SSD-001

AHEAD – Anticipatory Humanitarian Action for Displacement (Foresight Model)

Download PDF
South Sudan Sub-Saharan Africa Low income Operational Deployment (Limited Rollout) Confirmed

Danish Refugee Council (DRC)

At a Glance

What it does Prediction (including forecasting) — Trend and shock forecasting
Who runs it Danish Refugee Council (DRC)
Programme AHEAD – Anticipatory Humanitarian Action for Displacement (Foresight Model)
Confidence Confirmed
Deployment Status Operational Deployment (Limited Rollout)
Key Risks Not assessed
Key Outcomes AHEAD triggered an early response in May 2024 in Akobo County, South Sudan, that prevented 2,800 people from being displaced — saving €6.
Source Quality 4 sources — Other, Academic journal article

The AHEAD (Anticipatory Humanitarian Action for Displacement) model is a machine learning system operated by the Danish Refugee Council (DRC) that forecasts forced displacement at the sub-district level in conflict-affected countries, including South Sudan, Somalia, and the Liptako-Gourma region in the Sahel covering Burkina Faso, Mali, and Niger. The system is designed to enable humanitarian actors to anticipate displacement crises and intervene before they escalate, shifting from reactive to anticipatory humanitarian response (DRC 2024).

The AHEAD model builds on DRC's earlier Foresight platform, which was originally established by DRC in partnership with IBM to scale up predictive analytics capacity in the humanitarian sector (Red Social Innovation 2020). The model is grounded in a theoretical framework focusing on the root causes and pre-disposing factors of displacement. It aggregates open-source data from 18 sources, including the World Bank development indicators, ACLED, UCDP, EMDAT, and UN agencies such as UNHCR, WFP, FAO, and IDMC, encompassing a total of 148 indicators. These indicators are organised into five categories: economy (e.g. unemployment, GDP, poverty), violence (e.g. civilian fatalities, number of conflict events), governance (e.g. corruption, access to public services, democracy), environment (e.g. food security, natural hazard events), and social/population (e.g. presence of vulnerable groups, urbanisation, population size). The categories were identified based on DRC field experience and standard fragility groupings used by the OECD and similar bodies (Red Social Innovation 2020).

Technically, the system employs an ensemble machine learning approach using gradient boosted trees and linear models to generate point forecasts. Model hyperparameters are determined through grid search. Each year-ahead forecast uses a separate model, training a set of ensemble models for different time horizons. Confidence intervals are generated using an empirical bootstrap method, with source error distributions produced through retrospective analysis. The model training data covers the period from 1995 onward (Red Social Innovation 2020). ACLED's conflict data are a critical input to the model, providing indicators on violent events, kidnappings, lootings, extremist attacks, and violence against civilians, which enable AHEAD to identify long-term conflict patterns and predict displacement trends. ACLED provides researcher-led data collection covering 244 countries and territories with at least seven years of historical data for each (ACLED 2024).

The model uses historical data from key displacement drivers to predict how many internally displaced persons (IDPs) there will be at a regional level, generating location-specific forecasts intended to improve operational responses. These forecasts are populated with open-source data on conflict, health, environment, food insecurity, numbers of IDPs, and income (DRC 2024). The model is hosted on an online platform where users can access the underlying data, view forecasts for different countries, and develop scenario-based forecasts of displacement (Red Social Innovation 2020). DRC provides public access to the AHEAD model through an interactive dashboard (DRC 2024).

In terms of deployment and documented outcomes, the Foresight model covers 26 displacement-producing countries. Of more than 150 forecasts produced across these countries, approximately 50 per cent have a margin of error of 10 per cent or below, and two-thirds have a margin of error of 15 per cent or below. The model can predict next year's displacement with an average margin of error as low as 6 per cent in its best-performing country. The model generally outperforms the accuracy of planning figures used in Humanitarian Response Plans (Red Social Innovation 2020). It has also been used by external partners to inform strategic planning, including in the Humanitarian Needs Overview (HNO) process for Central America and by OCHA CERF in funding allocation decision-making (Red Social Innovation 2020).

In May 2024, AHEAD triggered an early response in Akobo County, South Sudan, that prevented 2,800 people from being displaced, achieving a cost-effectiveness ratio of 6.60 euros saved for every 1 euro spent. This ratio is consistent with findings from a 2025 UN OCHA report concluding that every dollar invested in anticipatory action can yield up to seven dollars in avoided losses (ACLED 2024). The programme is funded by the European Civil Protection and Humanitarian Aid Operations (ECHO), the Swedish International Development Cooperation Agency (Sida), the Danish International Development Agency (Danida), and the Dutch Ministry of Foreign Affairs (DRC 2024).

Classifications follow the DCI AI Hub Taxonomy. Hover over field labels for definitions.

Social Protection Functions

Implementation/delivery chain
Assessment of needs/conditions + enrolment primary
SP Pillar (Primary) The social protection branch: social assistance, social insurance, or labour market programmes. Social assistance
Programme Name AHEAD – Anticipatory Humanitarian Action for Displacement (Foresight Model)
Programme Type The type of social protection programme, classified under social assistance, social insurance, or labour market programmes. View in glossary Emergency Cash Transfers
System Level Where in the social protection system the AI is applied: policy level, programme design, or implementation/delivery chain. View in glossary Implementation/delivery chain
Programme Description AHEAD (Anticipatory Humanitarian Action for Displacement) is a machine learning forecasting system operated by the Danish Refugee Council that predicts forced displacement at the sub-district level in conflict-affected countries including South Sudan, Somalia, and the Liptako-Gourma region. It uses an ensemble of gradient boosted trees and linear models trained on 148 indicators from 18 open-source datasets to generate location-specific displacement forecasts, enabling anticipatory humanitarian response. Funded by ECHO, Sida, Danida, and the Dutch Ministry of Foreign Affairs.
Implementation Type How the AI output is produced: Classical ML, Deep learning, Foundation model, or Hybrid. Affects validation, compute requirements, and governance profile. View in glossary Classical ML
Lifecycle Stage Current stage in the AI lifecycle, from problem identification through to monitoring, maintenance and decommissioning. View in glossary Monitoring, Maintenance and Decommissioning
Model Provenance Origin of the AI model: developed in-house, adapted from open-source, commercial/proprietary, or accessed via third-party API. View in glossary Developed in-house
Compute Environment Where the AI system runs: on-premise, government cloud, commercial cloud, or edge/device. View in glossary Not documented
Sovereignty Quadrant Classification of data and compute sovereignty: I (Sovereign), II (Federated/Hybrid), III (Cloud with safeguards), or IV (Shared Innovation Zone). View in glossary Not assessed
Data Residency Where the data used by the AI system is stored: domestic, regional, or international. View in glossary Not documented
Cross-Border Transfer Whether data crosses national borders, and if so, whether documented safeguards are in place. View in glossary Not documented
Decision Criticality The rights impact of the decision the AI supports. High criticality requires HITL oversight; moderate requires HOTL; low may operate HOOTL. View in glossary Moderate
Human Oversight Type Level of human involvement: Human-in-the-Loop (active review), Human-on-the-Loop (monitoring), or Human-out-of-the-Loop (periodic audit). View in glossary HOTL
Development Process Whether the AI system was developed fully in-house, through a mix of in-house and third-party, or fully by an external provider. View in glossary Fully in-house
Highest Risk Category The most significant structural risk source identified: data, model, operational, governance, or market/sovereignty risks. View in glossary Not assessed
Risk Assessment Status Whether a formal risk assessment, informal assessment, or independent audit has been conducted for this system. Not assessed

Risk Dimensions

Operational and system integration risks

Impact Dimensions

  • Human oversight protocol
  • Independent evaluation
CategorySensitivityCross-System LinkageAvailabilityKey Constraints
Administrative data from other sectorsNon-personalLinks data across multiple systemsCurrently available and usedRelies on open-source conflict event data (ACLED, UCDP) and disaster data (EMDAT); data quality depends on reporting coverage in conflict-affected areas where access is limited
Geospatial and remote sensing dataNon-personalLinks data across multiple systemsCurrently available and usedEnvironmental and food security indicators used at sub-national level; resolution and timeliness of environmental data may limit forecast granularity
Survey and census dataNon-personalLinks data across multiple systemsCurrently available and usedUses aggregate development indicators from World Bank, UN agencies (UNHCR, WFP, FAO), and IDMC; historical data from 1995 onward; indicator availability may vary by country and time period

ACLED (2024) 'Helping the Danish Refugee Council anticipate and prevent forced displacement', ACLED. Available at: https://acleddata.com/helping-danish-refugee-council-anticipate-and-prevent-forced-displacement (Accessed: 23 March 2026).

View source Other

Kjærum, A. and Madsen, B.S. (2025) 'Pushing the boundaries of anticipatory action using machine learning', Data & Policy, 7, e8. doi: 10.1017/dap.2024.88.

View source Academic journal article

Danish Refugee Council (2024) 'Anticipatory Humanitarian Action for Displacement (AHEAD) model', DRC. Available at: https://drc.ngo/what-we-do/innovation/anticipatory-action/ahead/ (Accessed: 23 March 2026).

View source Other

Red Social Innovation (2020) 'Foresight, using machine learning to predict refugees displacement', Red Social Innovation. Available at: https://red-social-innovation.com/en/solution/foresight-using-machine-learning-to-predict-refugees-displacement/ (Accessed: 23 March 2026).

View source Other
Deployment Status How far the system has progressed into real-world operational use, from concept/exploration through to scaled and institutionalised. View in glossary Operational Deployment (Limited Rollout)
Year Initiated The year the AI system was first initiated or development began. 2017
Scale / Coverage The scale and geographic or population coverage of the deployment. AHEAD operational in Liptako-Gourma (45 provinces across Burkina Faso, Mali, Niger) and South Sudan (Akobo County pilot May 2024). Foresight predecessor covers 26 countries with 214+ historical forecasts (Kjærum & Madsen 2025; Red Social Innovation 2020).
Funding Source The source(s) of funding for the AI system development and deployment. European Civil Protection and Humanitarian Aid Operations (ECHO); Swedish International Development Cooperation Agency (Sida); Danish International Development Agency (Danida); Dutch Ministry of Foreign Affairs
Technical Partners External technology vendors, academic partners, or development partners involved. Custom ML pipeline developed by DRC Foresight team; non-proprietary algorithms using open-source data
Outcomes / Results AHEAD triggered an early response in May 2024 in Akobo County, South Sudan, that prevented 2,800 people from being displaced — saving €6.60 for every €1 spent ACLED; Model can predict next year's displacement with average margin of error down to 6% in best-performing countries; approximately 50% of 150+ forecasts have margin of error of 10% or below Red-social-innovation

How to Cite

DCI AI Hub (2026). 'AHEAD – Anticipatory Humanitarian Action for Displacement (Foresight Model)', AI Hub AI Tracker, case SSD-001. Digital Convergence Initiative. Available at: https://socialprotectionai.org/use-case/SSD-001 [Accessed: 1 April 2026].

Change History

Created 30 Mar 2026, 08:41
by v2-import (import)