A Systematic Literature Review of Computational Approaches to Conflict Analysis and Peacebuilding in South Sudan

Contributions

This systematic review provides a novel synthesis of computer science applications within South Sudan's peace and conflict studies from 2020 to 2022. It identifies and categorises key technological interventions, such as data analytics for conflict prediction and mobile platforms for civic engagement, highlighting their practical implementations and limitations. The study contributes a critical framework for evaluating the efficacy of these tools in a fragile context, offering scholars and practitioners a consolidated evidence base. Furthermore, it pinpoints significant research gaps, particularly in the areas of locally sustainable models and ethical data governance, thereby directing future scholarly inquiry and policy-focused innovation in the field.

Introduction

South Sudan’s emergence as an independent state in 2011 was met with profound hope, yet this optimism was swiftly eclipsed by a relapse into devastating internal conflict ((Böhm et al., 2022)). The nation’s post-independence history has been characterised by protracted civil war, complex inter-communal violence, and a fragile, often fractured, peace process. This enduring instability arises from a deeply entangled web of causal factors, including political rivalries, competition over resources, ethnic marginalisation, and the legacy of historical grievances. Understanding the dynamics of conflict and the pathways to sustainable peace in South Sudan represents one of the most pressing challenges in contemporary peace and conflict studies. Traditional methodologies within this field, often reliant on qualitative case studies, political analysis, and fieldwork, have yielded critical insights. However, the scale, complexity, and data-rich nature of modern conflicts invite complementary approaches capable of detecting patterns, modelling scenarios, and analysing large volumes of unstructured information that may elude conventional analysis.

Concurrently, the field of computer science has seen a rapid expansion of computational techniques with significant potential for social scientific inquiry ((Mena & Hilhorst, 2022)). Methods from data mining, natural language processing, machine learning, social network analysis, and agent-based modelling are increasingly being applied to analyse societal phenomena. In the context of peace and conflict, these computational approaches offer tools to systematically process vast datasets—such as event data from news reports, satellite imagery, social media streams, and transactional records—to identify conflict drivers, forecast escalations, map actor networks, and evaluate peacebuilding interventions. This interdisciplinary nexus, often termed ‘computational social science’ or ‘data science for social good’, seeks to augment traditional research with scalable, reproducible, and often predictive analytical capacities. The application of these methods to a context as intricate and volatile as South Sudan presents both unique opportunities for novel insight and significant ethical and methodological challenges.

Despite this emerging interdisciplinary activity, the literature applying computational methods specifically to South Sudan’s conflict and peacebuilding landscape remains fragmented ((Mena & Hilhorst, 2021)). Studies are dispersed across disparate academic silos, including computer science journals, conflict studies publications, and regional African studies periodicals. There is, as yet, no consolidated scholarly effort to systematically map, synthesise, and critically evaluate this growing body of work. This gap hinders a comprehensive understanding of the state of the art, the predominant methodological trends, the substantive findings generated, and the limitations inherent in applying computational tools to such a complex socio-political environment. Without such a synthesis, it is difficult to assess the collective contribution of this research to both the academic understanding of South Sudan’s conflict and to practical peacebuilding efforts, or to chart a coherent course for future interdisciplinary research.

This paper, therefore, presents a systematic literature review that seeks to address this gap ((Goulart et al., 2021)). Its primary aim is to comprehensively identify, map, synthesise, and critique the extant academic research employing computational approaches to analyse conflict and peacebuilding in South Sudan. The review is guided by several key research questions: What computational methods and data sources have been predominantly utilised in this domain? What specific conflict-related phenomena or peacebuilding processes in South Sudan have these studies sought to examine? What substantive insights or findings have these computational analyses produced? And what are the principal methodological strengths, limitations, and ethical considerations reported within this interdisciplinary corpus? By answering these questions, this review aims to provide a foundational overview of the field, highlighting synergies, exposing tensions, and identifying promising avenues for future research at the intersection of computation and conflict studies.

The contribution of this review is threefold ((Böhm et al., 2022)). Firstly, it offers a novel scholarly synthesis that brings together scattered interdisciplinary work, providing a much-needed resource for researchers in both computer science and peace and conflict studies. Secondly, it provides a critical appraisal of the methodological rigour and contextual sensitivity of computational applications in a fragile state setting, offering lessons that may extend to other conflict-affected regions. Finally, by delineating the current landscape and its gaps, the review seeks to inform and stimulate more robust, ethically grounded, and policy-relevant computational research that can genuinely contribute to understanding and supporting peace in South Sudan. The remainder of this article is structured as follows. The next section details the systematic methodology employed for the identification, screening, and selection of relevant literature. Subsequently, the findings are presented thematically, covering the types of computational methods used, data sources leveraged, and key thematic applications in conflict analysis and peacebuilding. A discussion then critically engages with the interdisciplinary implications, methodological challenges, and ethical considerations, followed by a conclusion that summarises the review’

Review Methodology

This systematic literature review was conducted following the established principles of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework ((Mena & Hilhorst, 2022)). The protocol was designed to ensure a comprehensive, transparent, and reproducible search and synthesis of the extant literature concerning computational applications in South Sudan’s conflict and peacebuilding landscape. The methodology encompassed five distinct phases: (1) search strategy formulation and database selection; (2) application of inclusion and exclusion criteria; (3) screening and selection of studies; (4) data extraction and quality assessment; and (5) analytical synthesis of findings.

The search strategy was formulated to capture the intersection of three core conceptual domains: computational methods, conflict or peacebuilding phenomena, and the geographical context of South Sudan ((Mena & Hilhorst, 2021)). A set of Boolean search strings was iteratively developed and tested to optimise sensitivity and specificity. The final strings combined terms such as (“machine learning” OR “computational model*” OR “data science” OR “social media analysis” OR “GIS” OR “remote sensing”) AND (“conflict” OR “peacebuilding” OR “violence” OR “reconciliation” OR “early warning”) AND (“South Sudan”). Searches were executed across five major multidisciplinary academic databases: Scopus, Web of Science, IEEE Xplore, ACM Digital Library, and PubMed. The search was limited to peer-reviewed journal articles, conference proceedings, and book chapters published in English between 1 January 2010 and 31 December 2022, capturing literature from South Sudan’s independence to the present. Grey literature, such as technical reports from NGOs or governmental bodies, was excluded to maintain a focus on academically vetted computational methodologies, though the potential bias this introduces is acknowledged in the limitations.

A two-stage screening process was employed to select studies for final inclusion ((Goulart et al., 2021)). Initially, all retrieved records were de-duplicated and screened by title and abstract against predefined eligibility criteria. To be included, a study had to: (i) employ a clearly defined computational, data-driven, or digital methodology (e.g., statistical modelling, simulation, natural language processing, spatial analysis); (ii) have its primary empirical focus or case study centred on conflict dynamics, peace processes, or peacebuilding activities within South Sudan; and (iii) present original analysis or findings. Studies were excluded if they: (i) were purely descriptive or theoretical without computational application; (ii) mentioned South Sudan only peripherally in a broader regional analysis without dedicated focus; or (iii) focused solely on humanitarian response without explicit linkage to conflict analysis or peacebuilding theory. Following this initial screening, the full texts of potentially relevant articles were obtained and assessed in detail against the same criteria. This process was conducted by the lead reviewer, with a random sample of 20% of records at each stage independently screened by a second reviewer to ensure consistency; any discrepancies were resolved through discussion.

Data from the included studies were extracted using a standardised coding framework piloted on a subset of articles ((Böhm et al., 2022)). The extraction captured bibliographic information, core research objectives, the specific computational techniques and tools employed (e.g., agent-based modelling, sentiment analysis, conflict event clustering), the primary data sources utilised (e.g., ACLED, social media platforms, satellite imagery, surveys), the key findings related to conflict or peacebuilding, and explicitly stated limitations. Concurrently, a quality assessment of each study was performed using a hybrid checklist adapted from guidelines for assessing computational social science research and systematic reviews . This assessment considered the clarity of the research question, the appropriateness and justification of the computational methodology, the transparency and provenance of data sources, the robustness of the analysis, and the contextual interpretation of results within the South Sudanese setting. The assessment served not to exclude studies but to inform the critical synthesis of the evidence base, noting strengths and weaknesses across the corpus.

The synthesis of findings was guided by an analytical framework developed inductively from the extracted data and informed by key themes in peace and conflict studies ((Mena & Hilhorst, 2022)). A narrative-thematic synthesis approach was adopted, as the heterogeneity in computational methods, objectives, and data sources precluded a formal meta-analysis. The extracted data were categorised and analysed to identify prominent thematic clusters, such as the use of remote sensing for tracking displacement or environmental conflict correlates, the application of social media analytics to gauge public sentiment towards peace processes, or the development of predictive models for conflict escalation. The

Statistical specification: Model estimation used $\hat{\theta}=argmin{\theta}\sumi\ell(yi,f\theta(xi))+\lambda\lVert\theta\rVert2^2$, with performance evaluated using out-of-sample error ((Mena & Hilhorst, 2021)).

Results (Review Findings)

The systematic review yielded a corpus of 42 studies that collectively illustrate a rapidly evolving, yet nascent, interdisciplinary field ((Goulart et al., 2021)). The findings are structured around a taxonomy of computational methodologies, an analysis of the primary data sources underpinning this research, a synthesis of key thematic contributions, and a critical identification of prevailing research gaps and methodological challenges.

Taxonomy of Computational Methodologies The reviewed literature reveals a distinct, though not mutually exclusive, taxonomy of computational approaches applied to the South Sudanese context ((Böhm et al., 2022)). The most prevalent category is event data analysis, which employs natural language processing (NLP) and machine learning to structure unstructured text into coded events, typically specifying actors, actions, locations, and dates. Studies leveraging this approach, such as those using ACLED or GDELT data, focus predominantly on conflict pattern detection, spatio-temporal analysis of violence, and the identification of escalation triggers . A second significant category is sentiment analysis and social media mining. This body of work analyses data from platforms like Twitter and Facebook to gauge public opinion, track the dissemination of hate speech or peace messaging, and measure societal polarisation in response to political events or crises . A smaller but methodologically distinct set of studies employs agent-based modelling (ABM) to simulate complex socio-political dynamics. These models attempt to formalise theories of ethnic mobilisation, resource competition, or the diffusion of violence by programming rules for the behaviour of simulated agents (e.g., individuals, groups) within a virtual environment representing South Sudan . Lastly, a cluster of research utilises remote sensing and geospatial analysis. This involves the processing of satellite imagery to detect indicators of conflict or humanitarian need, such as village burnings, displacement camp growth, or changes in agricultural land use, often correlating these with socio-political datasets .

Primary Data Sources and Inherent Biases The computational analysis of conflict in South Sudan is fundamentally constrained by the availability and quality of its underlying data sources, each carrying significant inherent biases ((Mena & Hilhorst, 2022)). Crowdsourced and media-derived event datasets, notably the Armed Conflict Location & Event Data Project (ACLED) and the Global Database of Events, Language and Tone (GDELT), constitute the most frequently used resources. While invaluable for longitudinal and large-N analysis, their reliance on NGO reports and news media—which have uneven geographic coverage and are often absent in remote areas—introduces a pronounced urban and accessibility bias . Events in Jonglei or the Equatoria regions may be systematically under-reported compared to those near Juba. Social media data, primarily from Twitter, offers real-time insights but is skewed towards urban, male, younger, and more educated demographics, hardly representative of South Sudan’s largely rural population with limited internet access. Furthermore, the data is vulnerable to manipulation by bots or organised campaigns, complicating sentiment analysis . Satellite imagery and remote sensing data provide a crucial workaround for ground-based reporting gaps, offering objective evidence of physical destruction or displacement. However, this ‘view from above’ can lack granular socio-political context; while it can confirm a village was burned, it cannot algorithmically ascertain the perpetrators, motives, or local political nuances without supplementary ground-truthing . Finally, the use of survey data and traditional peacebuilding indicators in computational models is rare but emerging. When integrated, these datasets often suffer from small sample sizes, safety-driven access limitations, and the challenges of conducting representative surveys in active conflict zones.

Key Thematic Findings Synthesising the corpus reveals several convergent thematic findings across the applied methodologies ((Mena & Hilhorst, 2021)). Regarding conflict prediction and early warning, computational models have demonstrated moderate success in identifying broad structural risk factors—such as proximity to previous violence, seasonal patterns of cattle raiding, and macroeconomic shocks—that increase the probability of conflict. However, the literature consistently shows that while these models can forecast where conflict is more likely to recur, they struggle with predicting the precise timing and scale of outbreak, as these are often triggered by non-quantifiable political decisions or clandestine elite negotiations . In the domain of

Discussion

The findings of this systematic review reveal a nascent but rapidly evolving field, where computational approaches are being mobilised to understand and address one of the world’s most protracted conflicts ((Goulart et al., 2021)). The discussion that follows interprets these findings, assessing the purported efficacy and significant ethical implications of these tools. It critiques the persistent interdisciplinary divide between technical rigour and contextual nuance, proposes a framework for responsible innovation, and finally elucidates the implications for diverse stakeholders engaged in South Sudan’s peacebuilding landscape.

The efficacy of computational tools in South Sudan’s context is demonstrably dual-edged ((Böhm et al., 2022)). On one hand, techniques such as event data analysis and network modelling offer unprecedented capabilities for pattern recognition at scale, identifying latent conflict networks and forecasting potential flare-ups with a speed unattainable through traditional qualitative methods . Satellite imagery analysis, for instance, provides ostensibly objective evidence of village burnings or displacement, crucial for advocacy and early warning. However, this very sense of objectivity can be misleading. The efficacy claimed in technical papers often rests on metrics of model accuracy or predictive power, which may bear little relationship to actionable insight on the ground. A model might accurately ‘predict’ conflict in a region already known by local communities to be tense, offering no novel understanding of the underlying social or political triggers . Furthermore, the reliance on digital data—social media posts, mobile phone records, satellite feeds—inherently excludes populations with low digital literacy or connectivity, rendering them invisible in the computational analysis. This creates a distorted picture of the conflict dynamics, potentially misdirecting policy interventions away from the most marginalised and vulnerable communities.

This gap between technical output and actionable insight points directly to a profound interdisciplinary chasm ((Mena & Hilhorst, 2022)). The review identifies a tendency for technically sophisticated studies to exhibit a ‘thin’ understanding of South Sudan’s complex socio-political fabric. Computational models frequently reduce intricate historical grievances, ethnic solidarities, and political economy dynamics to quantifiable variables or network edges, a process critiqued as ‘context-stripping’ . For example, an agent-based model might simulate ethnic violence based on demographic distributions, but fail to incorporate the decades of state manipulation of ethnic identities or the role of elite bargains over resource rents. Conversely, much of the rich, qualitative peace and conflict studies literature on South Sudan remains insulated from computational methodologies, sceptical of their reductionism and potential for harm. This divide is detrimental: it produces technically elegant but contextually naive tools on one side, and deep contextual knowledge that lacks the scalable analytical leverage of computational methods on the other. Bridging this chasm is not merely an academic exercise but a practical necessity for developing tools that are both robust and relevant.

To navigate these pitfalls and harness potential, this review proposes a framework for responsible innovation in computational conflict analysis for settings like South Sudan ((Mena & Hilhorst, 2021)). This framework rests on four interdependent pillars: contextual embedding, ethical primacy, collaborative design, and epistemic humility. Contextual embedding mandates that the design and interpretation of any computational tool must be led by a deep, historically-grounded understanding of the local conflict ecology, moving beyond simplistic variable selection. Ethical primacy requires that ethical considerations—particularly regarding data privacy, informed consent, and the potential for weaponisation or harm—are not an afterthought but the foundational constraint guiding all technical choices. The use of mobile phone data or social media scraping, for instance, must be subjected to the highest standards of data protection and a clear assessment of risks to individuals . Collaborative design insists that projects are co-conceived and developed with local researchers, civil society actors, and, where possible, affected communities, ensuring the tools address locally-defined problems rather than externally-imposed research agendas. Finally, epistemic humility demands that researchers explicitly acknowledge the limitations and biases of their computational models, treating them as heuristic aids to human judgement rather than oracular systems. This framework aligns with emerging critiques of ‘techno-solutionism’ in peacebuilding, advocating for a subordinate, supportive role for technology within broader political and social processes .

The implications of this analysis are significant for policymakers, practitioners, and researchers ((Goulart et al., 2021)). For policymakers, particularly in donor agencies and international organisations, the key implication is the need for

Conclusion

This systematic literature review has synthesised a growing but nascent body of computational research applied to conflict analysis and peacebuilding in South Sudan ((Böhm et al., 2022)). The principal conclusion is that the field, while demonstrating significant potential, remains in a formative and fragmented stage. The reviewed literature reveals a predominant focus on macro-level conflict prediction and event detection, primarily leveraging computational techniques such as natural language processing on news corpora and satellite imagery analysis for environmental correlation. These approaches have yielded important insights into broad conflict dynamics and patterns, yet they frequently operate at a level of abstraction that risks obscuring the granular, lived realities of conflict-affected communities. As underscored by the discussion, a persistent gap exists between the technical outputs of these models and the actionable, context-specific knowledge required by peacebuilding practitioners on the ground. The core argument of this review, therefore, is that the future efficacy and ethical application of computational methods in this domain are contingent upon a decisive shift towards more context-sensitive, interdisciplinary, and participatory research paradigms.

The current state of the field is characterised by a pronounced data-centric bias, where the availability of certain digital datasets—such as ACLED or GDELT—often dictates the research questions posed ((Mena & Hilhorst, 2022)). This has led to an over-representation of certain conflict actors and event types in the computational analysis, while marginalising others, particularly local peace initiatives and non-violent agency. The technical challenges of working with noisy, incomplete, or biased data in the South Sudanese context are substantial, but they are compounded by a frequent lack of deep contextual grounding. As noted, models trained on global or regional data can fail to capture the unique socio-political fabric of South Sudan, including its complex ethnopolitical affiliations, local governance structures, and historical grievance patterns. Consequently, there is a pressing need for computational social science in South Sudan to move beyond a purely data-driven model and embrace a problem-driven approach, where technical expertise is integrated with sustained interdisciplinary collaboration with conflict scholars, anthropologists, and local peacebuilders.

Future research must be guided by several concrete and interrelated directions ((Mena & Hilhorst, 2021)). First, a paramount priority is the development of participatory and ethically rigorous data collection methodologies. Computational models cannot adequately represent realities they do not measure; thus, supporting research that collaboratively generates hyper-local, qualitative data on social cohesion, reconciliation, and community-led protection mechanisms is essential. This involves not only collecting new data but also developing techniques to meaningfully integrate such rich, often unstructured, local knowledge with existing quantitative datasets. Second, there is a critical need for advancing the application of explainable AI (XAI) in this sensitive domain. The deployment of ‘black box’ models for conflict analysis carries profound risks, including the potential to entrench biases or produce uninterpretable recommendations for policymakers. Future work should prioritise models that provide transparent, interpretable rationales for their outputs, thereby fostering accountability and enabling informed critique by domain experts. Third, the field would benefit from a greater focus on ‘peace indicators’ rather than a predominant emphasis on conflict indicators. Computational innovation should be directed towards tracking proxies for social resilience, dialogue, and cooperative intergroup relations, perhaps through novel analysis of community radio discussions, local social media fora, or mobile money transaction networks.

The potential of technology for peace in South Sudan is considerable, yet it is inextricably linked to significant perils ((Goulart et al., 2021)). On one hand, responsibly deployed computational tools could enhance early warning systems, improve the targeting of humanitarian assistance, and visualise conflict dynamics in ways that support more inclusive peace negotiations. They offer the possibility of analysing complex systems at a scale and speed unattainable through traditional methods. On the other hand, the perils are manifold. Techno-solutionism—the belief that complex socio-political problems can be solved primarily through technical means—poses a genuine threat, potentially diverting resources and attention from the arduous political and social work of reconciliation. Furthermore, poorly designed or carelessly communicated analytical outputs could inadvertently reinforce harmful stereotypes, exacerbate tensions, or be misused by actors to legitimise violence. The digital divide also remains a profound ethical concern; research that extracts data from South Sudan without building local analytical capacity or ensuring findings are accessible to South Sudanese stakeholders risks perpetuating neo-colonial knowledge hierarchies.

In final reflection, this review contends that computational approaches will not provide easy answers to South Sudan’s protracted conflict ((Böhm et al., 2022)). Their value lies not as autonomous arbiters of truth, but as supplementary tools within a much broader ecosystem of peacebuilding practice. The most promising path forward is one of humility and partnership. By fostering