Policy and Complex Systems
ing ( Bagherpour , Donaldson , & Scharpnick , 2016 ). Our research seeks to fill this gap through the modeling and computational exploration of human trafficking , with a particular focus on the vulnerable populations of migrant flows and the role of trust in detecting trafficking victims .
Model
Agent-based models are increasingly recognized as an effective tool for simulating systems that are characterized by complex interactions among their components . The success of the victim-centered approach is strongly dependent on the evolving dynamics of trust between migrants and authorities . We use an agent-based model that accounts for heterogeneous characteristics of migrants who interact spatially and over networks for scenarios of state immigration policies . Our model instantiates a migrant population and seeds the agent set with some number of interspersed trafficking victims . We measure the effectiveness of the victim-centered approach by its detection success of these seeded victims .
The model is a highly stylized and abstract representation of a migrant population . The model is developed on a grid , with each cell holding an agent object that represents migrants . The grid dimensions are 41 × 41 ; thus , all model runs are executed with a total migrant population of 1,681 . The motivation for using a grid layer was to use the Moore neighborhood ( Gilbert & Troitzsch , 2005 ) as a representation of spatial neighbor influences within the densely packed migrant population on the country borders of a state — an individual is assumed to have greater word-of-mouth communication opportunities with those who are directly adjacent to them . For this research , our model will be in context of the European migration crisis and will attempt to calibrate model parameters to represent a migrant population on the German border .
In the context of migration flows , networks play a key role in the problem of human trafficking . Of particular note are the local kin- and community-based migration networks which migrants rely on to identify potential destinations based on economic opportunity and / or greater safety and security ( Salt , 2000 ). We also wanted to capture network-based communications for these migrant populations ( e . g ., familial networks , country of origin networks ), and we approximate these relationships by linking all of our migrant agents in the population to each other on an underlying preferential attachment network . We assume that the clustering characteristics for this class of networks provide a sufficient proxy in this research to represent real-world migrant networks — some individuals have greater connectivity within the population than others do .
Figure 2 shows a screenshot of our model interface , which features user controls for adjusting population sizes and model parameters for the migrant agents and policies of the state . The underlying preferential attachment network is not visualized . Colors represent the cooperation strategy of each agent — to cooperate ( yellow ) or deceive ( red ). For example , from a trafficking victim agent ’ s perspective , we perceive “ cooperate ” as representing an intention to self-identify , and “ deceive ” as the intention to not . Shapes indicate if they have interacted with the state yet , and the interaction outcome ( accepted or rejected ). Using colors and shapes , the researcher may observe a visual representation of the evolution in the system over time and how cooperation may build and / or erode over time . The example realization depicted in Figure 2 shows clustered communities of cooperators persisting among a larger system population of deceivers .
151