ http://t.co/y36fv3AZNn” TPS Reply: “@BeingUncosmo We’re no where near a ’launch’ Stephanie, but we keep plodding along~Thanks for your support! @AnneMarieBatten @Torontopolice”
Conversely, TPS replies were almost non-existent when it came to socially or politically sensitive issues. In the case of the Sammy Yatim shooting, for example, the lack of police engagement on the topic may reflect the TPS’ inability to properly control and manage the influx of user counterclaims and interpretations in the aftermath of the controversial incident (see Schneider, 2016a, pp. 99-122). Although more than 200 tweets attempted to bring some aspect of Sammy Yatim’s shooting to the TPS’s attention, officers replied only twice, both times in response to the same instigating tweet: Instigating tweet: “Wonder if Officer Forcillo [the officer accused of shooting Sammy Yatim] was strip searched like all others who have to be in court cells? My guess: nope” TPS Replies: “@selwynpieters @LawyerToronto The SIU investigation is ongoing—please contact them directly if you want more info” “@Torontopolice @selwynpieters @LawyerToronto The @Torontopolice search procedure is public document—NO automatic strip search + each case on own merit”
Likewise, although there were 313 tweets that were critical of police conduct, TPS users replied only once: Instigating tweet: “On top of that @torontopolice are also w/holding important, related FOI video that IS ready. @afixedaddress @judemacdonald @GraffitiBMXCop” TPS Replies: @judemacdonald have sent a message to Cst Turner @TPSmkwa (twitter is new for her) @Torontopolice @afixedaddress @djjohnso @NicoleTanguay2
Thus, even though thousands of tweets were meant to catch the TPS’s attention, officers rarely replied. On those rare occasions when they did reply, the TPS usually responded to praise or tweets relating to community events and/or outreach, PSAs, or crime prevention: that said, just over half of all TPS replies were in response to tweets authored by other police officers.
The data indicate the TPS uses Twitter to disseminate relevant information and draw the public’s attention to its organizational successes. However, the data also show they avoid tweeting about sensitive issues, rarely reply to tweets sent by citizens (especially when those tweets broach sensitive issues), and reply most often to tweets sent by other police officers. Below we will discuss these patterns of usage in the context of civic engagement as well as image work and police legitimacy.
Discussion
The TPS has been using Twitter in some capacity since 2008 and is now recognized as being somewhat of a leader in the Canadian context (see Meijer & Thaens, 2013; Schneider, 2016b). According to Meijer and Thaens (2013), the TPS believes better police–citizen relations can be realized by “enabling authentic, individual external communication by police officers” (Meijer & Thaens, 2013, p. 348) and that social media platforms like Twitter allow for those communications to take place. But is this true? Is Twitter facilitating the kinds of civic engagement that ultimately lead to transparency, accountability, and trust? Not likely.
Consistent with the literature (see Haverin & Zach, 2010; Procter et al., 2013), our data indicate TPS officers rarely engage citizens via Twitter. Interestingly, this finding appears to contradict Schneider’s (2016b) work insofar as his study of the TPS determined “TPS officers [interact] with members of the public often” (Schneider, 2016b, p. 140). It is possible that methodological differences are at the heart of this discrepancy; for example, our decision to use @torontopolice as our selection criteria may have lead to an unknown number of TPS replies being overlooked. That said, because Schneider does not specify how “often” interactions took place relative to the total number of TPS tweets, exactly how different our results are is not entirely clear. Nevertheless, when officers did engage non-TPS users, it was typically in response to comments that appeared to take the legitimacy of the police organization and its operations for granted. In such cases, the police had only to reiterate a narrative for which a sympathetic audience already existed. This is not to suggest such exchanges are unimportant. Tweets about the success of a charity bicycle race, the fate of one’s favorite hockey team, or a community barbecue do personalize what is oftentimes a very guarded organization (see Meijer & Torenvlied, 2016; Schneider, 2016b); however, unlike O’Connor’s (2015), we argue such tweets do not constitute meaningful engagement because the police (and citizens, for that matter) do not have to acknowledge differences of opinion nor submit themselves to the kinds of interpersonal risks that are at the very heart of open, honest, and empathetic dialogue (see Kent & Taylor, 2002).
Perhaps most important is our finding that TPS users avoid addressing difficult subjects, whether by not initiating sensitive exchanges or by completely ignoring tweets that call their legitimacy into question. Thus, we would argue that O’Connor (2015) is incorrect when he says, “By utilising social networking sites, the police are forced to address in a very public way, and sometimes from anonymous sources, criticisms of their job performance [italics added]” (p. 10). As our data show, the police are not forced to acknowledge criticism at all; in fact, image work via Twitter involves officers making strategic decisions about whose tweets will be acknowledged and whose will be deliberately ignored, a pattern of use that is only visible when one takes all police-related tweets into consideration.
For decades, scholars have claimed police engage in image work by strategically managing their communications to construct a legitimate public image of their organization (Ericson, 1982; Manning, 1997; Mawby, 2002b). Our data involving the TPS suggest that police use of Twitter continues to reflect that organizational mandate insofar as officers use Twitter routinely to disseminate information (often of a self-congratulatory sort) that deviates rarely from culturally expected narratives highlighting their well-established role in order maintenance and law enforcement (see Doyle, 2003; Surette, 2007). The precepts of image work, which are firmly embedded in the culture of policing (see Crump, 2011) and which demand a carefully managed, if not entirely contrived, public image, are essentially negating the possibility of honest dialogue about issues of substance, including police violence, race relations, and issues of police accountability—the very issues that continue to strain police/citizen relations in communities across North America (Dyson, 2017; Wortley & Owusu-Bempah, 2011; Wortley & Tanner, 2003). Indeed, any suggestion that Twitter facilitates the kinds of civic engagement that constitute an important “democratic turn” is misguided.
Should the TPS engage non-TPS users on Twitter more often while openly and honestly addressing socially and politically sensitive issues? We think so. Although we recognize officers would not be able to comment on matters currently before the courts (see Schneider, 2016b), at the very least they could improve on the degree to which they acknowledge the comments and opinions of Twitter users who bring sensitive issues to their attention. As Kent and Taylor (2002) suggest, acknowledging the voice of others is at the heart of honest public relations. These are tall orders to be sure, especially given resource implications and officers’ tendency to call the public’s trustworthiness into question (Loftus, 2010; Skolnick, 1966). Nevertheless, we believe even incremental improvement in this regard would be worthwhile.
Before discussing the shortcomings of our study, a few words about our methodology are in order, especially given the extent to which it differs from existing literature. One of the advantages of CAQDAS with text mining is efficiency. Once the analytic model has been configured sufficiently to accommodate idiosyncrasies and context-specific meanings (Fielding et al., 2013), it can be used with semantically similar data sets well into the future. Of course, scholars can also make incremental tweaks to the model to continuously improve its analytic performance. While it is true that one’s analytic model is only as good as the software’s NLP capabilities and the integrity of its linguistic dictionaries, a noteworthy feature of SPSS Modeler with text-mining is that users can import custom dictionaries created by other scholars working on related projects; thus, improved and more comprehensive dictionaries become a shared resource. Of course, all of this means one’s analytic model can continue to improve over time all the while taking advantage of improved processing power to analyze truly massive text-based data sets.
Our study has several shortcomings that should be acknowledged. First, we make the implicit assumption the TPS is a good model for how Twitter is used by other police services across Canada. Whether this is a valid assumption is difficult to discern. As O’Connor (2015) and Meijer and Thaens (2013) note, police departments have their unique social media styles; moreover, some departments lack the necessary resources to capitalize fully on the technology which, in turn, alters their patterns of usage (LexisNexis, 2012). Nevertheless, we are fairly confident that most large police departments in Canada use social media platforms like Twitter in similar ways. Also, unlike Schneider (2016b) who monitored TPS Twitter activity for 21 months (collecting more than 100,000 tweets), our study collected data over a shorter 2-month period. 11 That said, we don’t believe collecting more data over a longer period of time would have yielded different patterns of usage in part because what drives the TPS’s strategic use of Twitter are social structures deeply embedded in the organization and which have, thus far, seemed rather impervious to change.
Conclusion
As police organizations continue to incorporate social media technology into their organizational frameworks, it is imperative that we understand how social media platforms like Twitter are changing the nature of police-related communications. Although Twitter has not replaced traditional forms of communication, it has surely provided new ways for citizens and police to interact. Thus, this article sets out to investigate the relationship between Twitter use, police image work, and public engagement in the Canadian context, and it did so while addressing a methodological limitation present in existing scholarship; specifically, scholars have not properly assessed the relationship between Twitter use, civic engagement, and image work because they have ignored the significance of all police-related tweets sent by citizens. Thus, conclusions have been based on rather one-sided analyses of police tweets. Understanding what police say on Twitter is important, but one must also assess how often police engage other users, whom they engage, and which issues they appear willing to address. Using a more meaningful definition of engagement, and despite TPS claims to the contrary, we argue the organizational precepts of image work continue to undermine any prospect of meaningful community engagement via Twitter and that, despite its dialogic potential, Twitter is being used, first and foremost, to legitimize the police organization. Moreover, we argue this image work via Twitter involves selective engagement whereby officers engage sympathetic users (albeit it infrequently) while ignoring those who call their legitimacy into question.
Undoubtedly Twitter has transformed how police communicate and there is clear evidence that officers are using Twitter to engage citizens in new ways (Schneider, 2016a, 2016b); however, any assessment of how police strategically engage citizens via Twitter must take the broader communicative context into consideration and adopt a more meaningful definition of engagement, one that goes beyond mundane exchanges to include meaningful interactions with substance. Indeed, one can only assume these deeper and perhaps even more authentic exchanges would contribute significantly to the development of a new police image grounded in transparency, accountability, and trust.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.