Mapping Unseen Boundary: Social Network Analysis for Human-Wildlife Conflict Prediction

The Unseen Web of Conflict: Why Traditional Approaches Fall Short

Human-wildlife conflict (HWC) is rarely a simple dyadic interaction between a farmer and a tiger. It is embedded in a complex web of social, ecological, and institutional relationships that traditional conflict mapping often overlooks. As conservation practitioners, we have long relied on spatial analysis and incident reporting, but these methods capture only the symptoms, not the underlying relational dynamics that drive conflict. Social Network Analysis (SNA) offers a paradigm shift: it maps the unseen boundaries of influence, trust, resource flow, and information exchange among stakeholders, wildlife populations, and landscape features. In this guide, we will demonstrate how SNA can transform HWC prediction by revealing hidden patterns of interaction that precede visible conflicts. This is not a theoretical exercise; it is a practical toolkit for anyone tasked with reducing conflict in complex socio-ecological systems.

Traditional approaches typically treat each conflict event as an isolated incident, analyzing spatial clusters of crop raiding or livestock depredation. While useful, this perspective ignores the social fabric that shapes human behavior—community networks, kinship ties, and power dynamics that influence tolerance, retaliation, and reporting. For instance, a farmer may not report a crop raid if a neighbor who is a park ranger's relative has warned against it. SNA captures these subtle ties, providing a more accurate picture of conflict drivers.

The Relational Turn in Conservation

The shift toward relational thinking in conservation acknowledges that conflicts emerge from interactions among multiple actors with divergent interests, perceptions, and capacities. SNA formalizes this by representing actors as nodes and their relationships as edges, enabling quantitative analysis of network properties. A key insight is that conflicts often occur at the boundaries of tightly-knit communities—where information flow is weak and trust is low. For example, a study of farmers bordering a national park in Southeast Asia found that villages with denser internal networks but sparse connections to park management experienced more retaliatory killings. The network structure itself was a stronger predictor of conflict than crop loss or livestock numbers. This demonstrates that SNA can identify high-risk zones where intervention should focus on bridging network gaps rather than simply erecting fences.

Another limitation of conventional methods is their inability to capture feedback loops. A conflict event changes relationships—it may increase cohesion among affected farmers or erode trust between communities and authorities. SNA allows us to model these dynamics over time, predicting how a single incident can cascade through the network. For instance, a retaliatory killing of a wolf might be triggered not by direct predation but by a rumor spread through a social media group that amplifies fear. SNA can map these information pathways, enabling preemptive counter-narratives or trust-building measures. By treating conflict as a relational phenomenon, we shift from reactive mitigation to proactive prediction, targeting the root causes of tension before they escalate.

In practice, this means rethinking data collection. Instead of only recording conflict events, we must map relationships: who talks to whom, who shares resources, who has influence. This can be done through interviews, surveys, and even automated analysis of communication data (with ethical safeguards). The investment pays off by revealing leverage points—a single well-connected individual whose change in attitude can tip the balance across an entire community. As we will see in subsequent sections, SNA is not a replacement for spatial analysis but a complementary lens that adds a critical human dimension to conflict prediction.

Core Frameworks: How Social Network Analysis Reveals Hidden Patterns

To understand how SNA predicts human-wildlife conflict, we must first grasp its core concepts and how they relate to conflict dynamics. At its heart, SNA provides a set of metrics that quantify the structure and flow within a network. These metrics—centrality, density, modularity, and tie strength—each offer a unique window into the social mechanisms that underpin conflict. When applied to HWC, they help identify not just who is involved, but how their interactions create conditions for conflict or cooperation. This section breaks down these frameworks with concrete examples, explaining why they work rather than just defining them.

Centrality: Who Holds the Keys to Peace or Conflict?

Centrality measures identify the most influential nodes in a network. Degree centrality counts direct connections; a farmer with many ties to neighbors may be a key disseminator of information or a potential conflict amplifier. Betweenness centrality highlights nodes that bridge otherwise disconnected groups; a wildlife ranger who regularly attends both community meetings and park headquarters can act as a conflict mediator or a bottleneck for information. In one composite scenario, an agricultural extension officer with high betweenness centrality was found to be the sole link between a cluster of farming villages and the wildlife department. When that officer left the area, conflict reports increased by 40% because communication broke down. Centrality analysis would have flagged this dependency and prompted training of multiple liaisons. Eigenvector centrality, which considers the influence of a node's connections, can identify individuals whose attitudes spread rapidly. A village elder with high eigenvector centrality who advocates for coexistence can shift norms across the entire community.

Practitioners should compute centrality scores for all stakeholder nodes—including human actors, wildlife corridors, and even infrastructure like water points that act as attractants. The highest-risk nodes are often not the most central but those with high betweenness in a sparse network; they become single points of failure for communication. Conversely, nodes with high degree centrality but low betweenness may be locally influential but isolated from decision-makers. By mapping these patterns, we can target interventions: strengthen ties to isolated communities, diversify communication channels, and empower influential advocates. Centrality is not a static property; it changes as relationships evolve. Regular network surveys (e.g., annually) can track shifts and predict emerging tensions before they erupt.

Density and Modularity: The Community Structure of Conflict

Network density measures the proportion of possible ties that are present. Dense networks—where everyone knows everyone—can facilitate rapid information sharing and social cohesion, reducing conflict through peer pressure. However, extremely dense networks can also be insular, resistant to external conservation messages, and prone to groupthink. Sparse networks, on the other hand, may suffer from misunderstandings and lack of trust, leading to conflict. The key is to find an optimal density, which varies by context. For example, among pastoralist communities in East Africa, moderate density (where herders share grazing information but maintain strategic autonomy) was associated with lower conflict with lions compared to either very isolated or very cliquish groups.

Modularity detects subgroups (communities) within a network. High modularity means the network is divided into distinct clusters with few cross-group ties. In HWC contexts, these clusters often align with ethnic groups, land tenure types, or political factions. Conflict frequently occurs at the boundaries between modules, where miscommunication and resource competition are highest. A park bordering two ethnic groups with weak ties between them is a classic hotspot. SNA can quantify the strength of these boundaries and recommend bridge-building interventions—joint patrols, cross-community meetings, or shared water projects—that increase intermodular ties. By reducing modularity, we reduce the structural conditions for conflict. One project in Nepal found that communities with lower modularity (more ties between upstream and downstream villages) had 30% fewer crop-raiding incidents because they coordinated early warning and deterrents. Thus, modularity analysis directly informs where to invest in relationship-building.

Tie strength, often measured by frequency or emotional intensity, also matters. Strong ties (family, close friends) are sources of support and compliance but can also be conduits for rumor and fear. Weak ties (acquaintances, cross-group links) are bridges for novel information and innovation. In conflict prediction, a sudden increase in weak ties between previously separate modules may signal new cooperation—or new points of friction. Monitoring tie strength changes over time can provide early warning. For instance, if a conservation agency starts hiring local guides from only one community, previously weak ties to that community may strengthen, but other communities may feel excluded. SNA would detect this imbalance and prompt inclusive hiring. Overall, these frameworks turn the messy reality of human relationships into actionable metrics.

Execution: A Step-by-Step Workflow for Building a Predictive Network Model

Translating SNA theory into a practical predictive system requires a methodical process that balances rigor with feasibility. Based on experiences from several conservation projects, we outline a seven-step workflow that teams can adapt to their context. This workflow covers problem definition, data collection, network construction, metric calculation, conflict data integration, model building, and validation. Each step includes decision points, trade-offs, and common pitfalls drawn from real implementations. The goal is to produce a network model that can identify high-risk areas and relationships before conflicts occur, enabling proactive intervention.

Step 1: Define the Network Boundary and Nodes

The first and most critical step is to decide who and what to include. The network boundary must be meaningful for the conflict in question. For crop raiding by elephants, nodes might include farming households, village leaders, wildlife rangers, water sources, elephant corridors, and market towns. For livestock predation, nodes could be herder camps, grazing areas, predator dens, and veterinary posts. The principle is to include any entity that influences conflict dynamics, even if not directly involved. A common mistake is to include only human stakeholders; wildlife and landscape nodes are essential because they attract or repel conflict. However, the network must remain manageable; a rule of thumb is to limit to 100-200 nodes for a single analysis unit (e.g., one landscape). Overly large networks become unwieldy and may obscure patterns. Use participatory mapping with local stakeholders to identify key nodes, ensuring cultural relevance and buy-in. In one project, initial interviews missed a sacred grove that served as a predator refuge; adding it improved model accuracy by 15%.

Step 2: Collect Relational Data

Relational data is the foundation of the network. The most common method is a survey that asks each respondent about their ties to other nodes: who they talk to about wildlife issues, who they trust for advice, who shares resources, and who they would call for help. To reduce bias, use a roster method (list of names) for small networks or a snowball approach for larger ones. Also collect attribute data (age, land size, livestock number, conflict history) to contextualize ties. Ethical considerations are paramount; ensure informed consent, anonymize data, and be transparent about how the data will be used. In some contexts, trust may be low, and relationships may be sensitive (e.g., ties to poachers). Use proxy measures like participation in meetings or mobile phone call records (with consent) to supplement self-reports. A pilot test of the survey with 10-15 respondents can reveal ambiguous questions and adjust for cultural norms. For example, in some cultures, asking about 'trust' directly may be inappropriate; instead, ask 'who would you share a sensitive concern with?' Collect data over a defined time window (e.g., past six months) to capture recent dynamics.

Step 3: Construct the Network Graph

Once data is collected, encode it into an adjacency matrix or edge list. Each row and column corresponds to a node, and entries indicate the presence or strength of a tie. This matrix is then imported into network analysis software (e.g., Gephi, R, or Python's NetworkX). Visualize the initial graph to spot obvious errors—isolated nodes that should be connected, or unrealistic clusters. Check for reciprocity (if A names B, does B name A?) as a quality indicator; low reciprocity may suggest misunderstanding or missing data. Impute missing ties cautiously; for example, if a respondent is absent, you may need to exclude them or use average values. The graph should be directional or undirected based on the nature of ties; information flow is directional, while kinship is undirected. At this stage, compute basic descriptive statistics: number of nodes, edges, density, and average degree. These numbers give a first impression of network structure. If density is very low (0.5) may indicate that the network is too aggregated and needs differentiation.

Step 4: Calculate Network Metrics

With the graph built, compute the key metrics discussed earlier: degree centrality, betweenness centrality, eigenvector centrality, modularity, and tie strength distributions. For each node, generate a profile that can be used as features in the predictive model. Additionally, compute network-level metrics like overall density, average path length, and clustering coefficient. These metrics become the independent variables for conflict prediction. It is crucial to standardize metrics across networks if comparing different time periods or regions. For example, betweenness centrality should be normalized by the total number of node pairs to allow cross-network comparison. Also compute ego-network metrics for each node—the subgraph of a node and its immediate neighbors—which can capture local dynamics like the density of a farmer's personal network. A farmer with a sparse ego-network may be more vulnerable to false information and thus more prone to conflict escalation. Store all metrics in a structured database with node IDs.

Step 5: Integrate Conflict Event Data

The dependent variable in this model is conflict occurrence or severity. Collect historical conflict records from park reports, community logs, or government databases. Georeference each event and link it to the nearest node(s) in the network. If a conflict event occurs at a specific location, assign it to the closest village or water point node. For continuous conflict data (e.g., number of incidents per month per village), aggregate to node level as a conflict score. Alternatively, use a binary indicator (conflict vs. no conflict) for a defined time window (e.g., past year). It is important to align the time window of network data with conflict data; ideally, network data should precede the conflict period to allow prediction. For example, collect network data in January, then track conflicts from February to December. If using retrospective data, ensure the network likely predates conflicts. Also consider spatial autocorrelation: conflicts in neighboring nodes may be correlated. Include a spatial lag term in the model or use a network autoregressive model to account for this. Standardize conflict metrics across nodes to avoid bias from reporting differences.

Step 6: Build a Predictive Model

With network metrics as features and conflict data as the target, choose a modeling approach. For small datasets (fewer than 100 nodes), logistic regression or generalized linear models work well and offer interpretability. For larger datasets, random forests or gradient boosting can capture non-linear interactions among metrics. Network-specific models like exponential random graph models (ERGMs) or stochastic actor-oriented models (SAOMs) can model tie formation and conflict jointly, but they are complex and require specialized expertise. A pragmatic approach is to start with a simple model and test its performance, then iterate. Split data into training and test sets (e.g., 70% train, 30% test) and use cross-validation to avoid overfitting. Key features often include a node's betweenness centrality, its degree, and the density of its community. Also include attribute data like land use or household income as controls. Evaluate model performance using metrics like AUC-ROC (for binary outcomes) or mean absolute error (for count outcomes). If performance is poor, revisit earlier steps: perhaps the network boundary is wrong, or ties need to be weighted differently.

Step 7: Validate and Iterate

Validation is not a one-time event. The true test of the model is whether it predicts future conflicts not used in training. Set up a monitoring system that records new conflicts and updates the network periodically (e.g., every six months). Re-run the model and compare predictions with actual events. If the model fails to predict a known hotspot, investigate why—maybe a new road changed movement patterns, or a new leader shifted alliances. Use this feedback to refine the network: add new nodes, update tie strengths, or recalculate metrics. Over time, the model can be calibrated to local conditions. Also conduct sensitivity analysis: perturb the network (e.g., remove a node or change a tie) and see how predictions change. This reveals which nodes are critical and where interventions would have the largest impact. Finally, communicate results to stakeholders in an accessible format—a network visualization with color-coded risk scores can be more persuasive than a table of numbers. The goal is not a perfect model but a tool that improves decision-making and reduces conflict.

Tools, Stack, and Economic Realities

Implementing SNA for HWC prediction requires not only methodological knowledge but also practical choices about software, data management, and budget. The good news is that many powerful tools are open-source or low-cost. However, the economic realities of conservation projects often constrain options: limited funding, variable internet access, and diverse technical skills among team members. This section reviews the essential tool stack, compares popular options, and discusses cost-benefit considerations to help readers choose appropriately. We also address maintenance and sustainability, as network models require ongoing updates to remain relevant.

Software Options: From Desktop to Cloud

For network construction and visualization, Gephi is a free, user-friendly desktop application that supports interactive exploration of large graphs. It runs on Java, works offline, and has a low learning curve, making it ideal for initial analysis and stakeholder presentations. For more advanced statistical modeling, the R programming language with packages like igraph, sna, and statnet offers flexibility and a vast ecosystem. R is free but requires programming skills; teams with a data analyst can leverage its power. Python with NetworkX is an alternative for those already using Python for other tasks. For cloud-based collaboration, NodeXL (an Excel add-in) or web tools like Kumu (freemium) provide easier sharing but may have data limits. We recommend a hybrid approach: use Gephi for visualization and R for modeling, with data stored in a relational database (SQLite or PostgreSQL). For teams with very limited capacity, consider hiring a consultant to set up the initial pipeline, then train local staff on maintenance. A typical one-time setup cost (including software, training, and initial data collection) ranges from $5,000 to $15,000, depending on landscape size and data availability. Recurring costs (annual data updates, analysis) are lower, around $2,000–$5,000.

Data Collection Technology

Collecting relational data efficiently often requires mobile data collection tools. ODK (Open Data Kit) is a free, offline-capable platform that runs on Android devices and supports complex survey logic. It is widely used in conservation and can be paired with KoboToolbox for easier form design. For automated data capture, mobile phone metadata (call detail records or CDRs) can reveal communication patterns without surveys, but this raises privacy concerns and requires explicit consent and regulatory approval. In one pilot in East Africa, researchers obtained anonymized CDRs from a telecom company and derived tie strength from call frequency; the resulting network predicted conflict hotspots with 70% accuracy. However, such partnerships require negotiation and ethical oversight. A simpler alternative is to use social media data (e.g., WhatsApp group membership) but again with consent. The choice depends on the context: in areas with low phone penetration, face-to-face surveys remain essential. Budget for survey enumerators, training, and data entry should be factored in. A survey of 100 nodes might cost $1,000–$3,000 depending on local rates.

Economic Trade-offs and Sustainability

Conservation projects often operate on short funding cycles, yet SNA models require long-term investment to yield returns. A common mistake is to build a comprehensive model in the first year but fail to update it, leading to rapid decay of predictive power. To address this, integrate network data collection into existing monitoring routines. For example, park rangers who already conduct patrols can also administer a short relational survey during community visits. This reduces marginal cost and ensures regular updates. Another strategy is to focus on a small, high-impact area rather than a large landscape; the model will be cheaper to maintain and more actionable. Compare the cost of SNA implementation to the cost of unmitigated conflict: a single livestock depredation event can cost a family $500, and retaliatory killings may result in fines or lost tourism revenue of $10,000+. If SNA prevents even a few events per year, it pays for itself. However, be realistic about the timeline; it may take two to three years before the model is accurate enough to guide decisions. Secure multi-year funding from donors who understand the long-term benefit. Finally, consider sharing infrastructure with neighboring projects to spread costs.

Growth Mechanics: Building Institutional Capacity and Sustaining Impact

Adopting SNA for conflict prediction is not a one-off technical fix; it requires a shift in how an organization thinks about data, collaboration, and decision-making. Growth in this context means scaling the approach across landscapes, embedding it in institutional routines, and building a community of practice. This section explores the human and organizational factors that determine whether SNA becomes a lasting tool or a pilot that fades. We draw on experiences from networks of conservation practitioners to identify best practices for capacity building, knowledge management, and adaptive management.

Training and Skill Transfer

The most common barrier to sustained use of SNA is lack of local expertise. A typical project trains one or two staff members on SNA software, but if they leave, institutional knowledge vanishes. To mitigate this, adopt a train-the-trainer model. Conduct workshops for a cohort of 5-10 staff from different departments (research, community outreach, enforcement) and provide ongoing mentorship via virtual check-ins. Develop documentation in the local language, with step-by-step guides and video tutorials. Use low-stakes exercises—like mapping a small social network among colleagues—to build confidence before tackling real conflict data. Also, partner with a university or research institute that can provide long-term support and host a repository of code and templates. Over three years, aim for at least three staff members who can independently run the full workflow. In one project in the Amazon, a local NGO created a 'Network Lab' that trained indigenous community monitors; they now produce quarterly network reports that inform park management decisions. This kind of capacity building not only sustains the tool but also empowers local communities, improving trust and participation.

Institutionalizing Network Thinking

For SNA to influence real decisions, it must be integrated into existing planning cycles. That means linking network metrics to resource allocation: if a village has high betweenness centrality, it should receive priority for conflict mitigation funds. Create a dashboard that displays risk scores alongside other indicators (crop loss, wildlife sightings) and is reviewed in monthly meetings. Ensure that the model is not a black box; decision-makers need to understand why a particular node is flagged. Use simple visual metaphors—like traffic light colors for risk levels—to communicate findings. Over time, as trust in the model grows, it can inform larger strategies: where to build fences, where to rotate ranger patrols, or which communities to engage in conservation education. Document lessons learned in an adaptive management log; when predictions fail, note the reasons and adjust the model. This creates a feedback loop that improves both the model and organizational learning. Also, share findings with the broader conservation community through reports, webinars, and peer-reviewed publications to build the evidence base. The more the approach is used and refined, the more credible it becomes, attracting further funding and partnerships.

Scaling Across Landscapes

Once the model works in one area, consider replicating it in adjacent landscapes. However, resist the temptation to simply copy-paste; each landscape has unique social and ecological dynamics. Instead, develop a toolkit or 'SNA-in-a-box' that includes standard survey instruments (customizable to local context), a data processing pipeline, and an analysis protocol. Train a regional coordinator who can oversee multiple sites and facilitate cross-site comparisons. Use a common set of core metrics while allowing site-specific variables. For example, all sites might compute betweenness centrality, but the list of nodes (e.g., water sources) can vary. A network of sites can also be linked at a higher level: if wildlife corridors connect two landscapes, the networks themselves may interact. This meta-network analysis can reveal landscape-scale conflict drivers that individual sites miss. Scaling also involves advocacy: demonstrate to government agencies that SNA-based early warning systems can reduce conflict costs, making a case for policy integration. In some countries, conservation authorities have adopted SNA as part of their national human-wildlife conflict strategy, securing dedicated funding for network monitoring. This is the ultimate growth metric—when the approach becomes institutionalized beyond any single project.

Risks, Pitfalls, and Mitigations in SNA for Conflict Prediction

No methodology is without risks, and SNA applied to HWC is no exception. Practitioners must navigate ethical minefields, methodological traps, and practical challenges that can undermine the validity and acceptance of the work. This section catalogs the most common pitfalls—drawn from documented project experiences—and offers concrete mitigation strategies. By anticipating these issues, readers can design more robust and ethically sound projects from the outset. We focus on risks related to data quality, privacy, community dynamics, model interpretation, and sustainability.

Data Quality and Sampling Bias

The most frequent pitfall is incomplete or biased relational data. If key stakeholders are missed (e.g., migrant herders, women, or marginalized groups), the network is skewed and predictions may be unreliable. Mitigation: use multiple sampling strategies—snowball sampling starting from diverse seed nodes, and stratified sampling by known subgroups. Cross-check with local key informants who can identify missing nodes. Also, be aware of recall bias; people may forget weak ties or overreport strong ties. Use a roster of potential ties to prompt memory, and ask about specific behaviors (e.g., 'how many times did you speak to X about wildlife in the past month?') rather than vague 'relationships'. Triangulate with observation or communication logs when possible. Another quality issue is missing data from non-responses. If a node is absent, consider imputation using average ties of similar nodes, but document the imputation rate and test sensitivity. A rule of thumb: if more than 20% of nodes have significant missing data, the network may be unreliable. In such cases, focus on smaller subnetworks that are complete.

Privacy and Ethical Concerns

Network data is inherently relational and can reveal sensitive information about alliances, conflicts, and power structures. This poses risks to participants if data is mishandled. For example, a map showing that a particular farmer has strong ties to a known poacher could lead to retribution. Mitigation: anonymize data immediately after collection; use pseudonyms for node names in all outputs. Store raw data on password-protected, offline devices. Obtain informed consent that explicitly describes how data will be used and shared, and offer participants the option to withdraw at any time. In some contexts, it may be necessary to avoid collecting certain sensitive ties altogether (e.g., ties to illegal actors) and instead use proxy indicators. When presenting results to authorities, aggregate data to community level rather than showing individual nodes. Also, consider the power dynamics: if the conservation authority is seen as a coercive force, network data may be used to target communities, exacerbating conflict. Engage a community advisory board to oversee data governance and ensure that benefits (e.g., improved conflict response) are shared equitably. Ethical review board approval is essential for any research-oriented project.

Overinterpreting Network Metrics

Another common mistake is to treat network metrics as deterministic predictors. High betweenness centrality does not automatically mean a node is a mediator; it could be a bottleneck that, if removed, improves flow. Similarly, a dense community may be resistant to change or may be highly cooperative—context matters. Mitigation: always interpret metrics within the local social and cultural context. Ground-truth findings through qualitative interviews; ask stakeholders why a particular node is central. Use multiple metrics together to triangulate. For example, a node with high degree and high betweenness is likely influential, but also check its eigenvector centrality to see if it connects to other influential nodes. Avoid making absolute claims; instead, frame results as probabilities or hypotheses that need further investigation. When presenting to decision-makers, emphasize that SNA identifies patterns, not certainties, and recommend that any intervention be piloted and monitored before scaling. Also, be aware of the modifiable areal unit problem: changing the network boundary can alter metrics. Test sensitivity by varying the boundary and see if results hold.

Community Resistance and Mistrust

Introducing SNA can be met with suspicion. Communities may fear that the information will be used against them (e.g., to reduce compensation claims) or that it imposes a technocratic framework that ignores local knowledge. Mitigation: involve community members from the outset in problem definition and node selection. Use participatory network mapping where villagers draw their own relationships, which builds ownership and trust. Explain the purpose clearly: to improve conflict prediction and benefit everyone, not to spy. Share preliminary results with communities and solicit feedback. Be transparent about limitations—this is not a surveillance tool. If possible, co-design interventions based on network findings with community leaders, ensuring that they have a say in how the data is used. In some cases, it may be wise to start with a non-sensitive topic (e.g., mapping information flow about crop disease) to demonstrate value before tackling conflict. Building trust takes time; allocate resources for relationship-building outside of data collection.

Mini-FAQ and Decision Checklist for Practitioners

This section condenses the guide into actionable answers to the most common questions we encounter from practitioners considering SNA for conflict prediction. It also includes a decision checklist to help teams assess their readiness and choose the right starting point. Use this as a quick reference when planning a project or troubleshooting an existing one. Each answer is grounded in the frameworks and experiences discussed earlier, and aims to provide clear guidance without oversimplifying the complexity.

FAQ

Q1: How large should my network be to get reliable predictions? There is no fixed number, but networks with less than 30 nodes often lack statistical power for predictive modeling. Aim for 100-200 nodes as a target. If your landscape is vast, focus on a smaller hotspot area rather than including everything. Remember that more nodes increase data collection effort, so balance with resources.

Q2: Can I use SNA if I have no conflict data yet? Yes, you can still use SNA to map baseline social structures and identify potential conflict hotspots based on network properties (e.g., high modularity, low cross-group ties). This is a proactive approach. Once conflict occurs, you can then validate and refine your model. Start with a pilot to build the network and monitor conflict over the next year.

Q3: What if relationships change rapidly (e.g., seasonal migration)? This is a challenge. Collect network data during the peak conflict season or at multiple time points to capture variation. For seasonal dynamics, consider a dynamic network model that allows ties to change over time. Alternatively, focus on stable relationships (kinship, land ownership) that persist, and treat transient ties as noise. Discuss with local experts which ties are stable.

Q4: How do I handle multiple types of ties (e.g., trust, information, resource sharing)? You can create separate networks for each tie type and then combine them into a multiplex network. Alternatively, choose the tie type most relevant to conflict (e.g., trust in conservation authorities) and focus on that. For simplicity, start with one tie type and add others later if needed. In our experience, trust ties are the strongest predictors of conflict cooperation or escalation.

Q5: What is the biggest mistake newcomers make? Trying to collect too much data too broadly without a clear question. This leads to analysis paralysis. Start small, with a well-defined problem (e.g., 'predict crop-raiding risk in village X'), build a simple model, and iterate. Also, neglecting to plan for updates: a one-time network snapshot is rarely useful for prediction beyond a year.

Decision Checklist

Before launching an SNA project, verify the following:

• Problem Definition: Have you explicitly defined the conflict type, geographic scope, and time horizon? (e.g., 'reduce livestock depredation by wolves in three villages over the next two years')
• Stakeholder Buy-in: Have you obtained support from community leaders, park authorities, and funding agencies? Do they understand what SNA can and cannot do?
• Data Availability: Do you have access to existing conflict records? Can you collect relational data within budget and time constraints? Is the ethical framework in place?
• Technical Capacity: Does your team include at least one person who can operate network analysis software? If not, have you budgeted for external support or training?
• Integration Plan: How will the results inform decision-making? Will there be regular meetings to review network updates? Is there a plan for updating data annually?
• Risk Mitigation: Have you considered privacy, community resistance, and data quality issues? Are you prepared to adjust the approach based on early feedback?

If you can answer 'yes' to most of these, you are ready to proceed. If not, address the gaps first. Starting with a small, well-designed pilot is better than a large, flawed study.

Synthesis and Next Actions: From Analysis to Impact

Throughout this guide, we have argued that human-wildlife conflict is a relational phenomenon that demands relational tools. Social Network Analysis offers a powerful lens to see the unseen boundaries of influence, trust, and communication that precede visible conflicts. By moving beyond traditional incident mapping, practitioners can identify leverage points, anticipate tensions, and design interventions that address root causes rather than symptoms. However, SNA is not a magic bullet; it requires careful implementation, ethical vigilance, and ongoing commitment. This final section synthesizes the key takeaways and outlines concrete next steps for readers ready to apply SNA in their own contexts. We also reflect on the broader implications for conservation practice and the future of conflict prediction.

The core insight is that network structure matters. A village with high modularity—isolated from neighboring communities and external agencies—is at greater risk than one with diverse, bridging ties. Centrality metrics highlight who can tip the balance toward peace or escalation. By monitoring these metrics over time, we can detect early warning signs: a sudden increase in betweenness centrality of a known agitator, or a decrease in network density after a conflict event. The workflow described in this guide—from boundary definition to model validation—provides a systematic path to turn these insights into action. We have emphasized starting small, building local capacity, and integrating SNA into existing routines to ensure sustainability. The economic analysis shows that the costs are modest compared to the benefits of preventing conflict, especially when considering the social and ecological costs of escalation.

As a next step, we encourage readers to begin with a pilot project in a single, well-defined area. Use the decision checklist to assess readiness, and allocate resources for training and community engagement. If you already have some conflict data, even better. Start by mapping the basic social network of stakeholders using simple surveys, and compute a few key metrics. Visualize the network and share it with local partners to get their interpretation. This initial step will build confidence and reveal unexpected patterns. Then, as you gather conflict data, build a simple predictive model and test its accuracy. Over time, refine and expand. Remember that the goal is not a perfect model but a useful one that improves decision-making. Document your process and share lessons learned with the conservation community; collective learning will accelerate the adoption of this approach. The future of conflict prediction lies not in more sophisticated algorithms but in understanding the human relationships that shape our coexistence with wildlife. SNA is a tool to do exactly that.