Predicting forced displacement using a generalised and automated agent-based simulation

Abstract

Within the last decades, international migration demonstrated an escalating growth with more than 68 million people forcibly displaced worldwide. Forced displacement has a huge impact on society today as 40 million people internally displaced within their home country and 25.4 millionrefugeesfledtoneighbouringcountries. Forcibly displaced people face several concerns, namely, the choice to stay or flee, the choice to flee internally or across borders, and the choice of destination. These decisions are often based on economic, social and/or political push and pull factors in origin and destination countries. Existing international migration theories frequently cover specific aspects of migration processes, such as why human migration occurs and what effects it has on economies. However, the combination of all the factors and reasons for human movement requires expertise from various disciplines at once. Moreover, existing migration theories are not extensive enough for practical applications, statistical methods are outdated, and usually, do not account for forced population movements. To fulfil gaps within forced displacement predictions, we use computational models as they can contribute to a better understanding of forced displacement patterns and have potential due to their reduced ethical burden. We propose a generalised simulation development approach (SDA) to predict forced population movements in conflict regions. Our SDA consists of a systematic set of phases to build agent-based simulations, which includes a generic model to define a real system problem, and simulation development and validation for situation-specific scenarios. We also synthesise data from UNHCR, ACLED and Bing Maps to build and validate agent-based simulations of three major African conflicts, namely Burundi, Central African Republic and Mali, and predict the distribution of incoming forced migrants across destination camps. Our simulations consistently predict more than 75% of the population arrivals in camps correctly after the first 12 days. Our agent-based simulation tool can help save migrants’ lives by allowing governments and NGOs to conduct a better-informed allocation of humanitarian resources. Few researchers have investigated the effects of policy decisions, such as camp capacity changes, camp and border closures and forced redirection, on forced population movements. To make such a study accurate and feasible in terms of human effort, we automate our generalised SDA by introducing and applying the FabFlee automation toolkit. We use our automated SDA to analyse the South Sudan crisis by incorporating two capacity changes to Adjumani camp, a border closure between South Sudan and Uganda, and forced redirection between Ethiopian camps. We find that a reduction in camp capacity induces up to 16% fewer forced population arrivals while an increase in camp capacity results in a limited increase in forced population arrivals (< 4%) at the destination camps. In addition, border closure results in 40% fewer force population arrivals and an increasingly long travel journey to other camps. There is also a lingering effect in prolonged force population journey times once a border is again reopened and a clear boost in forced population arrivals when forced population are redirected to a reduced number of camps with larger capacities. To the best of our knowledge, we are the first to conduct such an investigation for forced displacement conflict situations.