Network reciprocity describes the emergence of cooperative behaviour where interactions are constrained by incomplete network connectivity. It has been widely studied as an enabling mechanism for the emergence of cooperation and may be of particular interest in explaining cooperative behaviours amongst unrelated individuals or in organisms of lower cognitive abilities. Research in this area has been galvanised by the finding that heterogeneous topology promotes cooperation. Consequently there has been a strong focus on scale-free networks; however, such networks typically presuppose formative mechanisms based on preferential attachment, a process which has no general explanation. This assumption may give rise to models of cooperation that implicitly encode capabilities only generally found in more complex forms of life, thus constraining their relevance with regards to the real world. By considering the connectivity of populations to be dynamic, rather than fixed, cooperation can exist at lower levels of heterogeneity. This thesis demonstrates that a model of network fluctuation, based on random rather than preferential growth, supports cooperative behaviour in simulated social networks of only moderate heterogeneity, thus overcoming difficulties associated with explanations based on scale-free networks. In addition to illustrating the emergence and persistence of cooperation in existing networks, we also demonstrate how cooperation may evolve in networks during their growth. In particular our model supports the emergence of cooperation in populations where it is originally absent. The combined impact of our findings increases the generality of reciprocity as an explanation for cooperation in networks.

Intensive human-environment interactions are taking place in Midwestern agricultural systems. An integrated modeling framework is suitable for predicting dynamics of key variables of the socio-economic, biophysical, hydrological processes as well as exploring the potential transitions of system states in response to changes of the driving factors. The purpose of this dissertation is to address issues concerning the interacting processes and consequent changes in land use, water balance, and water quality using an integrated modeling framework. This dissertation is composed of three studies in the same agricultural watershed, the Clear Creek watershed in East-Central Iowa. In the first study, a parsimonious hydrologic model, the Threshold-Exceedance-Lagrangian Model (TELM), is further developed into RS-TELM (Remote Sensing TELM) to integrate remote sensing vegetation data for estimating evapotranspiration. The goodness of fit of RS-TELM is comparable to a well-calibrated SWAT (Soil and Water Assessment Tool) and even slightly superior in capturing intra-seasonal variability of stream flow. The integration of RS LAI (Leaf Area Index) data improves the model's performance especially over the agriculture dominated landscapes. The input of rainfall datasets with spatially explicit information plays a critical role in increasing the model's goodness of fit. In the second study, an agent-based model is developed to simulate farmers' decisions on crop type and fertilizer application in response to commodity and biofuel crop prices. The comparison between simulated crop land percentage and crop rotations with satellite-based land cover data suggest that farmers may be underestimating the effects that continuous corn production has on yields (yield drag). The simulation results given alternative market scenarios based on a survey of agricultural land owners and operators in the Clear Creek Watershed show that, farmers see cellulosic biofuel feedstock production in the form of perennial grasses or corn stover as a more risky enterprise than their current crop production systems, likely because of market and production risks and lock in effects. As a result farmers do not follow a simple farm-profit maximization rule. In the third study, the consequent water quantity and quality change of the potential land use transitions given alternative biofuel crop market scenarios is explored in a case study in the Clear Creek watershed. A computer program is developed to implement the loose-coupling strategy to couple an agent-based land use model with SWAT. The simulation results show that watershed-scale water quantity (water yield and runoff) and quality variables (sediment and nutrient loads) decrease in values as switchgrass price increases. However, negligence of farmers risk aversions towards biofuel crop adoption would cause overestimation of the impacts of switchgrass price on water quantity and quality.

This thesis consists of four self-contained papers related to banking, credit markets and financial stability.    Paper [I] presents a credit market model and finds, using an agent based modeling approach, that credit crunches have a tendency to occur; even when credit markets are almost entirely transparent in the absence of external shocks. We find evidence supporting the asset deterioration hypothesis and results that emphasize the importance of accurate firm quality estimates. In addition, we find that an increase in the debt’s time to maturity, homogenous expected default rates and a conservative lending approach, reduces the probability of a credit crunch. Thus, our results suggest some up till now partially overlooked components contributing to the financial stability of an economy.     Paper [II] derives an econometric disequilibrium model for time series data. This is done by error correcting the supply of some good. The model separates between a continuously clearing market and a clearing market in the long-run such that we are able to obtain a novel test of clearing markets. We apply the model to the Swedish market for short-term business loans, and find that this market is characterized by a long-run nonmarket clearing equilibrium.    Paper [III] studies the risk-return profile of centralized and decentralized banks. We address the conditions that favor a particular lending regime while acknowledging the effects on lending and returns caused by the course of the business cycle. To analyze these issues, we develop a model which incorporates two stylized facts; (i) banks in which lendingdecisions are decentralized tend to have a lower cost associated with screening potential borrowers and (ii) decentralized decision-making may generate inefficient outcomes because of lack of coordination. Simulations are used to compare the two banking regimes. Among the results, it is found that even though a bank group where decisions are decentralizedmay end up with a portfolio of loans which is (relatively) poorly diversified between regions, the ability to effectively screen potential borrowers may nevertheless give a decentralized bank a lower overall risk in the lending portfolio than when decisions are centralized.    In Paper [IV], we argue that the practice used in the valuation of a portfolio of assets is important for the calculation of the Value at Risk. In particular, a seller seeking to liquidate a large portfolio may not face horizontal demand curves. We propose a partially new approach for incorporating this fact in the Value at Risk and Expected Shortfall measures and in an empirical illustration, we compare it to a competing approach. We find substantial differences.

This thesis focuses on stochastic processes and some of their properties are investigated which are necessary to determine the tools, the extremal index and the extremogram. Both mathematical tools measure extremal dependency within random time series. Two different models are introduced and related properties are discussed. The probability function of the Agent based model is surveyed explicitly and strong stationarity is proven. Data sets for both processes are simulated and clustering of the data is investigated with two different methods. Finally an estimation of the extremogram is used to interpret dependency of extremes within the data.

Zugl.: Kassel, Univ., Diss.

We construct and examine a model of adoption of a product or policy using, firstly, a system of differential equations and then secondly, through simulation, an agent- based model. Awareness must come before adoption, and we model this as a simple epidemic type model, where information is spread through advertising and contact with other agents in the population. Adoption is then conditional on awareness and occurs only if the agent finds the perceived cost acceptable. After simulating the system using an agent-based model, we introduce heterogeneity through the model parameters, which are then considered individual attributes and include influence rates, effectiveness of advertising, price sensitivity, and speed of adoption. We also examine the effects of various network topologies by organizing individuals into lattice and preferential attachment networks. From there, we add two extra components to the adoption mechanism by introducing a social influence factor by which an agent can be influenced by the adoption patterns of their neighbourhood, as well as a green factor, which assumes an environmental product or policy being adopted and is the likelihood that an individual will adopt based on environmental reasons alone.

Our epidemiology research found that the incident and prevalent rates for Diabetes mellitus (DM) and Diabetic End Stage Renal Disease (DM-ESRD) were at rise in Saskatchewan between year 1980 and 2005. Combining concerns regarding the rising trends reported by research studies with the concerns of the significant health and financial burden imposed by DM-ESRD on individuals and societies, we sought to project the number of DM-ESRD patients in Saskatchewan up to year 2025 with the cost required for caring for those patients. An agent-based model (ABM) is developed to simulate DM to ESRD progression, treatments for DM-ESRD patients, and the assessments and waiting list processes preparing patients for kidney transplants. The model parameters were estimated from a wide variety of data sources. The agent based modeling approach is chosen for projections regarding the DM-ESRD situation in Saskatchewan because of its advantage in capturing heterogeneities of individual patients, ability to retain biographical information on patients, capacity to capture time-varying competing risks, better presentations features and easy integration with existing models built in either agent based or System Dynamic methods. The approach was also attractive due to its flexibility for future expansion to represent social networks. The model projects the incident and prevalent case count, cost, and person years lived for the DM-ESRD population in Saskatchewan between year 1980 and 2025. The projections captured the great challenges brought by the fast growing number of DM-ESRD patients and substantial cost associated with managing the disease. In addition to producing projection results, the research presented here demonstrates how the model can be used by policy makers to experiment and evaluate different policy/interventions in a safe context. By capturing both the individual level records and population level statistics, the model provide a wealth of data for detailed analysis, which can help health policy makers gain insights in the current and future diabetic-ESRD situation in the province, aiding in resources planning for managing the fast-growing ESRD population and the growing need for dialysis services.

Land use and cover change (LUCC) is increasingly recognised as one of the most visible impacts of humans on nature. In rural areas, most of the observed LUCC is associated with agricultural activities. This has traditionally been attributed to the interplay of the socio-economic and political milieu, and the opportunities and constraints arising from the climatic conditions and physical attributes of land. Although there is no doubt that these factors influence farmer decisions, the mosaic of farming systems suggests that farmers do not always behave uniformly, even in areas with comparable socio-economic and environmental conditions. While the multi-facetted and varying nature of farmer decision-making is considered to be established knowledge in rural sociology, it is often neglected in LUCC models that typically describe it as homogeneous and rational in economic terms. This thesis presents an application of mixed-method social survey which aims at improving the representation of the diversity and complexity of farmer decision-making process in LUCC models. Different data collection methods (in-depth, semi-structured interviews, questionnaire) and analyses (thematic analysis, principal components analysis, cluster analysis, choice-based conjoint analysis) were used complementarily to identify the factors that facilitate or constrain farmer participation in environmental management practices (a), to identify the dominant farmer profiles (b) and to assess farmer preferences that influence land use decisions (c). Data collection was conducted in a study area located in the Canton of Aargau, Switzerland, where there is limited knowledge about farmer decision-making drivers and actions. Research findings were used to empirically inform an agent-based model that simulates farmer decisions. Paremeterised storylines were used to explore farmer decisions in alternative futures. An advanced and context-specific representation of human agents in modeling frameworks can make LUCC models valuable tools both for landscape analysis and policy making. In the face of new policy reforms, this thesis contributes to the achievement of this objective, by presenting an approach to explore and organize the heterogeneity of farmer behaviour and to make this usable in agent-based modeling frameworks.

This research addresses the design and development of a decision support tool to provide healthcare policy makers with insights and feedback when evaluating proposed patient flow and infection mitigation and control strategies in the emergency department (ED). An agent-based modeling (ABM) approach was used to simulate EDs, designed to be tuneable to specific parameters related to specification of topography, agent characteristics and behaviours, and the application in question. In this way, it allows for the user to simulate various ‘what-if’ scenarios related to infection spread and patient flow, where such policy questions may otherwise be left “best intent open loop” in practice. Infection spread modeling and patient flow modeling have been addressed by mathematical and queueing models in the past; however, the application of an ABM approach at the level of an institution is novel. A conjecture of this thesis is that such a tool should be augmented with Machine Learning (ML) technology to assist in performing optimization or search in which patient flow and infection spread are signals or variables of interest. Therefore this work seeks to design and demonstrate a decision support tool with ML capability for optimizing ED processes. The primary contribution of this thesis is the development of a novel, flexible, and tuneable framework for spatial, human-scale ABM in the context of a decision support tool for healthcare policy relating to infection spread and patient flow within EDs . The secondary contribution is the demonstration of the utility of ML for automatic policy generation with respect to the ABM tool. The application of ML to automatically generate healthcare policy in concert with an ABM is believed to be novel and of emerging practical importance. The tertiary contribution is the development and testing of a novel heuristic specific to the ML paradigm used: Genetic Programming (GP). This heuristic aids learning tasks performed in conjunction with ABMs for healthcare policy. The primary contribution is clearly demonstrated within this thesis. The others are of a more difficult nature; the groundwork has been laid for further work in these areas that are likely to remain open for the foreseeable future.

This research addresses the design and development of a decision support tool to provide healthcare policy makers with insights and feedback when evaluating proposed patient flow and infection mitigation and control strategies in the emergency department (ED). An agent-based modeling (ABM) approach was used to simulate EDs, designed to be tuneable to specific parameters related to specification of topography, agent characteristics and behaviours, and the application in question. In this way, it allows for the user to simulate various ‘what-if’ scenarios related to infection spread and patient flow, where such policy questions may otherwise be left “best intent open loop” in practice. Infection spread modeling and patient flow modeling have been addressed by mathematical and queueing models in the past; however, the application of an ABM approach at the level of an institution is novel. A conjecture of this thesis is that such a tool should be augmented with Machine Learning (ML) technology to assist in performing optimization or search in which patient flow and infection spread are signals or variables of interest. Therefore this work seeks to design and demonstrate a decision support tool with ML capability for optimizing ED processes. The primary contribution of this thesis is the development of a novel, flexible, and tuneable framework for spatial, human-scale ABM in the context of a decision support tool for healthcare policy relating to infection spread and patient flow within EDs . The secondary contribution is the demonstration of the utility of ML for automatic policy generation with respect to the ABM tool. The application of ML to automatically generate healthcare policy in concert with an ABM is believed to be novel and of emerging practical importance. The tertiary contribution is the development and testing of a novel heuristic specific to the ML paradigm used: Genetic Programming (GP). This heuristic aids learning tasks performed in conjunction with ABMs for healthcare policy. The primary contribution is clearly demonstrated within this thesis. The others are of a more difficult nature; the groundwork has been laid for further work in these areas that are likely to remain open for the foreseeable future.