Showing papers by "Gary Polhill published in 2014"

Standardised and transparent model descriptions for agent-based models

Abstract: Agent-based models are helpful to investigate complex dynamics in coupled human-natural systems. However, model assessment, model comparison and replication are hampered to a large extent by a lack of transparency and comprehensibility in model descriptions. In this article we address the question of whether an ideal standard for describing models exists. We first suggest a classification for structuring types of model descriptions. Secondly, we differentiate purposes for which model descriptions are important. Thirdly, we review the types of model descriptions and evaluate each on their utility for the purposes. Our evaluation finds that the choice of the appropriate model description type is purpose-dependent and that no single description type alone can fulfil all requirements simultaneously. However, we suggest a minimum standard of model description for good modelling practice, namely the provision of source code and an accessible natural language description, and argue for the development of a common standard. Agent-based models can be documented with different types of model descriptions.We differentiate eight purposes for which model descriptions are used.We evaluate the different description types on their utility for the different purposes.No single description type alone can fulfil all purposes simultaneously.We suggest a minimum standard by combining particular description types.

Advances in Computational Social Science and Social Simulation

Abstract: Readers will find results of recent research on computational social science and social simulation economics, management, sociology,and history written by leading experts in the field. SOCIAL SIMULATION (former ESSA) conferences constitute annual events which serve as an international platform for the exchange of ideas and discussion of cutting edge research in the field of social simulations, both from the theoretical as well as applied perspective, and the 2014 edition benefits from the cross-fertilization of three different research communities into one single event. The volume consists of 122 articles, corresponding to most of the contributions to the conferences, in three different formats: short abstracts (presentation of work-in-progress research), posters (presentation of models and results), and full papers (presentation of social simulation research including results and discussion). The compilation is completed with indexing lists to help finding articles by title, author and thematic content. We are convinced that this book will serve interested readers as a useful compendium which presents in a nutshell the most recent advances at the frontiers of computational social sciences and social simulation research

Combining diverse data sources for cedss, an agent-based model of domestic energy demand

Abstract: CEDSS (Community Energy Demand Social Simulator) is an empirical agent-based model designed and built as part of a multi-method social science project investigating the determinants of domestic energy demand. Ideally, empirical modellers, within and beyond social simulation, would prefer to work from an integrated dataset, gathered for the purposes of developing the model. In practice, many have to work with less than ideal data, often including processed data from multiple sources external to the project. Moreover, what data will be required may not be clear at the start of the project. This paper describes the approach to dealing with these factors taken in developing CEDSS, and presents the completed model together with an outline of the calibration and validation procedure used. The discussion section draws together the most distinctive features of empirical data collection, processing and use for and in CEDSS, and argues that the approach taken is sufficiently robust to underpin the model’s purpose – to generate scenarios of domestic energy demand to 2049.

Empirical agent-based modelling of everyday pro-environmental behaviours at work

Abstract: We report on agent-based modelling work in the LOCAW project (Low Carbon at Work: Modelling Agents and Organisations to Achieve Transition to a Low Carbon Europe). The project explored the effectiveness of various backcasting scenarios conducted with case study organisations in bringing about pro-environmental change in the workforce in the domains of transport, energy use and waste. The model used qualitative representations of workspaces in formalising each scenario, and decision trees learned from questionnaire responses to represent decision-making. We describe the process by which the decision trees were constructed, noting that the use of decision trees in agent-based models requires particular considerations owing to the potential use of explanatory variables in model dynamics. The results of the modelling in various scenarios emphasise the importance of structural environmental changes in facilitating everyday pro-environmental behaviour, but also show there is a role for psychological variables such as norms, values and efficacy. As such, the topology of social interactions is a potentially important driver, raising the interesting prospect that both workplace geography and organisational hierarchy have a role to play in influencing workplace pro-environmental behaviours.

Using Genetic Algorithms to Fit Species and Habitat Parameters for Modelling the Effect of Climate Change on Species Distributions with Stochastic Patch Occupancy Models

Abstract: Standard approaches to modelling the effect of climate change on species distributions model a direct link between climatic and other biophysical variables and species occupancy. Though these provide a reasonable estimate for the effects of climate change on species distributions in the future, there are a number of issues with these approaches that fail to account for dynamic landscape interactions. For example, the mass occupancy effect means that species may be observed in unsuitable habitat patches surrounding a well-populated area of highly suitable patches. Conversely, a highly suitable area may be too disconnected from other suitable patches to allow long-term species occupancy. The degree of isolation, however, is not fixed but depends on landscape dynamics. The dynamics of patch occupancy can be modelled using tools such as stochastic patch occupancy models, which use a habitat variable to represent the suitability of each patch for species occupancy, and include the computation of landscape habitat connectivity in determining whether a patch is occupied. Fitting biophysical and climatic variables to habitat suitability and using stochastic patch occupancy models to model the distributions offers a means to account for issues such as the mass occupancy effect, which can be partly responsible for autocorrelated error in purely statistical approaches, and also allows us to account for the effect of the rate and variability with which climatic variables change when modelling future species distributions. However, the method poses a more significant computing challenge to finding the fitting parameters. We report and reflect on preliminary work using genetic algorithms to search for these parameters.

Qualitative spatial representation in agent-based models

Abstract: One of the advantages of agent-based models as simulations of social systems is the ease with which it is possible to spatially embed the agents and their interactions. Spatially explicit representations in agent-based models most typically take the form of raster-based representations in which the space is represented as a grid of squares. More recently, vectorbased representations have been used, usually importing data for the polygons from geographical information systems (GIS). However, for some models, what matters about the space for the purposes of simulation is less the quantitative spatial relationships among entities (e.g. area, distance or direction) than the qualitative relations these quantitative data are used to determine: neighbourhood, and accessibility (which is a general term covering movement and sensing from one region to another). This paper gives consideration to the use of qualitative spatial representations in agent-based modelling, using a model of everyday pro-environmental behaviour in the workplace as an example.