Keywords

socio-economic data, empirical agent-based models, regime shift, behavioral change

Start Date

15-9-2020 2:20 PM

End Date

15-9-2020 2:40 PM

Abstract

Agent-based models (ABMs) are developed for different purposes. While having a stylized ABM could be useful as a learning tool or a theory-development exercise, addressing a policy-relevant question usually calls for empirical models. Empirical ABMs proliferate, fueled by the emergence of new data collection methods, especially on the social dynamics side. In addition to traditional aggregated data, such as census, land use and market trends, researchers increasingly rely on micro-level data revealing preferences, perceptions, mental maps defining the architecture of modelled decisions and even structure of social interactions. This talks draws on the experience of using multiple data sources, both on the dynamics of macro-patterns as well as on the assumptions about micro-level behavior and interactions. I will discuss the value of different behavioral and economic data for ABMs of coupled social-environmental systems. In addition to the methodological approaches to ABM validation, I would like to discuss two challenges. Firstly, the ability of data on past behavior to represent what we anticipate to happen in the future, especially in policy implications where behavioral changes are at the core. While some changes may by nature be marginal, others can take a transformational form, leading to regime shifts. Secondly, ABM community widely shares the need for clear theoretical and empirical microfoundations of rules guiding agents’ actions and interactions. Yet, the implications of operationalizing in the computer code social concepts, which may be vague at times (think of perceptions or values), are rarely assessed. Parametrization and validation in this case may face a challenge, which needs to be openly discussed.

Stream and Session

false

Share

COinS
 
Sep 15th, 2:20 PM Sep 15th, 2:40 PM

A quest for (behavioral) data in agent-based models for environmental policy

Agent-based models (ABMs) are developed for different purposes. While having a stylized ABM could be useful as a learning tool or a theory-development exercise, addressing a policy-relevant question usually calls for empirical models. Empirical ABMs proliferate, fueled by the emergence of new data collection methods, especially on the social dynamics side. In addition to traditional aggregated data, such as census, land use and market trends, researchers increasingly rely on micro-level data revealing preferences, perceptions, mental maps defining the architecture of modelled decisions and even structure of social interactions. This talks draws on the experience of using multiple data sources, both on the dynamics of macro-patterns as well as on the assumptions about micro-level behavior and interactions. I will discuss the value of different behavioral and economic data for ABMs of coupled social-environmental systems. In addition to the methodological approaches to ABM validation, I would like to discuss two challenges. Firstly, the ability of data on past behavior to represent what we anticipate to happen in the future, especially in policy implications where behavioral changes are at the core. While some changes may by nature be marginal, others can take a transformational form, leading to regime shifts. Secondly, ABM community widely shares the need for clear theoretical and empirical microfoundations of rules guiding agents’ actions and interactions. Yet, the implications of operationalizing in the computer code social concepts, which may be vague at times (think of perceptions or values), are rarely assessed. Parametrization and validation in this case may face a challenge, which needs to be openly discussed.