Skip to main content

2014 | Buch

Empirical Agent-Based Modelling - Challenges and Solutions

Volume 1, The Characterisation and Parameterisation of Empirical Agent-Based Models

herausgegeben von: Alexander Smajgl, Olivier Barreteau

Verlag: Springer New York

insite
SUCHEN

Über dieses Buch

This instructional book showcases techniques to parameterise human agents in empirical agent-based models (ABM). In doing so, it provides a timely overview of key ABM methodologies and the most innovative approaches through a variety of empirical applications. It features cutting-edge research from leading academics and practitioners, and will provide a guide for characterising and parameterising human agents in empirical ABM. In order to facilitate learning, this text shares the valuable experiences of other modellers in particular modelling situations. Very little has been published in the area of empirical ABM, and this contributed volume will appeal to graduate-level students and researchers studying simulation modeling in economics, sociology, ecology, and trans-disciplinary studies, such as topics related to sustainability. In a similar vein to the instruction found in a cookbook, this text provides the empirical modeller with a set of 'recipes' ready to be implemented.

Agent-based modeling (ABM) is a powerful, simulation-modeling technique that has seen a dramatic increase in real-world applications in recent years. In ABM, a system is modeled as a collection of autonomous decision-making entities called “agents.” Each agent individually assesses its situation and makes decisions on the basis of a set of rules. Agents may execute various behaviors appropriate for the system they represent—for example, producing, consuming, or selling. ABM is increasingly used for simulating real-world systems, such as natural resource use, transportation, public health, and conflict. Decision makers increasingly demand support that covers a multitude of indicators that can be effectively addressed using ABM. This is especially the case in situations where human behavior is identified as a critical element. As a result, ABM will only continue its rapid growth.

This is the first volume in a series of books that aims to contribute to a cultural change in the community of empirical agent-based modelling. This series will bring together representational experiences and solutions in empirical agent-based modelling. Creating a platform to exchange such experiences allows comparison of solutions and facilitates learning in the empirical agent-based modelling community. Ultimately, the community requires such exchange and learning to test approaches and, thereby, to develop a robust set of techniques within the domain of empirical agent-based modelling. Based on robust and defendable methods, agent-based modelling will become a critical tool for research agencies, decision making and decision supporting agencies, and funding agencies. This series will contribute to more robust and defendable empirical agent-based modelling.

Inhaltsverzeichnis

Frontmatter
1. Empiricism and Agent-Based Modelling
Abstract
Agent-based modelling is losing its niche character and gaining wider recognition as a valuable methodology in empirical policy related situations. This growing recognition roots in the increasing demand for methods that allow integrating indicators from various disciplines across a broader systems perspective. Agent-based modelling gains its integrative strength from a combination of factors, in particular its ability
Alex Smajgl, Olivier Barreteau
2. A Case Study on Characterising and Parameterising an Agent-Based Integrated Model of Recreational Fishing and Coral Reef Ecosystem Dynamics
Abstract
This chapter reports a case study on the characterisation and parameterization of an agent-based model (ABM) of recreational fishing behaviour. Data collected through distributed surveys are used to produce the structure for agent behaviour. Specifically, agent behaviour is described using econometrically estimated site choice models, which allow for choice behaviour to be driven by angler characteristics, site attributes, the time of year, the direction of the agent’s trip, and so on. These choice models are statistical models based on accepted theories of choice behaviour. They underpin the simulation of the location, frequency, duration, and seasonality of an angler’s fishing trip as well as the angler’s expected fish catch. This case represents a relatively new approach to characterising agent-based models: using econometrically estimated models to drive agent behaviour.
Lei Gao, Atakelty Hailu
3. An Agent-Based Model of Tourist Movements in New Zealand
Abstract
This chapter outlines the development of an agent-based model to simulate international tourist movements and their “spatial yield”. A combination of interviews with tourists and front-line tourism workers was used to inform the development of tourists' decision-making processes. Data collected by departing tourists on destinations, length of stay, nationality and several other parameters were used to characterise the so later tourist groups. Both data sets allowed for the development of the conceptual model for tourist behaviours and for the parameterisation of the simulation model.
C. Doscher, K. Moore, C. Smallman, J. Wilson, D. Simmons
4. Human-Ecosystem Interaction in Large Ensemble-Models
Abstract
Managing the effects of human activities on environmental assets must surely be one of the most vexing jobs imaginable. The nature of human interactions changes with social, economic, political and technological changes and management strategies are implicitly difficult to assess: as soon as one is in place, the baseline for comparison changes. Unfortunately with growing population and associated demands it is an issue that cannot be avoided.
Randall Gray, Elizabeth A. Fulton, Richard Little
5. Using Spatially Explicit Marketing Data to Build Social Simulations
Abstract
To construct a population of artificial agents, modellers either can use available large-scale e.g. demographic orland-use data of built-up areas. Or they rely on detailed data on cognitive and behavioural variables e.g. gathered through a domain-specific survey to craftspecific behavioural agent rules.However, both scales cannot easilybe connected. This chapter describes a method of using data stemming from geo-marketing research to support this scaling-up process with lifestyles and their localisation that are used as an empirical bridge between the micro and the macro levels.
Andreas Ernst
6. Parameterisation of AgriPoliS: A Model of Agricultural Structural Change
Abstract
This model description follows the ODD protocol (Overview, Design concepts and Details) as proposed and described in Grimm et al. (Ecol Model 198:115–126, 2006) and Grimm et al. (Ecol Model 221:2760–2768, 2010). The ODD protocol allows a standardized description of agent-based models what improves the clarity and the comparability of models. The following ODD protocol has already been published by Sahrbacher et al. (ODD-protocol of AgriPoliS. Technical report. IAMO. Halle, 2012).
Christoph Sahrbacher, Amanda Sahrbacher, Alfons Balmann
7. The Parameterisation of Households in the SimPaSI Model for East Kalimantan, Indonesia
Abstract
This Chapter is based on participatory research we conducted in collaboration with the Government of Indonesia. The participatory modelling process aimed for facilitating a learning experience that involved three tiers of governance. The participatory process included the co-design of the research proposal by stakeholders who were actively involved in its implementation process by carrying out many research tasks. This was enabled by substantial capacity building activities, for instance in agent-based modelling. Following the categories of participation suggested by Barreteau et al. (2010) our process falls into the sixth category of participatory research processes, co-building and control over model use.
A. Smajgl, E. Bohensky
8. Parameterisation of Individual Working Dynamics
Abstract
How do European rural areas evolve? While for decades the countryside in many regions of Europe was synonymous with inevitable decline, nowadays, some areas experience a rebirth, even in areas where until recently development was not considered possible. Our modelling effort aims at better understanding these heterogeneities. To deal with this objective, the modelling and the parameterisation should be strongly constraint by available data. This chapter focusses on the modelling of the individual working dynamics describing how we can design the entering on the labour marking, the job search decision and process and every other process related to work from available data. We argue about the utility of large existing databases to design complex integrated individual dynamics.
S. Huet, M. Lenormand, G. Deffuant, F. Gargiulo
9. How to Characterise and Parameterise Agents in Electricity Market Simulation Models: The Case of Genersys
Abstract
This chapter describes the process of characterisation and parameterisation of computer agents in the case of decision making of profit-driven companies in a competitive electricity market. It focuses on adaptive behaviour of generation and investment companies in Australia’s National Electricity Market (NEM) as modelled by Genersys. Through initiatives such as formal focus group meetings, gathering observations of industry experts, analysing market data and a selective approach in representing real systems, modellers can improve the design and potential future use of their simulation systems.
George Grozev, Melissa James, David Batten, John Page
10. An Agent-Based Model Based on Field Experiments
Abstract
This chapter described the empirical calibration of a theoretical model based on data from field experiments. Field experiments on irrigation dilemmas were performed to understand how resource users overcome asymmetric collective action problems. The fundamental problem facing irrigation systems is how to solve two related collective action problems: (1) the provision of the physical and ecological infrastructure necessary to utilize the resource (water), and (2) the irrigation dilemma where the relative positions of “head-enders” and “tail-enders” generate a sequential access to the resource itself (water). If actors act as rational, self-interested, agents, it is difficult to understand how irrigation infrastructure would ever be constructed and maintained by the farmers obtaining water from a system as contrasted to a government irrigation bureaucracy. Wittfogel (1957) argued that a central control was indispensable for the functioning of larger irrigation systems and hypothesized that some state-level societies have emerged as a necessary side-effect of solving problems associated with the use of large-scale irrigation.
Marco A. Janssen
11. Companion Modelling with Rice Farmers to Characterise and Parameterise an Agent-Based Model on the Land/Water Use and Labour Migration in Northeast Thailand
Abstract
An agent-based model (ABM) was co-designed with a group of rainfedlowland rice farmers from a village in Northeast Thailand to investigate the interactions between wateravailability and labour migration in rice production. A preliminary agrarian system analysis of the study site (LamDome Yai watershed in Southern Ubon Ratchathani Province) suggested to define three main types of farminghouseholds according to their agro-ecological constraints and opportunities, farming practices, and socioeconomicstrategies. To reflect this typology, a group of 11 farming households were selected. The wives andhusbands of each household were involved in the co-design of the conceptual ABM through a series of fieldworkshops based on role-playing games (15-18 participants). To specify the rule-based algorithms related to ricecropping activities, a smaller group (6-10 participants) engaged in the design of UML activity diagrams.Implemented with the Cormas platform, the corresponding computer simulation model was introduced to theparticipating farmers. Showcasing the “business-as-usual” scenario enabled to fine tune the calibration and themeans of observation. The participants were then requested to suggest scenarios of interest to them to beexplored with the model. In this 4-year long process, the successive use of role-playing games, UML activitydiagrams and computer simulations with the participating farmers accounted for about two-thirds of the varioussources of information (as a complement to the one-third represented by farm surveys and secondary data) usedto specify the model. It resulted in increasing the sense of ownership of the model by the farmers. At completionof the process, a special seminar was organized at the regional Faculty of Agriculture of Ubon RatchathaniUniversity during which four collaborative farmers presented and discussed at length “their” model in front ofseventy graduate students and faculty staff.
C. Le Page, W. Naivinit, G. Trébuil, N. Gajaseni
12. Building Empirical Multiagent Models from First Principles When Fieldwork Is Difficult or Impossible
Abstract
This Chapter informs the reader about how to create and parameterize empirical multiagent models from first principles when fieldwork is difficult or impossible to conduct and data is primarily of qualitative nature. Empirical multiagent models have become ever more popular over the last decade. While informing models using statistical and geospatial data can orient itself on more established techniques and standards, methodological challenges persist in regards to using qualitative data for informing and parameterizing models. Protocols such as ODD are welcome and helpful devices—and hence used in this Chapter—but qualitative data comes with its own peculiarities. The most important of which is, for modeling purposes, that qualitative data tends to inform the logic of agent behavior. The emphasis I thus put on qualitative data to make model design decisions based on evidence and first principles will be reflected by soft adaptations of the ODD protocol. Arguably this may amount to a deeper insight the Chapter is providing: Whereas the usage of such frameworks as ODD increases model reliability, validity is built using qualitative empirical data for informing and parameterizing the agent and model behavior.
Armando Geller
13. Designing Empirical Agent-Based Models: An Issue of Matching Data, Technical Requirements and Stakeholders Expectations
Abstract
Despite their diversity, the 11 examples of empirical agent-based model design described in this Volume enable not only a consolidation of the CAP framework described in Chap. 1, but also an exchange of experiences in designing empirical agent-based models. The detailed descriptions of the example modelling processes showcase the methodological diversity and the state of art practiced within the emerging community of empirical agent-based modelling. All these examples have their own limitations as a matter of empiricism that the framework aims to structure. In this final Chapter we discuss effectiveness and robustness of the Characterisation and Parameterisation (CAP) framework, which we revised during the process of editing this Volume. Then, we discuss how the distinction of particular cases performed, which is followed by a discussion on the diversity of methods. Finally, we use the cases presented here (admittedly small in number) to provide some initial insights for the selection of suitable methods.
Olivier Barreteau, Alex Smajgl
Metadaten
Titel
Empirical Agent-Based Modelling - Challenges and Solutions
herausgegeben von
Alexander Smajgl
Olivier Barreteau
Copyright-Jahr
2014
Verlag
Springer New York
Electronic ISBN
978-1-4614-6134-0
Print ISBN
978-1-4614-6133-3
DOI
https://doi.org/10.1007/978-1-4614-6134-0