Enhancing Agent-Based Models with Artificial Intelligence for Complex Decision Making

Shaheen A. Abdulkareem

Research output: ThesisPhD Thesis - Research UT, graduation UT

401 Downloads (Pure)

Abstract

In 1953 the Netherlands saw the worst devastating coastal flood of the last century. The government stepped in and pledged that this should never happen again, brining science and technology, engineering and construction to create the most sophisticated and reliable flood defense protection system in the world. However, the government did not realize that it stepped into a vicious circle: the better the land in coastal areas is protected, the more attractive it becomes for people to locate, the higher the demand for and economic value of these flood prone areas are, and, therefore, the more government needs to invest in the protection of these areas. This could have been avoided if there was a better understanding of the feedbacks and relationships between the macro-scale governmental policies and the micro-scale individual homeowners behavior in a land market.

Coastal zone management policy in the Netherlands aims at reducing risk, which is defined as the probability of a disaster multiplied by economic damage. Direct economic damage depends on land patterns and value of properties under risk, which, in turn, are the outcomes of individual microeconomic interactions in a land market. Governmental policy might use instruments (e.g. taxes, insurance, educational programs) to affect individual motivations and rules of local interaction in order to direct land markets in coastal areas towards desired macroscopic outcomes (e.g. more safe allocations).
However, the transition from micro-behavior to macro-measures used by policy-makers is discontinuous, non-linear and may be associated with new, emergent effects and properties. Lack of understanding of micro-foundations of micro-phenomena (such as total economic value of the area and spatial pattern of location) can make coastal zone management and spatial planning policies inefficient and unpredictable.

The main goal of this thesis is to get insights into how aggregated economic phenomena in space emerge from interactions of individual economic agents in a land market.
Specifically, this study seeks to indentify traceable connections between micro and macroeconomic scales exploring a hypothetic city, which replicates the structure and complexity of a typical Dutch coastal town. Although the application is specific, the model is flexible and can be used in many other cases where economic behavior needs to be modeled in a spatially explicit way that involves consideration of environmental amenities, natural hazards and spatial externalities. The conventional economic approach assumes a representative rational agent and a unique equilibrium in the system. To accomodate more spatial and agent heterogeneity and to allow the study to be spatially explicit, this thesis adopts an agent-based approach, which helps to understand the effects of relaxing some of the conventional economic assumptions and their implications for coastal risk management policy.
Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • Department of Governance and Technology for Sustainability
Supervisors/Advisors
  • Filatova, Tatiana , Supervisor
  • Mustafa, Yaseen Taha, Co-Supervisor
  • Augustijn, Ellen-Wien, Co-Supervisor
Thesis sponsors
Award date10 Apr 2019
Place of PublicationEnschede
Publisher
Print ISBNs978-90-365-4748-2
DOIs
Publication statusPublished - 10 Apr 2019

Keywords

  • Risk perception
  • Machine learning
  • Bayesian networks
  • Infectious diseases
  • Adaptation
  • Human behaviour

Fingerprint

Dive into the research topics of 'Enhancing Agent-Based Models with Artificial Intelligence for Complex Decision Making'. Together they form a unique fingerprint.

Cite this