The “Digital Turn” in Transitional Justice Research: Evaluating Image and Text as Data in the Western Balkans

Introduction

Technological advances have changed not only how people engage with transitional justice (TJ) processes but also how scholars can study them. New platforms for social interactions and enhanced availability of the data produced by political and TJ institutions, such as national parliaments, or court trials and truth commissions, on the one hand, as well as the general public’s engagement through online platforms, such as Twitter or Facebook, permit analysis at an unprecedented level of precision and granularity. The transformation in the way we communicate and interact online and how the ever-growing flood of digital media affects our perceptions of the world around us and changes our behaviours has been dubbed the “digital turn” (Westera 2012). More accessible computing resources coupled with ever-improving algorithms and methods make possible new insights into the fabric of TJ mechanisms and their effects. At the same time, the development and application of new analytical procedures hinges upon a deeper understanding of the structure of a broad range of data sources that are used in TJ research.

The proliferation of new modes of data (chiefly, text and audiovisual) in TJ reflects a wider and notable development in the social sciences more broadly. Release of court records, digitisation of archives, wide availability of news articles in a machine-readable format, the proliferation of images and video footage, detailed minutes taken by civil society TJ initiatives and the spread of social media have made a treasure trove of new data available to TJ scholars.^[1]

This trend, combined with advances in local and, more importantly, cloud computing and social science methodology has made possible analysis on a new scale of both qualitative and quantitative datasets. While the broader debate revolves around the role of information and algorithms in generating new insights (King 2016, vii–x; Mayer-Schönberger and Cukier 2013) in contexts laden with recent adversarial history and politics, data cannot be assumed to produce unbiased results as ever-larger quantities of it are collected. TJ literature would therefore benefit from a more thorough discussion of the promises and challenges of new data. Integral to this discussion are ethical dilemmas involved in working with this data. New forms of data amplify “old” ethical questions, as well as raise new ones. Their consideration also directly impacts “ethical” modes of analysis, which in turn may also affect the kind of knowledge about TJ that is produced.

In this article, we contribute an opening gambit to this discussion by considering the opportunities and challenges for TJ scholarship and practice presented by data and analytic tools made available by new digital technologies in various domains. We draw on research undertaken in a major interdisciplinary research project involving visual arts, the humanities and social and political science, “Art and Reconciliation: Conflict, Culture and Community” (2016–2019), involving digital archives, visual images, transcripts of trials and hearings and deliberations around TJ and social media (Facebook). We consider how these new datasets and the use of new analytical tools open up novel ways of thinking and knowing about TJ, by grounding theory in practice-based understandings and in data generated by TJ processes themselves. In this article, we address opportunities and dilemmas that are presented to scholars and that derive from the digital nature of this new data, contributing to the emerging discussions about technology and TJ. These discussions have focused on the role of technology in enhancing the bases of claims to justice by affected populations, as well as in being a source of insecurity to those who seek justice (Pham and Aronson 2019, 1–6). However, contradictory implications of technology from the perspective of methodological advancement and the questions of research ethics, as well as their relation to knowledge creation in TJ—to which we turn—are less understood.

Data and Knowledge in Transitional Justice

Advocates and critics of TJ have long debated the relative merits and drawbacks of a range of approaches to dealing with past atrocity and the relative merits and drawbacks of different methods for evaluation and measures of “success”. In 2008, the Centre for International Policy Studies at the University of Ottawa conducted a comprehensive review of the field and found that there was insufficient evidence to support any of the claims made about possible negative or positive impacts of TJ (Thoms, Ron, and Paris 2008), echoing the earlier lament by Fletcher and Weinstein (2002, 585) that the “primary weakness” of TJ research was the “paucity of empirical evidence to substantiate claims”. The TJ field responded with an array of empirical studies, but a less than satisfactory consensus on the meaning of all of this for theory, policy and practice. The “primary weakness” now is not the lack of evidence, but the tendency to use the available evidence to make judgements about the success or failure of different initiatives based on a preconceived set of criteria or expectations, and a preordained preference for particular methodologies (Kerr 2017, 125–6; Shapiro 2015, 13–25). Measures of success and failure have largely been determined by theoretical assumptions about what TJ is for, and neglects to ask for whom or how they function. Goals are either reductionist—seeking to measure what is “measurable” such as the number of new human rights instruments enacted, or expansionist, questioning whether TJ ought to be expected to fulfil a transformative agenda of socio-economic and political justice or to contribute to the fulfilment of key development goals. Moreover, in the past few years, tensions have arisen between those of a more quantitative orientation versus those who favour qualitative methodologies.^[2] More evidence through the production of new data may not resolve these tensions if the problem continues to be framed in normative terms. This article does not seek to intervene in that debate so much as to move it forward by focusing on the data first. New datasets and innovative methods for analysis have the potential to create exciting opportunities to produce robust evidence that will move the field of TJ research forward in bounds.

After all, we know that processes of TJ, and their effects, are never neat and straightforward. Nor should the enterprise be viewed in terms of a trajectory of progress for “good”. Rather, it is a complex and messy process involving a disparate range of political, social, emotional and psychological factors, all of which operate at different levels—individual, familial, group, societal and state. Moreover, pinning meanings on highly contested concepts of justice, peace and reconciliation is difficult, embedded as they are in particular social, cultural and political contexts. We might better address these aspects if we conceive of TJ as a particularised social and political process and study its enactment rather than its instrumentality. What can the data tell us about the mechanics of TJ? How is TJ enacted and articulated by those involved in the process? Can we ground knowledge of TJ in the data created in and around the processes themselves? New digital technologies offer huge opportunities in this regard, allowing us both to access rich new datasets created and curated in digital format, including archives, transcripts, discourses and images, thus expanding the empirical evidence available exponentially, as well as offering opportunities for new insights into their mechanics, while being rigorous about how data is produced and curated in these sites.

While appraising the potential of data to open up new research prospects in the context of the “digital turn” in TJ, the issue of research ethics looms large. TJ research which deals with deeply personal and distressing issues of pain and suffering has spurred the thinking on research ethics. In particular, the focus has been on how to conduct research while mitigating harm to participants. More recently, scholars have discussed harm to researchers conducting work on these sensitive issues (Kostovicova and Knott 2022). Ethical considerations in TJ are commonly derived from research methods, often conducted in the field, and that requires close contact with victims of war crimes and human rights violations (Clark 2017, 424–39). Meanwhile, ethical issues regarding data are reduced to the issue of data security and anonymisation. Data in digital formats draws attention to ethical concerns that stem both from the format of the data and its means of production, as well as the analytic strategies employed.

We focus our discussion on TJ processes in the Western Balkans, which here refers to the post-Yugoslav space. The region is an appropriate case study for the “digital turn” for three reasons. First, it involved conflicts along ethnic lines, which allows for comparison with other similar conflicts around the world. Second, multiple TJ mechanisms were implemented in the region, most of which with limited success. This included an international criminal tribunal, domestic war crimes trials and various civil society initiatives. These have all generated vast amounts of data for researchers to examine. Third, the region has been extensively researched but with limited exploration of the potential of (large volumes of) text and images as data.^[3] The post-Yugoslav states in the Balkans thus provide a typical case of ethnic conflict that resulted in a variety of TJ responses, which generated vast amounts of data and where the “digital turn” offers an opportunity for significant new insights. The remainder of this article discusses different types of data in a variety of domains and how we might creatively use the “digital turn” to help us better to “see” what is there and use it to leverage more nuanced and refined insights on the complex political and social phenomenon that is TJ.

Data Archives

Archives are a rich source of data for TJ scholars. Yet, they often are underutilised, perhaps both due to TJ’s focus on the recent past and because TJ remains dominated by legal scholars and political or social scientists, who have traditionally preferred to use other sources of data—surveys, interviews, focus groups—for their analyses. The value of the archive for TJ can be seen in (at least) three regards. First, archives are part of the TJ process. This was shown in the Latin American transitions, where archives often helped support nascent TJ interventions (Collins 2010). Going beyond helping other TJ mechanisms function, archives have in some contexts played an even more central role in transition. In East Germany, for example, opening access to the Stasi archives was presented as a substitute for otherwise absent TJ intervention after reunification (Bruce 2009). Second, archives are of historical value when they are produced as a result of TJ. A detailed analysis of these documents can capture, for example, the chain of command in the perpetration of mass atrocity. In this regard, Christopher Browning’s ground-breaking account of Nazi atrocities, Ordinary Men, relied heavily on the Nuremberg archives (Browning 2001; Hilberg 1985). The use of trial records in historical research also challenges the, now largely accepted, position that war crimes trials tend to produce a poor reflection of the lived experience of war (e.g. Dembour and Haslam 2004, 151–77; for contrast, see Redwood 2021). Indeed, TJ archives should be seen as sites of memory, a monument to past acts of atrocity. Their very existence can indicate that some form of TJ had, in fact, taken place.

This is not to say that these records are untainted recordings of the past. This view of archives as neutral sites of knowledge production has, for some time now, been replaced with an understanding of the archive as inherently political. Like all archives, the information stored in TJ archives remains strongly framed by the priorities of the institutions that produced them (Humphrey 2003, 176). The structured and constrained nature of data is, however, true of all forms of data, and acknowledging this opens up the archive to a third use. As Stoler’s (2002, 87–109) work on colonial archives shows, this directs the researcher to explore why the data exists in a particular way and what this can tell us about the institutions that produced this data and the broader structures they operate within. This should lead researchers to ask: from which perspective is knowledge produced? Who is excluded? Which crimes are captured, and which are ignored? (Redwood 2020, 271–96). And what does this tell us about the functioning and priorities of the institution or intervention that produced the data, and more directly, about the unfolding TJ process we want to examine?

If this is the value of archives for TJ research, then the digitisation of archives on the surface seems to enhance these possibilities. New technologies greatly aid protection against data decay—as many old archival documents are crumbling, photographs fading, letters decomposing—digitisation of archival material has become routine practice. If these archives are seen as sites of memory, then digitisation not only helps to preserve data but, in the context of TJ, can help preserve memory itself.

Digitalisation also allows for access to an unprecedented amount of data. Even a few years ago, access to many archival records was made considerably more difficult because of either the lack or poor functionality of digitised and online archives. However, recently, for example, records at the International Criminal Tribunals (ICTs) for the former Yugoslavia and Rwanda have been almost completely digitised, with the introduction of a new archive interface (UNMICT, Judicial Records and Archives Database). The combination of innovation in building digital infrastructures and new methods such as graph databases provides more advanced means of access to facts in the documents (Blanke and Kristel 2013, 41–57). The new ICTs archive interfaces, for instance, now allow researchers to search across case files enhancing their historical value. Using powerful search engines also helps overcome the problem of data dispersal, where, as demonstrated in Holocaust research, information about one person may be spread across many databases in different parts of the globe (Nikolova et al. 2018). As we explore in subsequent sections, the digital revolution, moreover, opens archiving to, arguably, more democratic and socially progressive possibilities as data is produced in a more collaborative and consensual way. Technology has, for instance, helped in trying to sensitively deal with the curation and access to records pertaining to indigenous populations (Christen 2011, 185–210). All of these advancements directly contribute to the reach and robustness of TJ research. For example, scholars were able to use the newly digitised ICTY archives to systematically study the behaviour of perpetrators of violence (Vukušić 2021, 66–86), to uncover the often-hidden cases of sexual violence against men (Drumond 2019, 1271–87), and even to interrogate the emotional entanglements and ethics of research on violence in the Western Balkans (Biddolph 2021, 530–55).

Digitisation is not, however, value-neutral, nor is it the panacea for problems faced by archivists and archive users. First, whilst on the surface digitisation seemingly aids preservation, the extreme volume of data to be digitised and stored is logistically difficult, and the proper maintenance of digital archives is costly and requires a new kind of technical expertise with which many archives may not be fully equipped. The regularity with which systems change means that storage and access solutions are in danger of quickly becoming obsolete, which means that some files, if not updated regularly, might become inaccessible.

The second issue is the perception of the totality of the archive. Many digital archives—particularly those linked to TJ mechanisms—contain so much information that the problem becomes one of “too much data”. And yet, these archives remain as partial, selective, and therefore politicised as other archives (Tesar 2015, 101–14). With the digital revolution often occurring long after the creation of an archive, the digital archives often remain only a partial reflection of their physical counterpart.

The digitalisation of archives, further, brings about a new set of ethical concerns. Scholars of the Holocaust, for example, have pointed out the potential of further dehumanising and depersonalising Holocaust victims by anonymising, numbering and classifying their stories (Einwohner 2011, 415–30), a process that—even if unintentionally—will be amplified by digitalisation. With increased access, moreover, greater concern must be paid to the issue of consent, especially when dealing with the digitalisation of private archival documents (Subotić 2021, 342–54).

Whilst the “digital turn” creates a number of possibilities to enrich TJ research; it can also be seen as amplifying difficulties that come with analogue archives. As the identification of novel archiving practices in relation to indigenous archives suggests, however, there is considerable scope for the digital revolution to force a more outright challenge of what it means to archive, which could have a significant impact on each of the uses of the archive identified above. This could help feed in a wider variety of voices into the TJ process, enable a richer and deeper understanding of the historical past and help democratise the way knowledge is produced and accessed.

Court and Truth Commission Transcripts as Text Data

Digital TJ archives are often dominated by transcripts from the court proceedings and Truth and Reconciliation Commission (TRC) hearings. Many of the potential uses of these transcripts align with the previous discussion of the value of archives: as a source of historical knowledge, they contain insights into the workings of TJ processes and thereby provide a means to analyse how TJ interventions produce specific articulations as to their purpose and function. Considering the multifaceted value of the transcripts, the following section focuses on some of the knowledge gains and ethical dilemmas of deploying court and TRC transcripts as text data and methods, with particular attention to those that are made possible by the “digital turn” and development of computing power.

Availability of court and TRC transcripts has spurred methodological innovation and pluralism in the analysis of this data. They allow a window into the past, and, almost paradoxically, the skewed nature of this window can be read from the data to reveal the priorities and principles underpinning the TJ mechanisms. Analysing transcripts in this way offers a way into exploring power relations within the courts and truth commissions but also helps us ask fundamental questions about who and what these TJ practices are for (Redwood 2021).

A number of scholars employing qualitative social science methods have drawn on ethnographic approaches to examine courtroom practices (Eltringham 2013, 342), while sociolinguistic approaches have explored inner dynamics of the courtroom and, in turn, the institutions that these processes function within (for an exception see Perrin 2016). This and other types of qualitative textual analysis have offered insights into how macro-level discourses (such as class, ethnicity and gender) influence both how participants in these processes speak and why certain acts of testimony are believed and others rejected (Conley, O’Barr, and Riner 2005; Erickson et al. 1978, 266–79). Fine-grained analyses of court and TRC transcripts have also provided insight into silences, which, in turn, points to a marginalisation of specific harms, such as wartime sexual violence and of voices of marginalised groups.

On the quantitative side, manual quantitative content analysis of TRC transcripts has revealed variation, such as how the race of survivors or perpetrators impacts engagement with justice (Chapman and van der Merwe 2008). More recently, quantitative text analysis methods have offered a solution to what is a major issue with transcripts, whether of war crimes trials or of TRCs: the challenge of “too much data”. For example, the transcripts of civil society debates in the Western Balkans about a regional fact-finding commission (Regional Commission for the Establishment of Facts about War Crimes and Other Serious Violations of Human Rights Committed in the Former Yugoslavia from 1 January 1991 until 31 December 2001, RECOM) amount to around four million words, roughly the equivalent of eight volumes of Leo Tolstoy’s War and Peace (RECOM Reconciliation Network 2021).^[4] “Too much data” is not an unfamiliar challenge to political scientists. They have developed sophisticated quantitative text analysis (QTA) methods (Grimmer and Stewart 2013) and applied them to large volumes of textual data, such as parliamentary debates, public consultations, party manifestos, newspaper reports (as well as blogs, Twitter, and online comments) (Laver, Benoit, and Garry 2003, 311–31; Mueller and Rauh 2018, 358–75; Proksch and Slapin 2010, 587–611).

QTA is well suited to uncovering patterns in textual data and capturing its prominent features, and as such, it has allowed scholars to begin to uncover the effects of TJ initiatives in new ways. Emerging research applying the “text as data” approach in TJ, as in the case of applying the dictionary method to study reconciliation in interethnic civil society debates about justice for war crimes in the Western Balkans, has demonstrated the value of a regional approach to TJ (Kostovicova 2017, 154–75; Kostovicova and Bicquelet 2018). Combining computer-assisted quantitative text analysis (including the topic analysis) and human-coded content analysis has revealed how women are silenced in speaking sequences during TJ debates in the Western Balkans (Kostovicova and Paskhalis 2021, 263–76). At the same time, the application of the word embedding method in the study of the digitised transcripts of the public hearings of the South African Truth and Reconciliation Commission has identified different understandings of reconciliation by victims and perpetrators (Alonso 2019). The field of TJ has yet to reap theoretical benefits by applying a full range of quantitative text analysis methods to newly available digitised text data. Their application holds out the prospect of providing new insights into the content of TJ processes and their effects, and advancing our understanding of the expressive purpose of transitional justice (Aloisi and Meernik 2017; cf. Chlevickaitė, Holá, and Bijleveld 2021). Approaching TJ data as big data stands to enhance the evidence base relying on qualitative discursive analytic approaches (such as critical discourse analysis, as well as narrative, thematic or frames analysis). New insights can reveal the features of the content and their variation based on a range of TJ actors’ attributes (such as gender, ethnicity, race, victim vs perpetrator, witness vs defendant, state vs non-state actor, rural vs urban, etc.) in a range of different TJ mechanisms. In combination with social network analysis and sentiment analysis, quantitative text analysis can reveal novel manifestations of interactivity in TJ processes by identifying links between TJ actors or themes (Kostovicova and Paskhalis 2021; Kostovicova, Sokolić, and Paskhalis 2019; Vinck 2019).

TJ scholars can now take analytical advantage of the development of computing power with ever more sophisticated software packages and language processing tools (Evans et al. 2007, 1012). The point about language is particularly pertinent for TJ scholars, who work with a wide range of languages. Thanks to technological advances, even if the researcher does not restrict the corpus to English language sources, there are options that she might pursue. For example, the emergence of crowdsourcing platforms, such as Mechanical Turk (Berinsky, Huber, and Lenz 2012, 351–68) and Figure Eight, allow us to scale coding of a vast volume of text by relying on a large number of Internet workers coding small chunks of text in English, but also in smaller European languages (Benoit et al. 2016, 278–95). Furthermore, the advances in machine translation mean that large quantities of text can be processed at a relatively affordable price. At the same time, translation makes possible the use of analytical tools that are still only available for the English language. The development of social science methods that can be used to analyse vast quantities of data in TJ research, including those bespoke for textual data, will also help practitioners of TJ who have been faced with an “information overload” (Gavshon and Gorur 2019, 71–91).

Overall, the digital availability of court and TRC transcripts and recent innovations in natural language processing and statistical modelling of textual data herald major gains for enhancing evidentiary bases for TJ claims. These advances ought to be attentive to ethical concerns which derive both from the type of data and how methods are applied to analyse it. Transcripts are often redacted to protect witnesses from having sensitive data revealed. This process is far from perfect, and so researchers have an obligation not to reproduce information if it might lead to identifying particular persons and, where these breaches are identified, notify the institution responsible for maintaining the records. Text mining methods, with their “bag-of-words” approach, where a word is a unit of analysis, largely mitigate against harms related to the identification of individuals directly or indirectly through their specific personal experiences. But, this hardly makes the QTA method ethics-free. Looking forward, the release of ever more data in a digital form, which holds the potential of research advances in TJ through text mining methods, also raises a delicate “old” ethical dilemma of consent for secondary use and analysis of this data (van Wel and Royakkers 2004, 129–40). Lastly, making the court transcripts digitally available may also be ethically dubious if this is framed in political terms as serving “the reconciliatory purpose”, that even the trials have not been able to achieve, as is the case with the ICTY (Kaye 2014, 381–96).

Facebook Data

The “digital turn” has increased the use of Facebook and other social media platforms to discuss TJ topics and thus opened new opportunities for research (cf. Vinck 2019, 105–12). Social media platforms have made digital data ubiquitous and versatile. Facebook data, as one type of data generated in communication on social media, has resulted in the expansion of empirical evidence available to study TJ by increasing its geographical and longitudinal scope. This data allows us to learn more about the diversity of voices in the process and examine how TJ is received and examined by different publics. This section examines Facebook data that is produced and data that Facebook collects, which can be gathered without researcher intervention. This new type of data in TJ also comes with a host of new ethical challenges.

The range of Facebook users is immensely heterogeneous, and Facebook provides several tools to easily, and with little cost, recruit large, diverse and representative samples through various techniques (see, for example, Kosinski et al. 2015, 543–56; Zhang et al. 2018, 558–64). It reflects the range of voices in TJ, which many appraisals miss. This type of data can, therefore, be obtained from targeted populations, such as young people, those in rural locations, members of the LGBT community, women, stigmatised groups, and specific ethnic minorities (Lijadi and van Schalkwyk 2015, 8; Mangan and Reips 2007, 233–36). Their perspectives are routinely overlooked in the TJ literature due to financial, security and accessibility constraints. As shown by emerging research involving youth from different ethnic groups in Bosnia and Herzegovina, Facebook allows researchers to bypass ethnic segregation and involve people of different ethnicities in conversation (Sokolić 2019). Moreover, Facebook’s ubiquity enables the collection of global data, thereby allowing for an exploration of TJ processes across cultures and large numbers of cases, which the field has struggled with.^[5] Facebook’s reach highlights that TJ, viewed as a process, has different meanings to different people, and its reception and contestation are varied.

Facebook data can be produced, for example, through interviews and group discussions, where Facebook is used as the method of communication between the researcher and participants. This data does not entail a significant change in existing research frameworks since it is largely comparable to non-digital types of data.^[6] Facebook data, however, also involves different modes of communication, such as photos, videos, songs and links to websites, which allow for a combination of textual and visual data analysis. The data can be approached and analysed qualitatively or quantitatively, and the barriers to entry are minimal; simply being able to use Facebook is enough. The scope to expand data collection is extensive, including developing apps that only require a click from a user to donate data to researchers (Kosinski et al. 2015, 543).

Facebook also allows for access to existing data about user behaviour and preferences without researcher intervention (Wilson, Gosling, and Graham 2012, 547). The full range of data recorded is constantly evolving, but it includes demographics, user-generated content, social network structure, user preferences and activities and information about friends (Kosinski et al. 2015, 547). The reliability of Facebook data has been questioned because it allows participants “free reign to construct a public or semi-public image of him- or herself” (Wilson, Gosling, and Graham 2012, 213). This, however, provides a further analytical avenue for the exploration of how individuals construct their identities in the postconflict context. It can help us better understand how identities are presented and manipulated when individuals engage with TJ. Behaviour and preferences on Facebook are a proxy for real-world behaviour (Kosinski et al. 2015, 548).

This type of Facebook data is also available retrospectively (if consent is provided), and users can be tracked across studies and time. In the postconflict and transitional contexts, this is unique. It can help us analyse how the process is received by intended beneficiaries in different geographical locations, but also over time. It means that researchers can access real-world data that was recorded on Facebook before a conflict started or ended, depending on the research purposes. These insights can help expand and better define the scope of transition in TJ, an often-debated topic. This data is potentially less biased than self-reported data, imbued with distorted and selective memories.

However, Facebook data is limited in terms of representativeness. Despite the platform’s ubiquity, the data is not perfectly representative since it requires internet access and a device to access it on. The platform has also been or is currently blocked by countries on political or religious grounds, including China, North Korea, Iran, and Syria, among others. Facebook is, therefore, of limited use in certain TJ contexts. Finally, the data’s comparability is dependent on Facebook’s design choices. For example, “likes” were only introduced in 2009 and were preceded by free-text fields (Kosinski et al. 2015, 549).

Facebook presents both opportunities and challenges in terms of ethics. TJ research can be exploitative; researchers often access research sites for brief periods of time, ask participants to revisit past traumas and have little follow-up contact. Social media platforms allow for a more ethically considerate research relationship with participants since they permit follow-up, even continuous, contact with participants. Facebook data, nevertheless, introduces novel ethical challenges to researchers, many of which are yet to be properly addressed. The key ethical issue relating to Facebook data is informed consent. The readily available data on Facebook has been grossly misused by overreaching researchers; emblematic of this is the 2014 Facebook experiment on 700,000 users’ news feeds and the scandal surrounding Cambridge Analytica (where data was passed on to a third party for “research purposes”) (Arthur 2014; Cadwalladr and Graham-Harrison 2018). Facebook has consequently restricted access to its data significantly (see Schroepfer 2018). Even produced data can suffer from this problem since researchers may not be aware of how participant data will be used in the future. At best, researchers may be able to provide certain conditions under which it will be used, but they may not be able to reacquire consent if the data is reused (for a discussion of this, see Mackenzie 2018). Moreover, the lines between private and public information have become blurred in online environments, and users seem exceptionally willing to share deeply personal information with both friends and researchers, whereas they may otherwise choose not to in a face-to-face situation (Wilson, Gosling, and Graham 2012, 204). Users may not be aware that they have consented to information being public and accessible to researchers. Alternatively, in the process of data production, they may not be aware of whom they are sharing the information with (for example, in a group discussion). Consequently, it is the researcher’s duty to ascertain the risks and benefits of research, which can be particularly complex in postconflict environments, as well as to make participants aware of these novel risks associated with Facebook data.

Image as Data

For TJ scholars, images hold a range of possibilities to incorporate new voices and data sets that expand how processes of TJ are understood from, as of yet neglected, perspectives (Bleiker 2018). Visual research methods offer opportunities for TJ researchers to engage with the lived experiences and opinions of postconflict affected communities and to root their work within locally owned transitional processes.

Visual research opportunities have been amplified by the “digital turn”. Digital technology has made image production, storing and sharing affordable and accessible, increasing the central role of images in social communications and memory and meaning-making. Everyone can now be a photographer; digital cameras are cheaper and easier to use; camera phones are close to ubiquitous. Our capacity to share and disseminate images has transformed with the rapid spread of image-driven social media, messaging services and a range of online platforms for curating and presenting image portfolios and archives. The plural forms of digital images—photographs, paintings, drawings, film, found images, archival images, public and private image collections, researcher generated images, research subject generated images—have rapidly expanded the range of research data that scholars can harness and access. The digitalisation of image archives, data visualisation techniques, forms of artistic and cultural visual production and the ubiquity of images in social communications and media have created whole new data sets and research opportunities for TJ researchers.

Digital technology has democratised image-making, allowing for new voices to be included in the research process through participatory modes of data production using visual methods. Social research is embracing more engaged, applied, collaborative, innovative and public forms of scholarship. Within postconflict settings, this shift chimes with calls to recognise and endorse the role participatory approaches have to play in rethinking TJ conceptually and operationally in response to critiques of top-down, one-size-fits-all models of TJ (Lundy and McGovern 2008, 265–92). Participatory visual methods and approaches, such as photovoice (Wang and Burris 1997, 369–87) or participatory video (Roberts and Lunch 2015), enable communities to play an active role in generating knowledge and shaping programmes and interventions.

Working with images requires a reflexive and ethical engagement about the objective status of “data”. Images are multivocal and multilinear, and research that uses images has to engage with complexity, instability and address questions of cultural meanings and power (Rose 2007). While research projects that examine large digital image data sets using quantitative methods are starting to emerge, the emphasis on meaning and significance in image-based research means that qualitative methods and collaborative and action research approaches have been more prevalent (for example, Rose and Willis 2018, 411–27).

A brief consideration of some of the ways images are being used within TJ research settings in the Western Balkans indicates their range of application. Researchers have studied the use and impact of images within the context of national courts and International Criminal Tribunals, demonstrating the potential of images to reach beyond the courtroom and to shape attitudes (Gow, Michalski, and Kerr 2013; Petrović 2014, 89–109). The impact of the circulation of images on the delivery of criminal justice has been examined through a study of the creation, circulation and contestation of iconic images of the conflict in Bosnia (Petrović 2015, 367–85). And within the context of the ICTY, a number of artists, such as Miladinović (2020), have produced new artwork based on digitised archives, focusing on both images contained within the archives but also textual documents and evidence. Beyond formal judicial mechanisms, Simić (2016, 11–3) has considered the role of images, in the form of a photographic exhibition featuring the portraits and testimonies of women survivors of conflict-related sexual violence, in making distinct reparative contributions to TJ processes (Harper 2002; Poole and Rojas-Pérez 2010).

Images can be used not only as data in and of themselves but also as a means to elicit alternative forms of narrative and interview data. Methods such as photo-elicitation (Cieplak 2017), which enable a collaborative form of “active” interviewing and research (Holstein and Gubrium 1995), have been employed within mixed methods research to build a multifaceted picture of social identity among youth in the Western Balkans (Felicia et al. 2017).

Research participants can create images as part of the research process. Some scholars, acknowledging the marginalisation of victims and local communities in retributive TJ approaches, argue that a grassroots community approach to TJ should be an intrinsic part of the postconflict agenda. Researchers are using participatory visual methods to work with communities to develop bottom-up and community-based truth-telling mechanisms and conflict transformation strategies. Using photovoice with women in Bosnia and Herzegovina, Smith (2020) points to the transformative potential of participatory visual methods that allow for the emergence of a multiplicity of narratives. Other research into participatory photography and media projects with youth in Bosnia and Herzegovina has argued that these projects create opportunities for dialogical and localised forms of TJ (Fairey and Kerr 2020, 142–64).

Images provide a multitude of opportunities for TJ researchers. However, it is important to sound a note of caution. Digital technology and online platforms are not freely available to all, especially within communities in postconflict settings where a lack of resources, infrastructure and opportunities limits people’s access. Digital images themselves raise a labyrinthine set of ethical issues. They are easily manipulated using postproduction software, and as they are shared, disseminated and re-appropriated quickly become de-contextualised from their site of production and divorced from their original authors and captioning information (Gross, Katz, and Ruby 2003). They pose considerable ethical challenges around consent and anonymity as digital technologies allow images to move seamlessly between private and public spheres (Pauwels 2008, 243–57).

Furthermore, participatory visual methods require specialist expertise (Pauwels 2008) and have the potential to disempower, create negative outcomes and silence rather than enable voice if misused (Fairey 2018, 111–26). Participatory processes more generally are vulnerable to co-optation and misappropriation. There is also limited consensus over agreed forms of image analysis. Rose highlights that “interpreting images is just that, interpretation, not the discovery of their ‘truth’” (Rose 2007, 4). Researchers have to justify their interpretation and root it in an explicit methodology. Given that a range of methodological approaches informs the use of images across academic disciplines, this requires scholars creatively to adapt and develop approaches in accordance with the contexts in which they are working.

Conclusion

It is clear that, across a variety of domains and media, technological advances offer significant opportunities for TJ research. We have access to new evidence for evaluating TJ as a practice and new tools to conduct the analysis, such as software packages, and facilitating the innovation in methods applied to investigate old as well as new technology-facilitated data. Our research across these domains has demonstrated that the field is making great strides as technology opens it up to significant theoretical gains. Some of these gains provide a more robust empirical grounding through both qualitative and quantitative methods on the well-known axes/questions in TJ research, such as the relationship between individual- and society-level effects. Other gains are reflected in opening up fresh theoretical inquiries by allowing intimate insights into the world of victims, living pain and overcoming trauma, such as photovoice research.

Our comparative investigation also clearly shows that possibilities for research provided by the “digital turn” demand a heightened level of reflexivity and critical engagement both with data and with research methods. They require a constant appraisal of their benefits and limitations in relation to research aims. In particular, we caution against the veneration of either new data or new methods as an absolute terra nova of TJ research. In fact, some perennial dilemmas, such as a need to understand the context of the subject matter, whether it is victimhood or violence, persist (also see Gavshon and Gorur 2019). At the same time, new questions and challenges emerge. An example is the issue of veracity in computer-mediated communication.

In parallel to questions related to the rigour and robustness of research are ethical questions. Some “old” questions are amplified by technology, such as the anonymity of research subjects, and some challenges are new. Inevitably, technological advances come with their own dilemmas, such as the issue of what is digitised when. Paradoxically, some challenges are yet to be revealed to us as scholars push the frontiers of TJ research theoretically and methodologically. Ethical issues in the conduct of TJ research are amplified by the sensitivity of the research matter. This also explains why TJ scholars have made a significant contribution to the development of social science research ethics in general (Clark 2012, 823–39; Simić 2016, 134–68). Identifying and addressing the risks for research subjects and researchers deriving from the digitisation of data, as we have attempted to do here in response to the dilemmas encountered during our research, is part of an effort to rethink the ethical implications of the “digital turn” across social sciences and humanities.

As we demonstrate, this data also provides opportunities for novel types of analysis involving and combining qualitative and quantitative research. These also raise the research prospects for addressing the tensions in the TJ scholarship rooted in different methodological stances. Lastly, this evaluation of new datasets and methods and of old and new ethical problems also indicates that there is a need for strong interdisciplinary work that enhances understanding of TJ through robust triangulation of methods. Equally, there is a need for more challenging but potentially more revelatory, engaged and robust interdisciplinary conversations in order for technological innovation providing for more and better data and analysis to translate into better knowledge and understanding of TJ processes and their effects. We hope that this data-focused interdisciplinary review provides an opening gambit.