Publication:
A multiagent framework for the co-evolution of social service delivery networks and population dynamics

Thumbnail Image
Authors
Herrera-Restrepo, Oscar A.
Embargoed Until
Advisor
Medina-Borja, Alexandra
College
College of Engineering
Department
Department of Industrial Engineering
Degree Level
M.S.
Publisher
Date
2011
Abstract
A multiagent system for simulating the co-evolution of social service delivery networks is presented. In the first part, 63 potential variables were analyzed to identify significant predictors in the construction of behavior rules for each agent. An agent in the model represents one chapter or field unit of the network. These rules were focused on four (4) key responses. Thirteen (13) significant predictors were found through CHAID data mining techniques. Predictors demonstrated that chapter survival over time is mainly affected by financial, demographic, and service variables. Agent behavior rules were established based on CHAID results. These rules of behavior summarize a set of simple conditions to achieve certain states. Despite this simplicity, the emergent behavior of the model reproduces most of the co-evolutionary patterns of the case study site. An experimentation process was performed and the results confirmed the significant effects of factors such as: network configuration, market demographic characteristics (i.e. rate of migration, racial concentrations and median household income) over responses such as number of services, total donations, percentage of active chapters (i.e. in operation), and proportion of volunteers who remain in the network over time.

Un sistema multi-agente para simular la co-evolución de redes prestadoras de servicios sociales es presentado. En la primera parte, 63 potenciales variables fueron analizadas con el fin de identificar los predictores significativos en la construcción de las reglas de comportamiento de cada agente. Dichas reglas fueron enfocadas a 4 respuestas. Trece (13) variables predictoras significativas fueron encontradas a través de árboles de clasificación por el método CHAID. Los predictores demostraron que el status de un capitulo en el tiempo es principalmente afectado por variables financieras, demográficas y de servicio. Las reglas de comportamiento de los agentes fueron establecidas basadas en los resultados de la técnica CHAID. Las reglas de comportamiento resumen un conjunto de condiciones simples para alcanzar determinados estados. A pesar de dicha simplicidad, el comportamiento emergente del modelo reproduce la mayoría de los patrones co-evolutivos de la red caso de estudio. Los resultados experimentales confirmaron los efectos significativos de factores como el tipo de red, la tasa de migración, y la interacción de ciertas concentraciones raciales y el ingreso medio sobre respuestas como: el número de servicios prestados, las donaciones totales, la razón de capítulos activos y la razón de voluntarios.
Keywords
Multiagent system,
Social service delivery networks
Cite
Herrera-Restrepo, O. A. (2011). A multiagent framework for the co-evolution of social service delivery networks and population dynamics [Thesis]. Retrieved from https://hdl.handle.net/20.500.11801/1495