Complex events and feedback within and between different levels of aggregation: implications for the development of hybrid SD/ABM simulations

Bernd Wurth, Susan Howick, Niall MacKenzie

Research output: Contribution to conferenceAbstract

Abstract

Combining system dynamics (SD) and agent-based modelling (ABM) has a tremendous potential to model complex systems in various fields. Though the number such hybrid approaches is constantly increasing, some of the fundamental issues are still unanswered and compromise the practical applicability . This paper revisits the concept of hybrid simulations with a particular focus of combinations of SD and ABM. The desirability and feasibility of hybrid SD/ABM simulations are addressed, including philosophical, theoretical, and methodological considerat ions. Two key issues that separate SD/ABM from other hybrid simulation approaches are the conceptualisation of feedback across SD and ABM boundaries and inconsistencies with regard to emergent phenomena from the ABM module. A common approach for the interf ace between SD and ABM is aggregating agents' states, but this does not sufficiently address the complexity of dynamic systems with emergent properties. The implications of this issue are illustrated using the example of the collaborations between universi ties and industry. Entrepreneurial activities of universities are modelled using system dynamics and multiple universities interact with a set of agents, representing companies and start-ups in innovation ecosystems. Complex events, a collection of interre lated (simple) events, are introduced as a means to deal with emergent behaviour and feedback within and between different levels of aggregation. This has wide implications for combining SD and ABM but also for the study of complex systems in general. A model development process (MDP) for hybrid SD/ABM simulations that incorporates these ideas is provided based on the case study. The MDP is based on conceptualising feedback using causal loop diagrams before any separation between agents and aggregated variables and the operationalisation of complex events. The paper concludes with directions for future research.

Conference

ConferenceOperational Research Society Conference
CountryUnited Kingdom
CityLoughborough
Period12/09/1714/09/17

Fingerprint

Simulation
System dynamics
Agent-based modeling
Hybrid systems
Complex systems
Model development
Development process
Conceptualization
Entrepreneurial activity
Innovation
Inconsistency
Operationalization
Hybrid approach
Emergent properties
Compromise
Causal loop diagram
Ecosystem
Dynamic systems
Module
Industry

Keywords

  • system dynamics
  • agent-based modelling

Cite this

Wurth, B., Howick, S., & MacKenzie, N. (2017). Complex events and feedback within and between different levels of aggregation: implications for the development of hybrid SD/ABM simulations. Abstract from Operational Research Society Conference, Loughborough, United Kingdom.
@conference{57b6af1640d14718883f8b6102ea2712,
title = "Complex events and feedback within and between different levels of aggregation: implications for the development of hybrid SD/ABM simulations",
abstract = "Combining system dynamics (SD) and agent-based modelling (ABM) has a tremendous potential to model complex systems in various fields. Though the number such hybrid approaches is constantly increasing, some of the fundamental issues are still unanswered and compromise the practical applicability . This paper revisits the concept of hybrid simulations with a particular focus of combinations of SD and ABM. The desirability and feasibility of hybrid SD/ABM simulations are addressed, including philosophical, theoretical, and methodological considerat ions. Two key issues that separate SD/ABM from other hybrid simulation approaches are the conceptualisation of feedback across SD and ABM boundaries and inconsistencies with regard to emergent phenomena from the ABM module. A common approach for the interf ace between SD and ABM is aggregating agents' states, but this does not sufficiently address the complexity of dynamic systems with emergent properties. The implications of this issue are illustrated using the example of the collaborations between universi ties and industry. Entrepreneurial activities of universities are modelled using system dynamics and multiple universities interact with a set of agents, representing companies and start-ups in innovation ecosystems. Complex events, a collection of interre lated (simple) events, are introduced as a means to deal with emergent behaviour and feedback within and between different levels of aggregation. This has wide implications for combining SD and ABM but also for the study of complex systems in general. A model development process (MDP) for hybrid SD/ABM simulations that incorporates these ideas is provided based on the case study. The MDP is based on conceptualising feedback using causal loop diagrams before any separation between agents and aggregated variables and the operationalisation of complex events. The paper concludes with directions for future research.",
keywords = "system dynamics, agent-based modelling",
author = "Bernd Wurth and Susan Howick and Niall MacKenzie",
year = "2017",
month = "9",
day = "12",
language = "English",
note = "Operational Research Society Conference ; Conference date: 12-09-2017 Through 14-09-2017",

}

Wurth, B, Howick, S & MacKenzie, N 2017, 'Complex events and feedback within and between different levels of aggregation: implications for the development of hybrid SD/ABM simulations' Operational Research Society Conference, Loughborough, United Kingdom, 12/09/17 - 14/09/17, .

Complex events and feedback within and between different levels of aggregation : implications for the development of hybrid SD/ABM simulations. / Wurth, Bernd; Howick, Susan; MacKenzie, Niall.

2017. Abstract from Operational Research Society Conference, Loughborough, United Kingdom.

Research output: Contribution to conferenceAbstract

TY - CONF

T1 - Complex events and feedback within and between different levels of aggregation

T2 - implications for the development of hybrid SD/ABM simulations

AU - Wurth, Bernd

AU - Howick, Susan

AU - MacKenzie, Niall

PY - 2017/9/12

Y1 - 2017/9/12

N2 - Combining system dynamics (SD) and agent-based modelling (ABM) has a tremendous potential to model complex systems in various fields. Though the number such hybrid approaches is constantly increasing, some of the fundamental issues are still unanswered and compromise the practical applicability . This paper revisits the concept of hybrid simulations with a particular focus of combinations of SD and ABM. The desirability and feasibility of hybrid SD/ABM simulations are addressed, including philosophical, theoretical, and methodological considerat ions. Two key issues that separate SD/ABM from other hybrid simulation approaches are the conceptualisation of feedback across SD and ABM boundaries and inconsistencies with regard to emergent phenomena from the ABM module. A common approach for the interf ace between SD and ABM is aggregating agents' states, but this does not sufficiently address the complexity of dynamic systems with emergent properties. The implications of this issue are illustrated using the example of the collaborations between universi ties and industry. Entrepreneurial activities of universities are modelled using system dynamics and multiple universities interact with a set of agents, representing companies and start-ups in innovation ecosystems. Complex events, a collection of interre lated (simple) events, are introduced as a means to deal with emergent behaviour and feedback within and between different levels of aggregation. This has wide implications for combining SD and ABM but also for the study of complex systems in general. A model development process (MDP) for hybrid SD/ABM simulations that incorporates these ideas is provided based on the case study. The MDP is based on conceptualising feedback using causal loop diagrams before any separation between agents and aggregated variables and the operationalisation of complex events. The paper concludes with directions for future research.

AB - Combining system dynamics (SD) and agent-based modelling (ABM) has a tremendous potential to model complex systems in various fields. Though the number such hybrid approaches is constantly increasing, some of the fundamental issues are still unanswered and compromise the practical applicability . This paper revisits the concept of hybrid simulations with a particular focus of combinations of SD and ABM. The desirability and feasibility of hybrid SD/ABM simulations are addressed, including philosophical, theoretical, and methodological considerat ions. Two key issues that separate SD/ABM from other hybrid simulation approaches are the conceptualisation of feedback across SD and ABM boundaries and inconsistencies with regard to emergent phenomena from the ABM module. A common approach for the interf ace between SD and ABM is aggregating agents' states, but this does not sufficiently address the complexity of dynamic systems with emergent properties. The implications of this issue are illustrated using the example of the collaborations between universi ties and industry. Entrepreneurial activities of universities are modelled using system dynamics and multiple universities interact with a set of agents, representing companies and start-ups in innovation ecosystems. Complex events, a collection of interre lated (simple) events, are introduced as a means to deal with emergent behaviour and feedback within and between different levels of aggregation. This has wide implications for combining SD and ABM but also for the study of complex systems in general. A model development process (MDP) for hybrid SD/ABM simulations that incorporates these ideas is provided based on the case study. The MDP is based on conceptualising feedback using causal loop diagrams before any separation between agents and aggregated variables and the operationalisation of complex events. The paper concludes with directions for future research.

KW - system dynamics

KW - agent-based modelling

UR - http://www.theorsociety.com/Pages/Conferences/OR59/OR59.aspx

M3 - Abstract

ER -

Wurth B, Howick S, MacKenzie N. Complex events and feedback within and between different levels of aggregation: implications for the development of hybrid SD/ABM simulations. 2017. Abstract from Operational Research Society Conference, Loughborough, United Kingdom.