AI and the Politics of Humanitarian Prediction - TRT World Research Centre
Strategic Argument and Areas of DebateThe humanitarian transition towardanticipatory actionmediated bypredictive analyticsrisks institutionalising adata-driven hierarchy of suffering, where the pursuit. . .
Strategic Argument and Areas of DebateThe humanitarian transition towardanticipatory actionmediated bypredictive analyticsrisks institutionalising adata-driven hierarchy of suffering, where the pursuit oftechnological efficiencyinadvertently compromises theneutralityandimpartialityof life-saving interventions. This strategic shift creates a dilemma between the speed ofalgorithmic responseand the essential requirement forhumanitarian accountabilityin protecting the most vulnerable populations. Executive SummaryTheTRT World Research Centreanalysis explores howartificial intelligenceis fundamentally reshaping humanitarian governance by transitioning operations from reactive funding appeals to proactive, risk-based preparedness. Central to this evolution are initiatives such as theWorld Bankrefugee forecasting model inUgandaandUNHCR‘sCareLinecommunication system, which aim to optimise resource allocation for displaced persons from theDemocratic Republic of the CongoandSouth Sudan. However, the report finds that these technologies introduce systemic risks, includingautomation bias,technological dependencyon private firms, and the potentialdual-useof sensitive data for state surveillance or border enforcement. Analytical Framework and Key DriversAnticipatory Humanitarian Governance Transition: A structural shift from responding to crises after they unfold to managing potential risks through predictive data analysis and pre-agreed financing. Algorithmic Construction of Vulnerability: The process by whichpredictive modelsand specific datasets determine which crises become visible on institutional dashboards and which populations receive priority aid. Digital Gatekeeping and Information Risks: The deployment ofAI-supported chatbotsand automated platforms that manage refugee access toasylumprocedures, legal information, and essential services. Technological Sovereignty and Private Dependence: The growing reliance of international organisations on proprietary algorithms and cloud infrastructure owned by commercial technology companies, affecting institutional autonomy. Strategic Assessment & Empirical FindingsTheWorld Bankforecasting initiative inUgandaprocesses over90 variables—including conflict dynamics and food prices—to predict refugee arrivalsfour to six monthsin advance. UNHCR‘s testing of predictive analytics covers multiple regions, including internaldisplacementinSomalia, refugee return patterns inUkraine, and mixed-movement routes inLatin Americaand theSahel. The model utilised inUgandawas validated usingUNHCRdaily arrival data spanning the decade between2014 and 2023to ensure predictive reliability. AI-enabled systems likeCareLinein theAsia and the Pacificregion automate the classification of thousands of inquiries, though they risk producing “plausible fabrications” if not properly supervised. Data-scarce environmentslead to the systematic marginalisation of rural populations and linguistic minorities who lack the digital footprint required for algorithmic visibility. TheICRCidentifies significant protection risks when data collected for humanitarian assistance is repurposed formilitary operationsordeportationby state actors. Geopolitical Trajectories & Policy RisksUgandaand other major host nations face a strategic dependency onWorld Bankforecasts that may either delay funding through underestimates or divert resources from less visible but acute crises. TheICRCandUNHCRmust navigate the high-stakes risk ofdata repurposing, where information meant for protection is weaponised by state security apparatuses forborder enforcement. OCHAfaces the institutional challenge of standardisingpredictive analyticsthrough peer-review frameworks to preventautomation biasfrom overriding local, qualitative field assessments. Critical Policy Questions & ResponsesQuestion 1Why does the shift toward anticipatory humanitarian action challenge traditional international funding models?Answer:Traditional models are reactive, initiating appeals only after a crisis escalates, whereas theWorld BankandOCHAnow promote triggers based on predictive probabilities. This requires pre-arranged financing mechanisms that can be activated months before displacement occurs, shifting the focus from emergency response to systemic preparedness. Question 2How do predictive models in data-scarce environments impact the visibility of vulnerable populations?Answer:In conflict zones where digital infrastructure has collapsed, models relying on social media or mobile data overrepresent digitally visible communities while making rural or marginalised groups invisible. This creates a risk whereartificial intelligenceallocates resources based on data density rather than actual humanitarian need. Question 3What are the strategic consequences of humanitarian organisations’ dependence on private-sector AI infrastructure?Answer:Reliance on proprietary models and cloud systems owned by commercial firms limits transparency and compromisestechnological sovereignty. This transition redistributes authority, as external actors increasingly shape the analytical foundations upon whichUNHCRand other agencies make life-saving decisions. Question 4Why is the “human-in-the-loop” requirement often insufficient in preventing algorithmic errors?Answer:Staff members often suffer fromautomation bias, placing excessive confidence in machine-generated risk scores, especially when they lack the technical authority to challenge complex models. Without meaningful empowerment and local knowledge integration, human oversight becomes a symbolic gesture rather than a robust safeguard. Key Actors and Systemic DynamicsWorld Bank→ Supports →Ugandarefugee responseUNHCR→ Deploys →CareLinesystemOCHA→ Regulates →Predictive model standardsICRC→ Challenges →AI technical neutralityPredictive Models→ Influence →Resource allocationCommercial Tech Firms→ Control →Cloud infrastructureConflict Indicators→ Input for →Displacement forecastingData Scarcity→ Weakens →Rural visibilityAutomation Bias→ Constraints →Human oversightGenerative AI→ Creates risk of →Plausible fabricationDownload the Policy Outlook