This study builds on the findings of previous research into church-related tweets by considering and analysing the content of 1,004 tweets containing the word “church” which were posted from within Greater London during a 10-week period. The qualitative analysis of the resultant Twitter dataset enabled the research team to validate a previously proposed coding framework for church-related tweets, subject to a minor change, and provide new evidence around the types of content discussed in church-related tweets posted from across all London boroughs. This paper concludes by drawing-out some of the key lessons which sociologists of religion can learn by exploring church-related tweets.
Defining digital theology
Defining social media research
Setting the context
With 500 million posts per day , the popular microblogging site Twitter is increasingly capturing the attention of sociologists seeking to understand the attitudes and behaviours of social media users. Despite the abundance of church-related posts on Twitter, little research to date has focussed on such posts in detail. Where research investigating church-related tweets has been conducted, the focus has tended to be on the application of quantitative research methods to understand how Twitter data can be used to complement more traditional off-line research methods (e.g., Cooper, 2018, 2017). This paper seeks to address this imbalance by presenting a detailed qualitative analysis of over a thousand church-related tweets, seeking to understand the topics which were discussed in those tweets and the common themes which could be used to categorise such tweets.
In approaching this challenge, this paper builds-on previous work by Cooper (2014), which considered a small sample of 108 church-related tweets posted from within the London Borough of Camden. That study considered each tweet in turn, assigning content labels to the tweets and grouping together tweets containing discussion of similar topics. As an output of that study, the following coding framework for church-related tweets was proposed:
Table 1: Thematic coding labels used in a previous study to categorise 108 church-related tweets from the London Borough of Camden, with the volume of tweets captured by each thematic coding label and percentage of total tweet volume (Cooper, 2014). Thematic coding label Volume of tweets captured by each thematic coding label Percentage of total tweet volume Discussion about church attendance 59 55% Discussion about church service content 5 5% Discussion about theology, the Bible or belief 15 14% General church discussion 40 37% Miscellaneous 5 5% Non-church discussion 4 4% Unknown theme 10 9%
The research presented in this paper builds on Cooper’s work by repeating the method used in that 2014 study across a bigger dataset, to robustly analyse the content of a sample of church-related tweets and to determine whether the thematic coding framework presented in that previous study can be robustly validated for use in future research. Therefore, the dataset used in this study necessarily originates from the same time period (April 2014 to June 2014) but contains tweets from all London boroughs.
Defining digital theology
The emerging field of digital theology is an interdiscipline analogous to digital humanities which itself has been variously defined. Among the most useful in regard to the latter is Stan Ruecker’s definition, recorded from the ‘Day of Digital Humanities’ (2009):
I often say that humanities computing involves three distinct research areas. First, some researchers apply computing to research questions in the humanities. These might be questions they’ve always pursued but can now pursue faster or at a larger scale, or they may be questions that could not be addressed satisfactorily at all without computers. Second, some researchers take computing as an object of study using humanities methods. Examples include cyberculture and posthumanism. Third, some researchers take a generative approach, creating new online materials or tools for subsequent study and use. Most of my own work is in this third area.
This understanding of digital humanities might be simplified by folding Ruecker’s third category in with the second. As a result, we understand digital humanities to be a field that (i) applies the questions and methods of computing to the humanities (e.g., sentiment analysis of nineteenth century Scottish literature); and/or (ii) applies the questions and methods of the humanities to computing (e.g., a historical analysis of humanities computing) .
Digital theology, then, might be similarly understood as a field that (i) applies the questions and methods of computing to subjects within theology (e.g., sentiment analysis of the Bible; determining church activity with geo-location metadata); and/or (ii) applies the questions and methods of theology to computing (e.g., a theological reflection on the impact of social media on adolescence). Furthermore, digital theology as an academic field still in the making has the potential to go even further than this by facilitating paradigmatic transformation of both computing and theology where the subjects encounter each other in new ways. Thus, as an interdisciplinary field, digital theology has the potential to reshape intradisciplinary research within both parent fields. It should be added that while many definitions for the subject of theology exist, theology is understood within this paper to be Christian theology, which can be defined as “Christian reflection on the ideas of faith” .
Alongside this attempt to define Digital theology in light of digital humanities, in their recent paper on defining digital theology, Phillips, et al. (2019) propose an alternate definition of the field, comprising four different levels of digital theology:
- “DT1: The use of digital technology to communicate or teach theology as a traditional academic subject” 
- “DT2: Theological research enabled by digitality or digital culture” 
- “DT3: Intentional, sustained and reflexive theologically-resourced engagement with digitality/digital culture” 
- “DT4: A prophetic re-appraisal of digitality in the light of theological ethics” 
This alternate definition broadens the field considerably and takes digital theology beyond the scope of computing and theology, by creating the opportunity to appraise “the impact of technology on society at large drawing upon the technological toolkit of scripture and tradition” .
Although multiple definitions of digital theology exist at this early stage in the field’s lifecycle, it is clear from the diverse definitions available that digital theology is a broad area of study which encapsulates both theoretical and practical research with real-world applications always in mind. During the recent COVID-19 pandemic, for example, churches have had to respond quickly to evolving circumstances and stay at home orders, in order to be able to provide new and increased offerings of online services and community activities. Scholars of digital theology have been quick to explore and publish findings on these practical responses (e.g., Cooper, et al., 2021), to understand the different ways in which churches around the world have responded and to provide timely insight to guide and support church leaders taking similar decisions during the ongoing pandemic, which at the time of writing this article continued to limit the ways in which churches can operate in-person. At the same time, digital theologians have had to grapple with more theoretical and academic issues arising from the rapid up-take of online worship, such as the extent to which congregants can benefit from watching/participating in the sacrament of the eucharist online, therefore, not sharing in physical presence with the presiding priest (e.g., Parish, 2020). While such studies have tended to take a high-level approach to considering important theological issues and the use of technology, the real-world applications of such discussions are very much rooted in practice and have tangible relevance to, and impact on, those who worship and practise the Christian faith.
As research in the field of digital theology continues, it is important that digital theologians are able to harness a wide range of research methods and analytical techniques, to be able to provide the most accurate, insightful and timely analyses to guide real-world practice and, therefore, achieve optimum research impact. Such methods and techniques will include a wide range of quantitative and qualitative approaches, as well as mixed-method approaches such as social media research, which makes use of often freely available social data to better understand views, attitudes and behaviours.
Defining social media research
In defining the field of social media research, this study acknowledges the Oxford English Dictionary definition of social media:
“Websites and applications that enable users to create and share content or to participate in social networking.” 
Therefore, this study considers social media research to be the study of social media Web sites and applications to consider how such platforms are used, why such platforms are used and the impact arising from the use of such platforms. Social media research can, thus, either make passive use of social media platforms (e.g., capturing data from social media platforms to analyse discussion content or views/attitudes) or active use of social media platforms (e.g., creating online space in which to conduct a research interview or focus group digitally, using a social media platform).
Social media research methods are valuable within the field of digital theology as they allow the rapid interrogation of high-volume datasets to generate timely and impactful insight. They also permit researchers to explore views and discourse which are, generally speaking, less susceptible to observation bias than samples of data collected under study conditions. This means that it is often possible to generate timely understanding of the issues that people think about and talk about online, in order to test existing theoretical frameworks and generate new ideas for empirical testing. In the context of digital theology specifically, this means that it is possible, for example, to explore religious views and ethical debates present in the online discourse and assess these views and debates in the context of previous theology research. The fact that it is possible to quickly and easily collect and analyse new data means that it is possible to track changes in the online discourse temporally, enabling a greater understanding of the ways in which views form or change over time. It is also possible, using metadata on the location from which social media posts were created, to track changes in the online discourse spatially, enabling a greater understanding of the ways in which views form or change in different geographical locations.
Within this paper we seek to explore the content of a sample of church-related tweets posted from within Greater London, to assess the validity of an existing framework for qualitatively coding church-related tweets and to generate new insight into the topics which are discussed in church-related tweets.
Setting the context
Published studies have explored religious discourse across a range of social media platforms including blogs (Burgess, 2013), Myspace (Bobkowski and Pearce, 2011), and Twitter (Naaman, et al., 2010). Notably, Cheong (2014) published findings of a study which investigated common practice in the authorship of tweets which embed and are inspired by scripture, finding that church leaders were “appropriating new patterns of interactions including social media communication to relegitimize and construct new practices of authority” . In a separate paper on reading the Bible within digital culture, Phillips (2018) notes that analysis of the most tweeted Bible verses across several years shows “a gradual shift over time from broadly propositional texts to broadly therapeutic texts” .
Where previous studies have focussed on content analysis of religious discourse on Twitter, they have tended to use content analysis to deduce and propose coding frameworks for use in future similar studies. Codone (2014), for example, made use of a coding framework originally proposed by Naaman, et al. (2010) to analyse tweet content by two prominent church leaders from the U.S. Codone’s study made use of this existing framework to present findings around the content of tweets by those church leaders and additionally proposed a revised framework, specifically focussed on coding and categorising tweets by church leaders.
Similarly, Holmberg, et al. (2016) conducted a content analysis of tweets directed to @God. The authors described in that study how a Twitter account had existed with the handle @God for around nine years prior to the study and the account had a following of around 180,000 Twitter users at the time of the study. The authors posited that as well as capturing tweets aimed directly at the @God user account, they were also likely to capture many tweets written actually to God. As well as providing an analysis of the types of tweet posted by users and directed to @God, the authors proposed a qualitative coding framework for classifying tweets to @God (or, perhaps, to God).
Cooper (2018, 2017, 2014) has focussed a recent programme of research towards developing an understanding of how Twitter data can be used to improve and contribute to quantitative studies of church growth. This programme of work has included:
- content analysis of Twitter data (Cooper, 2014)
- analysis of the relationship between sentiment expressed in church-related tweets and the presence of church growth in the London borough from which the tweet was posted (Cooper, 2017)
- analysis of geolocated church-related tweets to understand whether it is possible to identify churches excluded from previous quantitative church attendance studies (Cooper, 2018).
This present paper builds on Cooper’s (2014) work, by repeating the methodology used in that small-scale study of tweets posted from the London Borough of Camden, using a bigger dataset of a sample of tweets posted from all London boroughs, during the same time period. The aim of this study is to explore whether the coding framework proposed in 2014 can be validated using a dataset comprising tweets posted from all London boroughs and to provide new insight into the types of discussion covered in church-related tweets.
This study seeks to answer a number of research questions, designed to further knowledge around how the church is discussed on Twitter. The main aim of this research is to consider a dataset of church-related tweets and classify them into themes, to provide an understanding of what types of topics are discussed in church-related tweets. Once this question is answered, the research seeks to explore, in light of these identified themes, what lessons can be learned by sociologist of religion, seeking to use online data to further their knowledge and understanding of churches and church communities. Finally, using a bigger dataset than has previously been used, this research seeks to explore whether the previously presented qualitative coding framework can be validated for future use against social media data.
Hence, the following research questions will be addressed:
RQ1: Is the existing qualitative coding framework for church-related tweets fit-for-purpose?
RQ2: What topics are discussed by social media users in church-related tweets?
RQ3: What can sociologists of religion learn from the content of church-related tweets?
Collection of Twitter data
A sample of Twitter data  were collected for analysis using the Twitter Application Programming Interface (API) . The API provides a freely available time slice of tweets matching set search criteria, rather than all tweets meeting the search criteria. While it is not clear how truly random resultant samples are, in the social science sense of the word, the API sample approach provides a manageable volume of tweets spread across the period of interest. This approach was considered appropriate for this study as it was comparable to previous studies of Twitter data.
The search criteria were set so that a sample of tweets were collected which were posted each Sunday over a 10-week period, commencing Sunday 20 April 2014. The tweets were written in the English language, contained the word “church” and were posted from within a 60 km. radius of an arbitrary point within central London. In total, 19,894 tweets were obtained within the sample; 1,234 (6.2 percent) of the tweets within the sample contained geo information (a latitude and longitude value for the location from where the tweet was posted). This low proportion of tweets containing geo information was consistent with the rates reported in previous academic studies (e.g., Cheng, et al., 2010). As part of the data cleaning and preparation process, tweets which did not contain geo information were discarded. The remaining 1,234 tweets were then analysed using the MapIt API  to establish the London borough from which the tweets were posted. As part of this process, 230 tweets which were determined to have been posted from outside Greater London were also discarded. Following the data cleaning and preparation process, the dataset contained 1,004 tweets which were used within the analysis.
The sample of 1,004 tweets analysed within this study were tabulated together for qualitative coding. As a coding framework for church-related tweets had already been designed and was being tested within this study for suitability and relevance, qualitative coding was conducted using the labels from that coding framework. The lead analyst commenced the research phase by independently working through the content of each tweet in turn and allocating as many labels as was relevant to each particular tweet. Where a tweet might also have warranted a new label, this was noted for later review around the suitability of the existing coding framework and the necessity for new labels. On completion of the lead analyst’s primary coding, the full dataset, including coding labels, was sent the supporting analyst for secondary coding. The role of the second coder was to review each tweet in turn and consider whether the allocated labels were suitable and appropriate. Again, where the secondary coder felt a tweet might also have warranted a new label, this was noted for later review. On completion of secondary qualitative coding, the primary and secondary coder worked through each tweet for which there was a coding disagreement, to discuss the content of the tweet and reach consensus over the most suitable and appropriate labels. Once the qualitative coding of the full dataset was agreed between the primary and secondary coders, a discussion took place on the appropriateness of the existing qualitative coding framework for church-related tweets, including a review of whether it was necessary to introduce new coding labels.
The analysis conducted for this study confirmed the suitability and applicability of the church-related tweet coding framework previously proposed in the small-scale thematic content analysis of church-related tweets posted from within the London Borough of Camden (Cooper, 2014), subject to the addition of one new coding label — Discussion about Hillsong . Although this addition alters the coding framework to some extent, it was not entirely unexpected — the original Camden study had already noted that “at least 38 of the 108 tweets (about 35 percent) within the sample appeared to be related to Hillsong church” (Cooper, 2014). Therefore, this change to the proposed coding framework is considered to be minor.
Each tweet coded as part of this study across all London boroughs was able to be allocated to one or more of the themes within that framework or the new Hillsong theme. Though a small number of tweets could have been allocated to a new theme entitled ‘church events’, both coders were satisfied this small grouping of tweets were well represented by the thematic coding label ‘general church discussion’ and, therefore, it was not deemed necessary to add any additional labels to the coding framework. Thus, RQ1 has been addressed affirmatively by this research study, subject to the one minor change. The coding framework and associated tweet volumes for all London boroughs are presented in Table 2:
Table 2: Thematic coding labels with the volume of tweets captured by each thematic coding label and percentage of total tweet volume for 1,004 tweets from all London boroughs. Thematic coding label Volume of tweets captured by each thematic coding label Percentage of total tweet volume Discussion about church attendance 605 60% Discussion about church service content 72 7% Discussion about theology, the Bible or belief 82 8% General church discussion 339 34% Miscellaneous 201 20% Non-church discussion 85 8% Discussion about Hillsong 51 5% Unknown theme 47 5%
A comparison between Table 1 and Table 2 reveals the proportion of tweets allocated to the thematic coding labels is roughly consistent between the sample of tweets coded from the London Borough of Camden and the sample of tweets coded from all London boroughs. The only difference greater than 10 percentage points was the thematic coding label ‘Miscellaneous’, which increased from five percent of tweets in the study of Camden tweets to 20 percent of tweets in the study of tweets from all London boroughs. This is likely to be as a result of increased diversity in the nature and content of the conversation, as the pool of church tweeters increased and diversified as the sample size of Twitter users increased. Though this study did not identify any new thematic coding labels which needed to be added to segment the ‘Miscellaneous’ tweets, future studies drawing church-related tweets from an even greater sample of church-tweeters could explore a research question around the continued suitability of the ‘Miscellaneous’ thematic coding label.
To illustrate differences between the various different London boroughs, the results of this study are differentiated by London borough, in Table 3, in which the following abbreviations are used:
- Discussion about church attendance — DCA
- Discussion about church service content — DCSC
- Discussion about theology, the Bible or belief — DTBB
- General Church Discussion — GCD
- Miscellaneous — MISC
- Non-church discussion — NCD
- Discussion about Hillsong — HLSG
- Unknown theme — UN
Table 3: Thematic coding labels with the volume of tweets captured by each thematic coding label and percentage of total tweet volume for 1,004 tweets from all London boroughs. Thematic coding label DCA DCSC DTBB GCD MISC NCD HLSG UNK Barking and Dagenham Borough Council 34 61% 8 14% 1 2% 14 25% 20 36% 2 4% 0 0% 1 2% Barnet Borough Council 17 68% 0 0% 1 4% 13 52% 9 36% 1 4% 0 0% 0 0% Bexley Borough Council 8 47% 2 12% 4 24% 8 47% 4 24% 3 18% 0 0% 0 0% Brent Borough Council 17 71% 1 4% 1 4% 3 13% 7 29% 2 8% 0 0% 1 4% Bromley Council 17 65% 2 8% 1 4% 10 38% 7 27% 0 0% 0 0% 1 4% Camden Borough Council  62 57% 5 5% 17 16% 40 37% 5 5% 5 5% 38 35% 9 8% City of London Corporation 7 44% 0 0% 1 6% 1 6% 8 50% 2 13% 1 6% 0 0% Croydon Borough Council 22 67% 3 9% 2 6% 15 45% 7 21% 2 6% 0 0% 1 3% Ealing Borough Council 16 70% 1 4% 3 13% 9 39% 7 30% 0 0% 0 0% 0 0% Enfield Borough Council 18 56% 3 9% 2 6% 8 25% 6 19% 2 6% 0 0% 3 9% Greenwich Borough Council 28 57% 4 8% 3 6% 14 29% 11 22% 3 6% 0 0% 3 6% Hackney Borough Council 17 40% 3 7% 0 0% 9 21% 6 14% 16 38% 0 0% 4 10% Hammersmith and Fulham Borough Council 6 67% 0 0% 0 0% 1 11% 4 44% 1 11% 0 0% 0 0% Haringey Borough Council 15 65% 3 13% 3 13% 3 13% 3 13% 3 13% 1 4% 3 13% Harrow Borough Council 8 67% 0 0% 1 8% 6 50% 0 0% 0 0% 0 0% 1 8% Havering Borough Council 8 36% 2 9% 0 0% 6 27% 5 23% 7 32% 0 0% 0 0% Hillingdon Borough Council 9 50% 2 11% 1 6% 6 33% 2 11% 1 6% 0 0% 2 11% Hounslow Borough Council 11 73% 1 7% 1 7% 6 40% 5 33% 0 0% 0 0% 0 0% Islington Borough Council 17 57% 0 0% 1 3% 9 30% 7 23% 1 3% 0 0% 3 10% Kensington and Chelsea Borough Council 20 69% 2 7% 2 7% 7 24% 1 3% 2 7% 0 0% 1 3% Kingston upon Thames Borough Council 6 50% 0 0% 0 0% 3 25% 2 17% 5 42% 0 0% 0 0% Lambeth Borough Council 22 48% 3 7% 3 7% 20 43% 7 15% 6 13% 0 0% 0 0% Lewisham Borough Council 30 73% 4 10% 2 5% 18 44% 4 10% 0 0% 4 10% 2 5% Merton Borough Council 14 54% 2 8% 4 15% 11 42% 4 15% 0 0% 0 0% 2 8% Newham Borough Council 18 64% 2 7% 2 7% 13 46% 3 11% 1 4% 0 0% 1 4% Redbridge Borough Council 9 56% 1 6% 0 0% 9 56% 5 31% 0 0% 0 0% 1 6% Richmond upon Thames Borough Council 6 43% 0 0% 2 14% 4 29% 3 21% 4 29% 0 0% 0 0% Southwark Borough Council 41 56% 4 5% 5 7% 35 48% 14 19% 8 11% 0 0% 2 3% Sutton Borough Council 5 83% 0 0% 1 17% 2 33% 3 50% 0 0% 0 0% 0 0% Tower Hamlets Borough Council 11 52% 6 29% 2 10% 6 29% 3 14% 2 10% 0 0% 4 19% Waltham Forest Borough Council 14 74% 2 11% 4 21% 5 26% 2 11% 2 11% 1 5% 0 0% Wandsworth Borough Council 19 76% 4 16% 2 8% 10 40% 13 52% 1 4% 0 0% 1 4% Westminster City Council 53 78% 2 3% 10 15% 15 22% 14 21% 3 4% 6 9% 1 1% 605 72 82 339 201 85 51 47
The data in Table 3 show that while there are variations in the types of content discussed in church-related tweets across the London boroughs, there tended to be a broad range of coverage across the various thematic coding labels in each London borough — other than the ‘unknown’ thematic coding label, there are few instances (49 of 231 possible instances, or, about 21 percent of possible instances) of the other labels having zero tweets recorded against them in the various London boroughs. This demonstrates that church-related tweets are diverse in content covered across the U.K.’s capital city, and the church-related tweet coding framework has a relevance in studies exploring church-related tweets from across all parts of London.
To illustrate the types of content discussed in the sample of church-related tweets analysed, and thus fully address RQ2, this paper will consider each thematic coding label in turn, presenting an agreed  definition of the coding label and a series of examples of the types of tweets allocated to each label.
Discussion about church attendance Tweet content which explored presence in church, e.g., indication that the tweeter, or tweet subject, had previously attended a church, was present in a church, might attend a church in the future or had made a decision not to attend a church.
After church, What next?(Posted at 12:23 on 27 April 2014)
@anonymised: I’m actually in church!! Banns being read today. If i see god I’ll give him a reght good, ay up odes.” Say one for me(Posted at 11:16 on 20 April 2014)
Ain’t been church inna while ima make sure I go next Sunday(Posted at 13:24 on 11 May 2014)
@anonymised: God forgive me for not coming church today I had no money for Oyster cardlolololol(Posted at 11:41 on 20 April 2014)
Discussion about church service content Tweet content which explored the content of a church service, e.g., a sermon, acts of worship or social interactions taking part within a church service.
Great reflective Church service today♥♥(Posted at 20:42 on 25 May 2014)
Idm going church but they talk for so long and my attention span can’t hack it(Posted at 10:11 on 11 May 2014)
Really good Pentecost worship at New Community. Many, many languages spoken, and sung, in worship. Just as church should be.(Posted at 15:35 on 8 June 2014)
What church has a DJ between Services? #becausewecan #ilovemychurch #hillsongfamily #warehousethingz @...(Posted at 15:47 on 8 June 2014)
Discussion about theology, the Bible or belief Tweet content which explored religious belief or theory, the nature of God or interpretation of religious texts.
“@anonymised: Easter is the oldest Christian holiday and the most important day of the church year. #HappyEaster”...Christmas...?(Posted at 13:58 on 20 April 2014)
Faith is taking the first step even when you don’t see the whole staircase. @ St Marylebone Parish Church(Posted at 14:48 on 18 May 2014)
I may not go to church every Sunday , but I comunicate with god daily(Posted at 19:49 on 18 May 2014)
The devil doesn’t care if you go to Church or read your Bible, as long as you don’t apply it to your life(Posted at 22:59 on 25 May 2014)
General church discussion Tweet content which contained discussion of church other than church attendance or church service content, e.g., church buildings, church bells or the role of church in society.
@anonymised: The Southend Greek Orthodox Church having to find £1500 to repair a stained glass window smashed by vandals :-((Posted at 07:11 on 8 June 2014)
Whilst the ringing of church bells is appealing to some, they’re just a pealing to everyone else.(Posted at 06:57 on 1 June 2014)
Why did my church move to Essex pls(Posted at 07:53 on 18 May 2014)
Church.(Posted at 10:28 on 20 April 2014)
Miscellaneous Tweet content related to church but fairly unique and varied in nature and volume, such that the content did not naturally align to the other thematic coding labels used within the framework.
I can’t believe the amount of people in church this morning who were taking selfies. I mean seriously, you’re in church!!!(Posted at 11:48 on 1 June 2014)
i don’t know what to wear for church :( also, last minute i’m playing during communion like uh no ty(Posted at 09:40 on 20 April 2014)
Jacket stolen in church....? Particularly weird when my bag is untouched!(Posted at 19:49 on 8 June 2014)
With a lot of messing about in Google Maps, managed to find name of church I was planning on. Walk begins.(Posted at 11:33 on 15 June 2014)
Non-church discussion Tweet content which was un-related to the Christian church, e.g. tweets related to the surname Church or bars/venues called church.
Back in London, lots of vintage furniture on Church Street today #Therestoration @ Stoke Newington ...(Posted at 14:15 on 8 June 2014)
Christ Church College, University of Oxford. The most beautiful campus I have seen. @ University of ...(Posted at 19:16 on 4 May 2014)
Doesn’t your school’s caf look like this? #hogwarts @ Christ Church College Oxford(Posted at 19:07 on 1 June 2014)
Football is My Religion. The Emirates is My Church!! Is not Easy to be an Arsenal fan but is hard not to LOVE this Team. #GOONERS♥☺♥(Posted at 15:18 on 11 May 2014)
Discussion about Hillsong Tweet content which was specifically related to Hillsong church.
#Church #SundayService #Hillsong #DominionTheatre #PraiseGod @ Hillsong London / Central London(Posted at 14:45 on 4 May 2014)
A Wow experience at #hillsong church in London this am. #moving(Posted at 17:27 on 1 June 2014)
Easter Sunday Service! @ Hillsong Church London(Posted at 17:48 on 20 April 2014)
Now that’s what I call a church....amazing service at Hillsong this morning! X(Posted at 11:33 on 4 May 2014)
Unknown theme Tweet content which could not be thematically coded, due to coder inability to understand/interpret the nature of the discourse.
Church tins(Posted at 11:09 on 1 June 2014)
I lost my plan for tomorrow. I wanted to go out but I don’t know now, they’ve got a church most of friend.(Posted at 22:14 on 25 May 2014)
Red church blend @anonymised(Posted at 13:26 on 22 June 2014)
There’s a storm coming.. #storm #stormy #darksky #streetlight #church #newvsold #robot #thunder ...(Posted at 09:41 on 25 May 2014)
The fact this study has been able to affirmatively answer RQ1, by validating the 2014 coding framework is an important step forward for scholars working in digital theology and the sociology of religion. This finding demonstrates that, when testing innovative new research methods, exploratory studies using small sample sizes can add value and enable quick insights around the art of the possible for bigger, more robust, future studies. The validation of the coding framework with a much bigger sample size also means the research community now have a framework which can be used in future studies. Such studies could consider different time periods, different cities, even different countries — thus, increasing the accessibility of far greater understanding of online church discourse and how it can change/evolve.
In answering RQ2, this study generated important new insight around the nature of the topics which are discussed by social media users in church-related tweets. The finding that 60 percent of tweets analysed included discussion about church attendance, for example, provides encouragement to digital theologians that social media data might be of real value to future studies investigating church growth and/or decline.
The finding that five percent of tweets analysed included discussion about Hillsong indicates that social media data might, going forward, be a useful resource for furthering academic understanding around the activities and practices of particular churches and denominations. This study did not set out to make comparisons between particular churches and denominations, but in light of this finding, there is clear value in future studies within the field of digital theology specifically exploring the online discourse related to particular Christian communities. It might be possible through such research to improve understanding of the similarities and differences between the activities and operations of those communities and the impact which those communities have on the lived experience of attendees. It might also be possible for future studies to take a deep dive into particular communities, to better understand the online discourse related to these communities. Such studies might enable a greater and more informed understanding around the beliefs and theologies of those communities (a number of Hillsong tweets within this study talked about beliefs related to Easter Sunday), as well as the practical aspects of those communities such as their physical infrastructure (several Hillsong tweets within this study referred to the church’s Dominion Theatre venue), their service logistics (discussion about the queue to get into the service was observed within the Hillsong tweets) their social activities (discussions about going to church with friends and about after-church meals were observed within the Hillsong tweets), their leadership structures and their visions for the future.
The finding that seven percent of tweets analysed included discussion about church service content indicates that there is much that digital theologians can learn from social media data about the structure, format and content of acts of worship. Several church service content tweets talked about the topics which were prayed for in the services which the content posters had attended. This indicates that it might be possible for future studies to explore the topics of prayer which church attendees discuss online. By analysing these tweets, it might, depending on their representativeness, be possible to draw inferences around changes in church prayer focus over time or across geographical locations. This offers the potential for digital theologians to get much closer to understanding the topics close to the hearts of Christian worshipers and the ways in which those Christian worshipers present their petitions and concerns to God. Several church service content tweets also talked about approaches to worship (e.g., the liturgy spoken during services and the style of musical worship used during services) encountered within the churches which the tweets refer to. The presence of such tweets indicates that it might, in future, be possible for digital theologians to use social media research methods to further their understanding around the ways in which church services are structured and conducted. This might be of increased importance in contexts in which church service content is likely to be subject to rapid change — for example in the rapid transition from traditional in-person services to online services, in order to comply with stay at home orders required to reduce the spread of the COVID-19 virus, which is affecting day-to-day life across the globe at the time of writing this paper.
The finding that eight percent of tweets analysed included discussion about theology, the Bible or belief offers hope and potential to digital theologians that social media data might be a rich source of insight into the views and doctrines held by those who observe religious faith. Several tweets coded as covering content related to theology, the Bible or belief related to beliefs associated with the Christian festival of Easter. This observation is likely to have been driven, at least in part, by the fact that the date on which data collection for this study commenced (Sunday, 20 April 2014) happened to be Easter Sunday. This indicates that it would be helpful for future studies to explore the extent to which there is a seasonality effect within the content of church-related tweets and the extent to which significant dates in the church calendar influence/drive the topics observed in church-related tweets. Discussion around the issue of sexuality was also observed in several of the tweets which were coded as covering content related to theology, the Bible or belief. The issue of sexuality and the response of the church to evolving views of sexuality has been pressing for an extended period, and is likely to remain of significant academic importance for some time into the future. It would, therefore, be of clear benefit to the academic literature for future studies to investigate the extent to which online discourse of current issues of importance to faith communities reflect the full range of off-line views and discussions and are representative of the views held by members of the faith community. Where the online discourse captures all views held and is representative of all views held, there is clear benefit in social media analysis quickly assessing those views and presenting relevant and actionable findings to the research digital theology research community in a timely manner.
It is noteworthy, however, that the research method applied in this study is laborious, time-consuming and dependent on subjective human judgement. Human resource availability is clearly going to be a limiting factor in future studies and, therefore, sample sizes in such studies are likely to be constrained to the hundreds or low thousands. Given this constraint, future research in the field of digital theology should focus on the process of automating the qualitative coding of tweets . In the first instance such studies could use classification techniques to test whether qualitative coding of this framework can be automated [19. Assuming such techniques could be demonstrated to be effective, follow-on studies could test the use of segmentation techniques to validate the coding framework proposed in this study and to identify possible new coding labels.
One limitation of this study is that it only used the word ‘church’ when generating the dataset of church-related tweets. While other useful terms could have been included, this decision was a deliberate choice, to ensure the resultant dataset was as similar as possible to that used in the small-scale 2014 study of church-related tweets from Camden. Notwithstanding this decision, future studies could also explore tweets using a wider list of keywords including ‘chapel’, ‘cathedral’, etc.
The study is also limited in that it took only a sample of tweets, posted on a Sunday, and from within Greater London. Similarly, this was a necessity to ensure the comparability of results from this study with previous published work. This limitation was also considered to be fairly minor, as the approach taken in this study was exploratory and intended to test what is possible. Future studies could repeat this work while moving away from geographically and time-bound sample datasets, to explore whether the coding framework still applies to population datasets — although there is no reason to assume it would not hold. Where future studies were limited to geographical or time-bound datasets, they could broaden the analysis conducted to consider the dynamics of other metrics, for example tweet emotion, to consider how such metrics vary across location and time. Such studies have the potential to increase datapoints available for future data science studies investigating the use of church-related Twitter data in wider studies and investigations.
To ensure RQ3 is answered as clearly and as fully as possible, it is important to focus on what sociologists of religion can do with the findings of this study. Firstly, scholars of sociology of religion and digital theology now have a baseline for online church discourse on Twitter. This enables future studies to compare changes against this baseline. As previously suggested, such changes could be over time or across geographical locations and could enable researchers to conduct monitoring or tracking to test how discourse alters in response to church or world events — for example by considering how discussion about church attendance changes in a particular region after a local controversy or scandal. Secondly, having the coding labels presented in this study provides another data feature for future work blending computer science with sociology of religion/digital theology. Previous studies have taken tweets and enriched them with sentiment data (Cooper, 2017). This study has now provided further data enrichment with coding labels. Additionally, the raw data already contain rich metadata provided by the Twitter API. As research continues to enhance and grow the dataset, this opens up new possibilities for further innovative analysis, for example, using machine learning methods to understand whether Twitter data can be used to plug gaps in traditional off-line datasets. Finally, the simple fact that people even discuss church at all is interesting in and of itself. In a world where secularisation is often assumed (e.g., Bruce, 2002), this provides further evidence that religion, in particular, Christianity, still has a role to play in people’s lives — both their online lives and their off-line lives.
Digital theology as a field of research may be young, but recent scholarship indicates that it has the potential to become a vibrant, exciting and fast-paced field of research, which may be able to blend innovative online methods with more traditional off-line methods to further our understanding of research problems across the fields of computer science, theology (including digital theology) and sociology of religion. This study has demonstrated the use of innovative methods to answer cross-cutting research questions which add benefit to each of these fields and has posed a number of emerging challenges, which if gripped quickly by the research community, could enable far-reaching and widely applicable research impact.
The findings of this study have provided new insight into the types of topics which people discuss online in church-related tweets. They have also validated an existing qualitative coding framework, which is proposed for use in future studies exploring similar Twitter datasets. These findings have immediate consequences for the research community exploring sociology of religion and church growth — they open up the possibility of future research to understand how discussions of church on Twitter change over time and across geographical locations, thus enabling future researchers to understand the relevance of online church discourse to offline church engagement. Furthermore, this opens up exciting new possibilities to explore how such Twitter data might, in future, be used to complement more traditional datasets to more accurately measure and understand church growth and decline.
About the authors
Anthony-Paul Cooper is co-director of the Centre for Church Growth Research at Cranmer Hall, Durham University. Anthony-Paul has a background in social research, with previous research topics including new church use of “secular” and “sacred” space and the use of social media data to better understand church attendance and church growth.
E-mail: anthony-paul [dot] cooper [at] durham [dot] ac [dot] uk
E-mail: anthony [dot] p [dot] cooper [at] utu [dot] fi
Dr. Joshua Mann is President and CEO of Expositus, a research and education nonprofit working in the area of digital humanities. Previously Joshua was a Research Fellow at CODEC Research Centre for Digital Theology at the University of Durham. Joshua’s research engages subjects in digital humanities and biblical studies. He is particularly interested in the hermeneutics of technology.
E-mail: joshua [dot] l [dot] mann [at] utu [dot] fi
Dr. Erkki Sutinen is Professor of Computer Science at the University of Turku and an ordained priest. Erkki’s research interests include educational technology, computing education, ICT4D, co-design and digital theology. He has supervised circa 30 Ph.D.s and co-authored around 300 papers. Erkki is currently based in Windhoek, Namibia, having recently set up the first overseas campus of the University of Turku.
E-mail: erkki [dot] sutinen [at] utu [dot] fi
Dr. Peter Phillips is Director of the Centre for Digital Theology at Durham University. With a Ph.D. in John’s Gospel and many years’ experience of teaching/researching the New Testament, Peter now explores the interface between all things digital and theological.
E-mail: p [dot] m [dot] phillips [at] durham [dot] ac [dot] uk
We gratefully thank Emmanuel Awuni Kolog for his helpful comments on the dataset presented within this paper.
1. https://business.twitter.com/, accessed 16 September 2019.
2. A version of this definition first appeared here: “A brief description of digital humanities,” Joshua L Mann (blog), 21 June 2016, https://josh.do/brief-description-digital-humanities/, accessed 28 March 2020.
3. McGrath, 2017, p. xiii.
4. Phillips, et al., 2019, p. 37.
5. Phillips, et al., 2019, p. 38.
6. Phillips, et al., 2019, p. 39.
8. Phillips, et al., 2019, p. 39.
9. https://en.oxforddictionaries.com, accessed 19 May 2019.
10. Cheong, 2014, p. 3.
11. Phillips, 2018, p. 408.
12. The dataset used in this study has been analysed and reported on in a separate study investigating the relationship between the sentiment of church-related tweets and rate of church growth in the London borough from which the tweets were posted. See Cooper (2017).
13. See https://dev.twitter.com/, accessed 28 March 2020.
14. See http://mapit.mysociety.org/, accessed 28 March 2020.
15. For which, the following definition was proposed: “Tweet content which was specifically related to the Hillsong church”.
16. A small number of tweets posted from the London Borough of Camden were coded differently in this study than in the initial 2014 study. This is to be expected on the basis of the more robust dual-coder approach used in this bigger study. Therefore, Camden volumes in Table 3 do not exactly match corresponding values in Table 1.
17. Between the primary and secondary coder.
18. This will clearly require truly multi-disciplinary teams, with partnership between sociologists of religion and computer scientists; learning around how such research teams work and collaborate could be of academic interest in itself.
19. Since this article was drafted, research has been published exploring the use of machine learning to automate the qualitative coding of church-related tweets; see Cooper, et al., 2019.
Piotr S. Bobkowski and Lisa D. Pearce, 2011. “Baring their souls in online profiles or not? Religious self-disclosure in social media,” Journal for the Scientific Study of Religion, volume 50, number 4, pp. 744–762.
doi: https://doi.org/10.1111/j.1468-5906.2011.01597.x, accessed 11 August 2021.
Steve Bruce, 2002. God is dead: Secularization in the West. Oxford: Wiley-Blackwell.
Regina L. Burgess, 2013. “Understanding Christian blogger motivations: Woe unto me if I blog not the gospel,” Journal of Religion, Media & Digital Culture, volume 2, number 2, pp. 1–42.
doi: https://doi.org/10.1163/21659214-90000030, accessed 11 August 2021.
Zhiyuan Cheng, James Caverlee and Kyumin Lee, 2010. “You are where you tweet: A content-based approach to geo-locating Twitter users,” CIKM ’10: Proceedings of the 19th ACM International Conference on Information and Knowledge Management, pp. 759–768.
doi: https://doi.org/10.1145/1871437.1871535, accessed 11 August 2021.
Pauline H. Cheong, 2014. “Tweet the message? Religious authority and social media innovation,” Journal for Religion, Media and Culture, volume 3, number 3, pp. 1–19.
doi: https://doi.org/10.1163/21659214-90000059, accessed 11 August 2021.
Susan Codone, 2014. “Megachurch pastor Twitter activity: An analysis of Rick Warren and Andy Stanley, two of America’s social pastors,” Journal of Religion, Media & Digital Culture, volume 3, number 2, pp. 1–32.
doi: https://doi.org/10.1163/21659214-90000050, accessed 11 August 2021.
Anthony-Paul Cooper, 2018. “Using geotagged Twitter data to uncover hidden church populations,” In: David Goodhew and Anthony-Paul Cooper (editors). The desecularisation of the city: London’s churches 1980 to the present. London: Routledge, pp. 134–147.
doi: https://doi.org/10.4324/9781351167765, accessed 11 August 2021.
Anthony-Paul Cooper, 2017. “Assessing the possible relationship between the sentiment of church-related tweets and church growth,” Studies in Religion/Sciences Religiouses, volume 46, number 1, pp. 37–49.
doi: https://doi.org/10.1177/0008429816664215, accessed 11 August 2021.
Anthony-Paul Cooper, 2014. “Unwrapping Camden’s church tweeters: A small-scale thematic study of Twitter data,” In: Kandy Woodfield (editor). Social media in social research: Blogs on blurring the boundaries. London: NatCen Social Research.
Anthony-Paul Cooper, Emmanuel Awuni Kolog and Erkki Sutinen, 2019. “Exploring the use of machine learning to automate the qualitative coding of church-related tweets,” Fieldwork in Religion, volume 14, number 2, pp. 140–159.
doi: https://doi.org/10.1558/firn.40610, accessed 11 August 2021.
Anthony-Paul Cooper, Ilkka Jormanainen, Annastasia Shipepe and Erkki Sutinen, 2021. “Faith communities online: Christian churches’ reactions to the COVID-19 outbreak,” International Journal of Web Based Communities, volume 17, number 2, pp. 99–119.
doi: https://doi.org/10.1504/IJWBC.2021.114453, accessed 11 August 2021.
Kim Holmberg, Johan Bastubacka and Mike Thelwall, 2016. “@God please open your fridge! A content analysis of Twitter messages to @God: Hopes, humour, spirituality, and profanities,” Journal of Religion, Media & Digital Culture, volume 5, number 2, pp. 339–355.
doi: https://doi.org/10.1163/21659214-90000085, accessed 11 August 2021.
Alister E. McGrath, 2017. Theology: The basic readings. Third edition. Chichester: Wiley Blackwell.
doi: https://doi.org/10.1002/9781119158196, accessed 11 August 2021.
Mor Naaman, Jeffrey Boase and Chih-Hui Lai, 2010. “Is it really about me? Message content in social awareness streams,” CSCW ’10: Proceedings of the 2010 ACM Conference on Computer Supported Cooperative Work, pp. 189–192.
doi: https://doi.org/10.1145/1718918.1718953, accessed 11 August 2021.
Helen Parish, 2020. “The absence of presence and the presence of absence: Social distancing, sacraments, and the virtual religious community during the COVID-19 pandemic,” Religions, volume 11, number 6, 276.
doi: https://doi.org/10.3390/rel11060276, accessed 11 August 2021.
Peter Phillips, 2018. “The pixelated text: Reading the Bible within digital culture,” Theology, volume 121, number 6, pp. 403–412.
doi: https://doi.org/10.1177/0040571X18794139, accessed 11 August 2021.
Peter Phillips, Kyle Schiefelbein-Guerrero and Jonas Kurlberg, 2019. “Defining digital theology: Digital humanities, digital religion and the particular work of the CODEC Research Centre and Network,” Open Theology, volume 5, number 1, pp. 29–43.
doi: https://doi.org/10.1515/opth-2019-0003, accessed 11 August 2021.
Received 1 April 2020; revised 30 July 2021; accepted 2 August 2021.
Copyright © 2021, Anthony-Paul Cooper, Joshua Mann, Erkki Sutinen, and Peter Phillips. All Rights Reserved.
Understanding London’s church tweeters: A content analysis of church-related tweets posted from a global city
by Anthony-Paul Cooper, Joshua Mann, Erkki Sutinen, and Peter Phillips.
First Monday, Volume 26, Number 9 - 6 September 2021