Collaborative Coding Cultures: How Journalists Use GitHub as a Trading Zone

Abstract

As the use of “big data” in journalism has become an increasing practice in recent years, its exercise is no longer limited to newsrooms in Europe and North America. The prevalence of coding cultures and collaboration in newsrooms worldwide has risen, in part to tackle challenges related to information overflow. This collaboration has been facilitated partly by the popularization of low-cost code hosting platforms such as GitHub. GitHub is a well-known public code management system with a significant potential for transparency and collaboration between communities. This study aims to map how newsrooms employ GitHub as a trading zone for collaborative journalism. We analyze the GitHub activities of leading media outlets from 27 countries. Based on content analysis of the newsroom’s repositories and users’ profiles, we found that media organizations in the United Kingdom and the United States continue to be the primary code producers in GitHub. However, our results show that newsrooms in developing countries benefit from these codes by stargazing these projects. The results of this article have important implications for understanding how newsrooms across the globe use GitHub for collaboration and the role journalistic code plays in bringing together various collaborative initiatives and actors in journalism.

Keywords:

• Data journalism
• GitHub
• trading zones
• coding
• collaborative cultures

Introduction

Collaboration across news organizations is increasingly assumed to be necessary and desirable during the practice of data-driven journalism (Alfter 2019; Sambrook et al. 2018). Indeed, in an age of “big data” where information is abundant, the sheer volume of available data sets that are relevant for reporting but in need of high-skilled analysis has prodded media workers to participate in collaborative projects to tackle more complex reporting (Lewis and Nashmi 2019; Loosen, Reimer, and De Silva-Schmidt 2020). Some of these projects have gained significant international attention during the last decade, including reporting on the leak of over 11 million documents in the Panama and Paradise Papers case or the fact-checking initiatives during the COVID-19 global health pandemic (for an overview see e.g., Alfter and Cândea 2019; Carson and Farhall 2018).

The inter-organizational collaborative phenomenon in reporting, that is, between two media organizations, also points to an emergent form of collaborative journalism. Cooperation in journalism is in no way new (Dodds 2019). However, this new type of collaborative journalism is characterized by at least two novel features. Firstly, we are observing new methods—from network analysis to graph-based keyword search systems—for the reporting, coordination, and collaboration between media organizations under the frame of a new “global network of journalism” (Berglez and Gearing 2018) that has the potential to break down professional silos in journalism (Heft 2021). Secondly, this new form of reporting further embodies the diversification of professional roles inside newsrooms. To avoid information asymmetry—newsrooms’ inability to gather, analyze, or disseminate information due, not to the lack of it, but to the affluence of data points (Gray, Chambers, and Bounegru 2012)—new computer programmers, graphic designers, entrepreneurs, and data scientists have been included in the decision-making process of news production like never before (Brossi and Dodds 2019; Lewis and Usher 2014).

These characteristics in collaborative journalism allow contemporary multimodal newsrooms to share raw data and information, as well as the elevated costs associated with high-skilled analysis. As Carson and Farhall argue after examining award-winning cases in Britain, the United States, and Australia, “investigative collaborations and digital approaches are powerful antidotes to declining revenues and falling journalist numbers as we move deeper into the twenty-first century” (2018, 1909). In times of already dwindling trust levels towards the press and click-based business models, collaborative journalism can improve small and disadvantaged newsrooms’ practices and motivations, as well as their access to investigative projects and public-interest data (Dodds 2021).

Furthermore, collaborative journalism is also gaining traction for yet another reason. Studies in the United States have shown that this type of reporting could boost transparency and accountability by showing the methods journalists use to construct their stories (Curry and Stroud 2021; Lewis and Usher 2013). Journalistic programming code helps the audience understand the process behind the stories they are reading. It also works as a script for other newsrooms that would like to analyze similar data in local contexts.

Nevertheless, despite the increasing importance of collaboration in journalism, “very few have talked about data journalism in non-Western contexts—at least not in an academic setting, leaving many wondering whether it exists” (Mutsvairo 2019). Indeed, despite the widespread use of data in journalism worldwide (Wachs et al. 2022), research has traditionally only shone a light on particular countries in North America and Western Europe (Appelgren, Lindén, and van Dalen 2019). By omitting media systems outside of Western Europe and North America, academic research has created a geographical gap in the study of data-driven journalism and focused chiefly on liberal and democratic corporatist media systems. This Western-centric approach to the study of innovation in journalism has failed to include news organizations that deal with “big data” under strictly regulated media environments (Wu 2022) or are run by grassroots and marginalized communities (Dodds 2021). Moreover, media organizations in Western, educated, industrialized, rich, and democratic countries have more resources because of their more affluent memberships and higher income through advertising. As the costs of practicing data journalism increase due to the need to hire high-skilled analysts, data journalism will likely be implemented differently in developing countries. To understand how code mediates newswork, we need to observe how different actors are appropriating codes, scripts, and programming languages to their realities and sharing their experiences with other news organizations.

This paper aims to overcome these limitations and reveal the diversity, similarities, and differences between collaborative coding cultures in journalism across the globe. Collaborative coding cultures result from the popularization of open-source code platforms, and journalists need to analyze and report the abundance of public interest data. Depending on the specific conditions of any given country, such as their levels of data access, privacy laws, or information rights, new relational and dynamic values have emerged in newsrooms across African, Middle Eastern, Asian, and Latin American countries (Bebawi 2016; Broussard and Boss 2018; Wright, Zamith, and Bebawi 2019). GitHub, one of the most popular platforms for sharing and collaboratively developing open-source code, has emerged as one of media organizations’ most popular trading zones. Its popularity has increased so much during the last years that, according to Davis, “GitHub has […] become an active social network in and of itself, where people with shared interests can follow each other and see the progress on various projects” (2015, 162). As an example of a collaborative coding culture, GitHub enables developers, journalists, and researchers with different professional backgrounds to track programming changes and collaborate on source-code management within its platform.

Founded in 2008 and now with over 80 million users (Gonzalez, Zimmermann, and Nagappan 2020), GitHub provides access to remote repositories for Git—a version control system that allows users to keep track of the changes made in their code—a feature that results particularly beneficial in team-based work developments.

This paper further aims to map how newsrooms across a range of countries employ GitHub as a trading zone for collaborative journalism, if they do it at all. The lack of studies about collaborative coding for data journalism in non-Western contexts gives this article an exploratory nature. To guide this research, we discuss the following main question: How are news organizations using the code-sharing platform GitHub? To answer this main question, we developed the following sub-research questions: (1) To which media organizations belong the most prominent journalistic repositories on GitHub? (2) What type of journalistic projects and themes are most common on GitHub? (3) Who are the primary actors interacting with journalistic repositories on GitHub?

Collaboration Cultures and Trading Zones

GitHub is one of the key platforms that afford digital trading zones for journalism. As more data becomes available for newsrooms to gather, analyze, and present to the audience, there has been an increasing demand for collaboration between news organizations (i.e., collaborative cultures) (Hellmueller, Cheema, and Zhang 2017). To facilitate collaboration between newsrooms, media organizations are now taking steps to train and hire journalists with the necessary technical skills or bring in data scientists to work alongside journalists (Guo and Volz 2021). These data scientists are expected to contribute to journalistic collaboration by providing their expertise in advanced numerical and quantitative abilities, big data analysis, and website development (Boyles 2020).

Thus, collaborating cultures, including coding, have contributed to the emergence of trading zones in journalism. Trading zones are spaces “within which individuals coming from different traditions or with distinct expertise can gather, agree on rules of exchange, and engage in complex, coordinated activity around shared goals” (Haim and Zamith 2019, 80–81). For Lewis and Usher (2016), trading zone is a valuable concept for understanding the inclusion of heterogeneous actors in journalism, their social interaction with traditional reporters, and the impact they end up having on newswork. Zhang and Ho (2022) further argue that trading zones in journalism are a sound theoretical framework to capture cross-disciplinary collaboration in contemporary multimodal newsrooms, both offline and online.

As a sign of the evolving technological landscape within newsrooms, trading zones can lead to both fruitful and challenging interactions inside and between news organizations. Data scientists typically do not undergo formal training or education specific to journalism. Their educational background typically revolves around data science, statistics, computer science, or related fields, which could present challenges for their development inside news organizations (Fairfield and Shtein 2014). However, this also opens the door for non-data personnel to contribute to data science teams by generating story ideas, evaluating stories’ viability, and ensuring data’s accuracy. At the same time, data scientists can play a crucial role in breathing life into stories by accessing datasets through automated methods and analyzing intricate data points (Zamith 2023).

However, trading zones in themselves do not always guarantee increased collaboration between news workers or news organizations. For example, using computational tools like R and JavaScript may be understandable primarily to journalists with technical inclinations, which could inadvertently exclude individuals or organizations lacking proficiency in programming (Stalph, Hahn, and Liewehr 2022).

Drawing from Galison’s (1997) first theorization on the concept, in this article, trading zones are understood as virtual and physical spaces where journalists interact, cooperate, and develop new tools with a heterogeneous group of actors with similar interests. The popularization of trading zones also fulfills three primary needs in contemporary journalism: (1) they promote the necessary diversification of professional roles, (2) they bridge the gap in educational backgrounds, thus relieving stress on media workers, and (3) they help overcome workforce limitations and economic conditions shortcomings. In this section, we will review how trading zones assist in fulfilling each of these needs.

Collaboration between news organizations has expanded beyond national borders. Cross-border collaborative journalism has gained popularity primarily due to data journalism initiatives seeking to generate in-depth reporting on complex global topics. As Heft argues, “although journalism has long been considered a profession of lone wolves, its present and future are shaped by collaborative practices of various kinds” (2019, 191). Indeed, cooperation is the norm in processing newly available data (Larrondo-Ureta and Ferreras-Rodríguez 2021). Along with these new forms of local and global collaboration, media organizations are facing the need to professionally diversify their newsrooms and hire creative artists, data scientists, and developers who can analyze and reshape journalistic content to the latest platforms media organizations need to use to distribute their news (Royal and Kiesow 2021). The integration of traditional journalism with professionals from other fields has not been exempted from difficulties (Weber and Rall 2012). Yet, Zhang and Chen (2022) show how, under the trading zones framework, the benefits of interdisciplinary collaboration are positively perceived by media workers.

Moreover, trading zones help journalists overcome the shortcomings of their educational backgrounds, particularly in contexts in which limited education in data handling and mining is available. That is, in a context of rapid technological transformations, in which journalists need to feed a different number of platforms and produce multimodal texts and designs, the limitations in the educational background of media professionals have been bridged by incorporating novel professional roles in journalism (Russell 2016). Not doing so could have severe consequences for legacy media and traditional journalists. As Bossio and Nelson argue, “increasing reliance on new tools has been a source of mental and physical burnout for journalists working within new online and digital work environments and competing with new media actors” (2021, 1377). The need to keep up to date with new tools for creating news stories, data analysis, and their distribution has often translated into deteriorating mental health for journalists.

Collaboration in trading zones might be necessary to overcome workforce limitations and economic conditions shortcomings in contemporary newsrooms. As Madrigal and Meyer find out, in their quest to increase their audience metrics and push toward the convergence of their platforms, media organizations have fired hundreds of journalists (2018), favoring those professionals with the data skills and knowledge required to engage with new technologies. Trading zones allow experts from different domains to make their data available and the scripts for how to work with and analyze it. As such, smaller, low-budget newsrooms do not have to dedicate unobtainable resources to exploring a database. Instead, they can benefit from the work arranged by other actors or groups, increasing their production possibilities.

As data-driven journalism shifts further away from traditional cultures within legacy media, new platforms like GitHub and Slack are emerging as alternatives for journalists, hackers, and citizens to engage in the practice of collaborative journalism (Ausserhofer et al. 2020; Bunce, Wright, and Scott 2018).

Finally, successfully using these platforms to practice cross-border collaborative journalism could have another advantage. Creating virtual spaces where journalists and a diverse range of professionals from different countries and regions interact can aid in eradicating nationalistic viewpoints and fostering a better grasp of intricate global concerns (Sambrook et al. 2018). Yet, as Kunert et al. (2022) argue, investigative journalists worldwide are adopting innovative digital practices at different speeds, with journalists in North America and Europe adopting technologies faster than those in developing countries. Cueva Chacón and Saldaña also show that, for the Latin American case, “the practice of national and transnational collaboration in the region is still scarce” (2021, 201). Whether this gap between North America and Europe and the rest of developing countries could signal the reproduction of media asymmetries and cultural dependencies between news organizations on a global scale during the practice of cross-border collaborative coding journalism is a question that remains unanswered. Hence, in this article, we deliberately refrain from presuming a priori that inter-organizational collaboration (especially a balanced one) is the prevailing norm, particularly on a platform like GitHub. Although the literature shows that newsrooms in developing countries benefit from international collaborations in general, we do not assume that they collaborate with the repositories on GitHub equally as countries in North America and Europe. Instead, this is precisely what our study is designed to investigate.

The Social Coding in GitHub

Structurally, GitHub comprises public and private repositories. Repositories are “folders” or spaces where users can save their project files and store and track versions of their codes. This affordance allows for a greater level of data structure, which simplifies the management of information. Moreover, GitHub also supports various file types for users to include in their conversations, such as JPEG, PDFs, GIF, PNG, ZIP, and other text files and videos. The diversity in the supported files allows GitHub to be more than a “code dump” and attract hobbyists and dedicated developers alike (Dabbish et al. 2012).

Furthermore, one of GitHub’s main strengths is “the awareness and transparency features it provides to team, project and community members” (Zagalsky et al. 2015, 1906), which positively affect people’s likelihood to contribute to shared projects (Tsay, Dabbish, and Herbsleb 2012). Transparency plays an essential role in GitHub community building and engagement, for it allows users to learn from the experiences of others. As Dabbish et al. put it, “being able to watch how someone else coded, what others paid attention to, and how they solved problems all supported learning better ways to code and access superior knowledge” (2012, 1283). Dabbish et al. (2012) also point out how users within the platform follow specific actions, signaling both the developer’s status and the quality of their projects. As they put it, “the number of followers a developer had was interpreted as a signal of status in the community. Developers with lots of followers were treated as local celebrities” (2012, 1281).

Undoubtedly, open-source projects permit fellow developers, scientists, and journalists to track software improvement closely. It is also a quality of the platform that aligns with the new ethos behind collaborative journalism. As Porlezza and Splendore (2016) argue, transparency and accountability in journalism, both qualities that GitHub offers, can be seen as a form of incentivizing participation for audiences, investors, or fellow reporters, an indispensable quality for emerging media outlets.

Using GitHub for collaboration also expands the possible actors involved in news production. For example, Mergel (2015) shows how GitHub has allowed fruitful cooperation between public and private communities for years. The GitHub online community can reuse, contribute, and improve the work made available by government coders, and in turn, government workers contribute to other organizations’ software development.

In a different context, Flores-Saviaga and Savage (2021) show how, in the aftermath of the earthquake that hit Mexico on the 19th of September, 2017 (known as the #19S earthquake), different users across social media came together to assist media outlets, NGOs, and universities in their effort to verify information related to the disaster using collaborative tools like GitHub and Slack. Flores-Saviaga and Savage found that GitHub allowed some volunteers to “create digital tools that seemed to bridge everyday users and technical ones” (2021). Collaborating with more experienced developers or allowing them to follow their actions and problem-solving strategies shows how this platform could be a tempting option for low-budget newsrooms worldwide.

In recent decades, GitHub has become increasingly popular as an open-source platform that functions as a trading zone for journalists, entrepreneurs, graphic designers, developers, and other actors. GitHub has been critical in popularizing collaborative coding cultures for sharing and analyzing information. This study examines how different stakeholders in journalism collaborate through GitHub by contributing and bookmarking scripts. Through a content analysis, we identify (1) the different project types developed, (2) the distinct actors across the globe that contribute to an array of repositories, (3) and the media organizations that are producing the projects and repositories. These questions stem directly from the recent literature on trading zones and GitHub and allow us to observe whether GitHub is indeed functioning as a trading zone for journalism or whether the scarcity of national and cross-national collaboration described by Cueva Chacón and Saldaña (2021) is replicated on GitHub. Notably, as Zhang and Ho (2022) argue, one of the critical characteristics of trading zones is their cross-disciplinary collaboration. Thus, it is indispensable to see the type of actors involved in this coding culture. Furthermore, this is also an opportunity to see if GitHub is bridging the gap between newsrooms across a range of countries, thus allowing the development of projects that benefit collaborative journalism, which seems to be the ethos of modern investigative journalistic projects (Larrondo-Ureta and Ferreras-Rodríguez 2021). Lastly, as Ausserhofer et al. (2020) argues, it is essential to test if the most popular journalistic repositories belong to legacy media outlets or if it has effectively opened the door for alternative organizations and actors.

Methodology

We obtained 8,686 repositories of 140 media organizations by querying GitHub’s application programming interface (API) from 2010 until 2022 (∼12 years). Due to the restrictions imposed by the GitHub API on the number of requests, it was necessary to establish a maximum limit of 100 projects per organization on each call. Nevertheless, all the projects from the media companies were thoroughly extracted. To retrieve data from GitHub’s API, username accounts are needed. We gathered the usernames by manually inspecting a list of 950 organizations in the first 50 countries outlined in the World Press Freedom Index. In Figure 1, we present an overview of the global distribution of media organizations with and without GitHub accounts. Our findings reveal 170 unique GitHub accounts from the 950 media organizations inspected, with a majority being based in the United States. It is also important to note that among the media companies, two had multiple GitHub accounts. For instance, The New York Times operated both “nytimes” and “newsdev” accounts, while Politico managed “politico” and “The-Politico” accounts.

Figure 1. Overview of media organizations with GitHub accounts across different countries.

From the subset, we extracted metadata—including elements such as descriptions, coding languages, stargazer counts, and followers—to identify the most popular ones and their projects. The complete list of media organizations inspected and the different GitHub accounts can be examined in the following link: t.ly/c4P7.

With this approach, we gathered 8,900 repositories. However, those lacking information were excluded. Thus, in our final sample, we only analyzed 8,686 repositories from 140 different media companies, including those with double GitHub accounts (i.e., The New York Times and Politico). On average, each media company maintained approximately 61 repositories. The repositories included information such as the project description, the number of stargazers, the number of forks, the programming language, and the users that contributed to the project.

Operationalization of Interactions

Before proceeding with our analysis, we must define four essential concepts for this study: stargazers, watchers, forks, and contributors. Stargazers refer to users who have bookmarked a repository, indicating engagement and potential future interest. The number of stargazers also reflects a repository’s popularity or perceived value in the GitHub community.

Watchers, on the other hand, are users who subscribe to updates for a particular repository and receive notifications about various activities, such as new issues, pull requests, or general updates. Unlike stargazing, watching a repository indicates a higher level of interest and active engagement with the project. We examine the number of watchers to measure community involvement and sustained attention toward a repository.

A “fork” is when a user creates a personal copy of a repository on GitHub. This process duplicates the entire repository, including all files, commit history, and branches, into the user’s GitHub account.

Lastly, contributions refer to a range of actions users take to participate in a repository. We consider contributions as interactions that involve active engagement with the repository, such as making commits, creating branches, opening pull requests, submitting issues, and providing comments on existing code or discussions. By evaluating contributions, we aim to understand the level of involvement and collaboration within a repository.

Users

We sample all user accounts contributing to each repository, aiming to understand the location of the interactions. We extracted 15,726 unique users, including their biographies. Drawing from users’ biographies, we coded 14,080 profiles into six groups: journalists, developers, scientists, learners, entrepreneurs, and graphic designers. These categories are selected since prior research describes them as common professional roles in contemporary newsrooms due to increased data management practices (Engebretsen, Kennedy, and Weber 2018). Users were only categorized once using string detection (i.e., exclusively categorized) based on the presence and order of keywords in their biographies (Appendix 1, available in OSF). The categorization of users was tested through a logistic regression model, obtaining .98 accuracy.

Furthermore, we identified contributors’ countries of origin from those with a location in their profile. In total, we identified 122 different locations (Table 1). In addition, Figure 2 shows an overview of the media companies in our sample globally.

Figure 2. Distribution of media companies in our sample across the globe.

Table 1. Descriptives overview.

	Concept	Frequency
Media organizations	Total media organization	140
	Countries	27
Repositories	Total repositories	8,686
	Programming languages	81
Users	Users	15,726
	Users’ countries	122

Additionally, we extracted and analyzed the description of the 8,686 projects. Following prior studies, we used an inductive-deductive analysis to identify themes in the repositories (Boumans and Trilling 2016; Grootendorst 2022; Wiedemann and Fedtke 2022). The first description analysis was done through unsupervised topic modeling using a pre-trained topic model (BERTopic) (Grootendorst 2022). We decided to use BERTopic since it considers the semantic relationship among words compared to other models, such as LDA (Grootendorst 2022). Before implementing the model, we pre-processed the descriptions in preparation for analysis. This pre-processing step involved removing URLs, symbols, and stopwords. Then, we tokenized the works and processed the text using the “all-MiniLML6-v2” MiniLM language model for BERTopic. This language model has been chosen for its exceptional performance in ensuring topic coherence for short texts. This approach allowed us to obtain an initial list of 155 topics acknowledging the context of the words. The performance of BERTopic was reflected through a topic coherence of .75.

Later, we ran several simulations estimating dimensionality and the number of topics to assess the clustering for the best topic coherence performance. The results of the simulations pointed out that 25 topics would provide the best topic coherence scoring .44. The identification of the topics was performed manually by analyzing the loadings in the topic modeling, the prevalent words that were exclusive from other topics, and the documents that were representative of each topic (Wiedemann and Fedtke 2022). We used the topics assessed by BERTopic to build a codebook. The details of the codebook can be found in the following OSF link: t.ly/c4P7.

Results

The way media organizations worldwide use GitHub as a platform for collaboration displays a great diversity of actors and a variety of topics. In this section, we described the stakeholders, how users interact with the repositories, and the themes they engage with. To understand how news organizations in different countries use GitHub as a code-sharing platform, we present the overall results in Figures 3, 4, and 5.

Figure 3. Top ten media companies with highest stargazed projects.

Figure 4. Frequency of projects by topic.

In Figure 4, the frequency is calculated through the count of categorized repositories under a specific topic.

Figure 5. Frequency of user types interacting with topics^a.

^aIn Figure 5, the frequency is calculated through a count of topic categories observed across the different types of users.

Primary Companies

A list of media companies was analyzed, including the Guardian, BBC, Seattle Times, FranceTV, NRKO, and more. The top five media companies with more repositories are located in North America and Western Europe. For example, the Guardian had 1,823 repositories, followed by the BBC with 1,008, both companies located in the United Kingdom. The companies with the most repositories were Seattle Times (n = 453) and Los Angeles Times (n = 207), both located in the United States. Finally, the fifth media company with the most repositories was Naver (n = 207), a South Korean online platform. The OSF (Open Science Framework) provides a comprehensive list of the number of repositories for the extracted media companies, which can be accessed through the following link: t.ly/c4P7 (Table 2).

Table 2. Top media companies with repositories.

Media company	Location	Number of repositories
The Guardian	United Kingdom	1,823
BBC	United Kingdom	1,008
The Seattle Times	United States	453
Los Angeles Times	United States	207
NAVER	South Korea	207
Yahoo News	United States	203
The New York Times	United States	191
San Francisco Chronicle	United States	187
Bloomberg	United States	179
ABC News	Australia	167

Moreover, we have mapped the top ten media companies with the most frequently stargazed projects (as shown in Figure 1). Yahoo News, The New York Times, and the Naver are the leading media companies with the highest number of stargazed repositories. Notably, Yahoo News reaches almost 52,939 stargazed counts for their projects, whereas The New York Times has 44,382 stargazed times and Naver 26,714. Significantly, most of these media companies (e.g., Yahoo News, The New York Times, Bloomberg, The Onion) are based in the United States. At the same time, Naver is located in South Korea, and The Guardian and the BBC are based in the United Kingdom. We included a comprehensive list of the distribution of stargazers per organization and repository in the OSF link: t.ly/c4P7.

Primary Projects

The description of the projects served to classify the repositories. In total, we categorized 4,635 projects (i.e., repositories) (Figure 2). From this categorization, we observed that 630 projects were related to the Guardian’s internal deployment of repositories: “repo-genesis.” Since the Guardian utilizes a web-based application to streamline their deployment on GitHub, it is accessible for everyone in the organization to deploy repositories and maximize their data collection efforts. The second largest repository category was data visualization, with 469 projects focused on creating graphics, plots, and interactive visualizations. Another 449 projects were related to web libraries in languages such as Python and JavaScript, while 396 projects dealt with API development and 322 with website and frontend components. An additional 313 projects focused on news topics such as climate change, elections, and crime rates. From these projects, we also analyzed the most common programming languages. Since the website and frontend components were a common topic in the repositories, the most common programming languages are JavaScript (n = 2,659), HTML (n = 975), and Python (n = 818).

In addition, we examine the distribution of repositories across our sample. Most projects are developed in the United States (26.9%, n = 1,963) and the United Kingdom (38.6%, n = 2,821). In this regard, all the projects’ topics are covered in these countries (the United States and the United Kingdom), including emerging themes such as machine learning and the development of conversational bots. Interestingly, countries like Uruguay or Lithuania do not seem to have repositories under those categories.

Primary Actors

We extracted 15,726 unique actors. We categorize the users using actors’ biographies. We coded 14,080 profiles into five groups: journalists, developers, scientists, entrepreneurs, and graphic designers. This approach resulted in 3,331 journalists, 8,577 developers, 327 scientists, 1,689 entrepreneurs, and 156 graphic designers from 15,726 users. The remaining users (1,646) were only considered for an overview of contributions since the descriptions of their bios were not enough to categorize them. From this sample, we estimated the percentage of contributions by the type of user. We observed that from 21,377 contributions, 43% of the contributions were done by developers (n = 9,192), 11.66% (n = 2,493) by entrepreneurs, 7.33% (n = 1,569) by journalists, 3.48% (n = 744) were made by scientists, and 0.5% (n = 109) graphic designers. In our sample, we observed that, on average, each media company had 101 contributors. However, on average, users only contributed .019 to the different repositories across the media companies. The total number of contributors per media company can be found in our OSF: t.ly/c4P7.

Projects and Collaborations

Analyzing user interaction types, we see that developers are the primary contributors to all repository themes (e.g., web libraries, website and frontend components, and data visualizations). Journalists contribute mainly to data visualization projects or news coverage topics. Surprisingly, although data visualization projects are common, graphic designers are not the main contributors to this topic. Developers, entrepreneurs, scientists, and journalists interact more with data visualization repositories than graphic designers (i.e., DataViz). In contrast, journalists are subscribers of the Guardian’s public repository development, using the newsroom template to create their projects (Figure 3). Moreover, since GitHub is also used for learning, Figure 3 also shows that contributions to documentation and learning projects are usually made by journalists and entrepreneurs, in addition to developers.

By exploring the contributions of journalists for each country (Figure 6), we see that journalists mainly contribute to projects in the United States (US) (n = 58), the United Kingdom (n = 26), Australia (n = 10), Canada (n = 10) and New Zealand (n = 10). Interestingly, journalists mainly subscribe to projects in the United States (n = 41) and the United Kingdom (UK) (n = 20). Nevertheless, on average, users contribute only once to projects. Multiple interactions with the same repositories are not frequent.

Figure 6. Frequency of journalist interactions with media companies’ location^a.

^aIn Figure 6, the frequency is calculated through the count of observed interaction types across user types and location.

Furthermore, an analysis of contributions by country of origin is presented in Figure 7. The top ten contributors include developers from the United States (n = 888), United Kingdom (n = 386), Germany (n = 209), France (n = 141), and Canada (n = 103), as well as Sweden (n = 79) and the Netherlands (n = 76). Additionally, contributions from entrepreneurs in the United States (n = 292) and the United Kingdom (n = 107) and scientists in the United States (n = 115) are also among the top contributors. It is important to emphasize that while GitHub does offer a platform for collaboration across news organizations, the data indicates that contributions to repositories typically originate from specific users within each organization. Despite the potential for collaboration, users predominantly focus on their respective news organizations and may not contribute to other organizations as frequently as initially anticipated, as illustrated in Figure 8.

Figure 7. Top ten frequencies of contributions by users’ location^a.

^aIn Figure 7, the frequency is calculated through the count of observed contributors (a specific type of interaction) across user types and locations.

Figure 8. Top ten contributors to repositories in different media companies.

Furthermore, our observations reveal that most users are employed by the news organization with which they interact. This finding underscores the close association between users and their affiliated news organizations, likely influencing their contribution patterns. For a more comprehensive understanding of the relationship between news organizations and user affiliation, we included a dataset with detailed information on the OSF link: t.ly/c4P7.

Discussion

As journalists increasingly rely on and report about digital data, they rely on trading zones to find collaborators to analyze, interpret, and display the data, helping to alleviate the burden of information asymmetry (Gray, Chambers, and Bounegru 2012). Recent literature shows that a new collaborative coding culture has emerged from these practices, marked by cross-disciplinary cooperation between media professionals, transparency in using codes and scripts, and the accessibility of large data sets. However, our results complicate the idea of collaboration in coding cultures by showing different degrees of reciprocity or cooperation in the information flow on GitHub’s repositories.

This article aimed to study how media organizations across a range of countries use GitHub as a platform for collaboration. Focusing on GitHub was not an accident. Previous studies have shown the potential behind this platform to promote cooperation between actors with different professional backgrounds and skills (Zagalsky et al. 2015). Our results suggest that GitHub is playing the role of trading zone for journalistic work. Professionals with different backgrounds and skill sets come to this platform to engage in the production and consumption of coding materials. As we have described, developers are the primary actors in all the different projects contributing to GitHub repositories owned by media corporations. In line with previous literature (Brossi and Dodds 2019; Lewis and Usher 2014), we see a diversification of professional roles in news development. Developers as the main contributors might result from more familiarity with the platform, which coders and data scientists primarily use.

Aligned with these findings, users mainly contribute to the same news organizations they are employed by. In this sense, users’ familiarity with their respective organization’s inner infrastructure plays an essential role in driving their interactions with the repositories. Their familiarity likely fosters a sense of ease and efficiency in contributing to projects within their organization. As a result, although GitHub can function as a trading zone platform by bringing together different types of professionals, collaboration across different organizations may be less common due to users’ strong association with their employing organizations.

These results put tension on the ‘collaborative’ part of coding cultures. These findings suggest limited cross-organizational collaboration, particularly between countries of different geographical regions. Instead, we see a notable trend toward intra-collaboration, with journalists predominantly engaging in collaborative efforts with their own organizations and repositories.

Our results show that GitHub, a platform that has experienced exponential growth since its foundation in 2008, is not alleviating the burden of smaller low-budget newsrooms that cannot gather or analyze large data sets and bridge journalists’ work and developers’ skills. However, the directionality of information flow seems to benefit smaller low-budget newsrooms, as they consume repositories more than they collaborate with them. To test how news organizations use GitHub, we extracted metadata from news organizations’ GitHub accounts (e.g., descriptions, coding languages, stargazer counts, followers). In addition, we identified the most popular media organizations based on the stargazed counts of their projects and sampled the users’ accounts contributing to them.

Regarding our main research question, namely, how news organizations use the code-sharing platform GitHub, we found that those organizations in the United States host the projects with the highest stargaze counts. We found that media outlets in North America, Western Europe, and South Korea have the highest number of repositories and interactions. Although media organizations assisting other media organizations in news production is a positive outcome, future research needs to study if the reliance on newsrooms in North America and Western Europe’s repositories eventually translates into new self-owned projects for newsrooms in other countries around the world or this remains a dependent relationship between one and the other.

Our results showed that GitHub mainly fulfills reporters’ need for the graphical display of big data sets (Gray, Chambers, and Bounegru 2012). Following recent literature, the collaboration emerges as the need to adapt various journalistic content to the latest platforms utilized by media organizations for news dissemination (Royal and Kiesow 2021). Surprisingly, one of the main themes was how The Guardian facilitates the deployment of repositories among the organization to maximize the collected data points. Interestingly, this finding points out the need for analyzing and reporting the abundance of public interest data. As Zhang and Ho (2022) asserted, the trading zones show an interconnection between physical and digital settings. In this scenario, news organizations require not only a collaborative coding culture to analyze the growing number of data points (Gray, Chambers, and Bounegru 2012) but also to produce additional information that aligns with the organization’s primary objective of keeping the public informed (Hellmueller, Cheema, and Zhang 2017).

Future research should explore the relationship between media cooperation size, turnover, and their ability to contribute to collaborative data journalism. Our data indicates that many media companies among the top contributors to GitHub are large, successful corporations. However, in the context of media market concentration and “winner takes all” dynamics that have characterized the news media market in recent years, it remains an open question whether sharing the GitHub repositories serves as an equalizer or is, in reality, more likely to be a transparency tool that is costly to create and therefore further contributes to the economy of scale, thus perpetuating media asymmetries and cultural dependencies between news organizations on a global scale.

Finally, our results suggest that trading zones promote the necessary diversification of professional roles inside newsrooms, bridging the gap in educational backgrounds and helping small-budget newsrooms overcome workforce limitations and shortcomings caused by economic conditions. Perhaps these financial shortcomings help explain the imbalance in the “collaborative” part of coding cultures. As we posit at the beginning of this article, trading zones are both virtual and physical spaces, and the limitations of one (e.g., limited technological infrastructure in a low-budget newsroom) could affect their ability to produce coding repositories. Yet, collaborative (coding) cultures in journalism demand not an equal amount of supply but rather a nuanced understanding of how diverse skill sets, perspectives, and local expertise can, in a symbiotic relationship, benefit those proficient in coding and those who are still learning.

Conclusion

Based on our data, our results suggest that major English-language media outlets mainly produce emerging journalistic trading zones on the GitHub platform. Therefore, our results reveal an ongoing asymmetrical relationship between news organizations dominated by well-resourced media giants in North America and Western Europe. However, our data also show a great deal of diversity—the act of sharing digital resources, scripts, data, and learning content that connects a variety of contributors from newsrooms worldwide. Furthermore, media organizations in developing countries interact with the repositories of media companies in one way or another—to some extent, they speak the same language and share specific interests. From this perspective, the idea of a ‘global network of journalism’ seems to be reinforced (Berglez and Gearing 2018).

Future studies could start here and thus overcome one of the significant limitations of this study: Our initial media sample already includes most countries from North America and Western Europe. The focus on the ‘most robust’ media systems—the highest-ranking countries in the Press Freedom Index—seems plausible if the objective is to sample those countries with particularly active participants in the trading zone around media technology. Though this was, in fact, one of the aims of this study, it also meant that countries in Central and South America, Africa, and other regions of the world had been underrepresented from the very beginning. Moreover, the statements made here are based on a small and heterogeneous sample – raising the question of whether developing countries will reflect more diversity in the interaction between newsroom organizations worldwide. In any case, future studies should choose a more balanced country sample or one with the opposite constellation, for example, with the US and UK as the only representatives of North America and Western Europe. Since language, in particular, plays an essential role in adopting technologies (e.g., conversational bots), the digital divide often runs along the gap between low-resource and high-resource languages; it would be interesting to study English-speaking countries outside North America and Western Europe.

Furthermore, the analysis of topics on GitHub reveals a strong presence of the Guardian deployment system without offering insights into the range of topics the organization covers. Future research should broaden the representation of topics by employing the readme files as the unit of analysis.

Another limitation of the initial sample is that it consists exclusively of the GitHub accounts of traditional media companies. It is likely—and our data suggested so—that specific networks of actors that address media technology are not visible here. Future studies should focus on institutions beyond traditional mainstream journalism, such as smaller online-only media, news aggregators, research collectives, NGOs, individual reporters’ accounts, etc.

Qualitative studies could lay a foundation for this selection, identifying key actors and collaboration structures via interviews. A mixed-method approach could, for example, be used to identify critical stakeholders as potential interviewees on GitHub and to contact them via the platform. This would also answer whether GitHub is the primary platform for exchanging digital resources in all countries or whether trading zones can also be found elsewhere.

At the same time, our results suggest that the trading zone we considered also attracts actors whose connection with journalism is marginal at best. On the GitHub platform, journalistic interests overlap and interact with commercial and scientific interests. Future research could investigate how specific resources (e.g., individual scripts) transcend the boundaries of their original context and are recycled in other environments or adapted to different system logic.

Our study not only advances the current understanding of journalistic trading zones but also provides a roadmap for future research directions that can contribute to a more comprehensive and inclusive exploration of media technology landscapes globally. The intricate interplay of diverse actors and interests on GitHub underscores the evolving nature of journalism in the digital age, urging scholars to explore beyond traditional boundaries, contributing to a better understanding of the cross-pollination of ideas and practices within and beyond the journalistic sphere.

Disclosure Statement

No potential conflict of interest was reported by the author(s).