From Agent-Based Models to Network Analysis ( and Return ): The Policy-Making Perspective
1 . Introduction — Complexity and Policy
In the last two decades , complexity
economics has reached a considerable scientific cohesion and it is currently one of the most successful endeavors at the frontier of research . The boundary that needs to be crossed is now that of the policy domain .
It is beyond doubt that the ontology and epistemology of complex systems — heterogeneity , interaction , innovation , and adaptation — offers new insights both to scholars and policymakers ( Fontana , 2012 ); however in spite of a considerable number of case studies , there is no sign of an emerging unitary theory . 1 On the contrary , on the methods side considerable progress has been made .
Among the tools developed in the complexity field , agent-based modeling ( hereafter , ABM ) and network analysis ( hereafter , NA ) seem very important in sustaining the process of bringing complexity to bear on the policy world . The former allows modeling a variety of agents and mechanism of interaction in ways that are precluded from mathematical and econometric models ; the latter unveil the role in the structure of interaction to the diffusion of the effects of policy , in their efficiency and stability over time .
Moreover , they allow embedding a huge amount of data in user-friendly models — typically software — that improve the transfer of knowledge and competences from the academic world to the policy environment .
While models using these methods are currently thriving , the attempts at applying them jointly are not very frequent ( De Caux , Smith , Kniveton , Black , & Philippides , 2014 ; Edmonds & Chattoe , 2005 ; Hamill & Gilbert , 2009 ; Kirman
& Vriend , 2001 ; Weisbuch , Kirman , & Herreiner , 2000 ). In this paper , we argue that the combination of the two methods can increase enormously the potential of complexity-based policies and we propose a model that operationalizes the merger of the two from an innovative perspective . We conclude by proposing a project for a novel procedure of analysis that can deduce individual behavior from the structure of emerging network thereby diminishing the computational and informational burden that is required to devise policies in complex environments .
The rest of the paper is organized as follows : Section 1 discusses the current state of the literature on the joint use of ABMs and NA and emphasizes its potential benefits ; Section 2 introduces recipeWorld , an agent-based model that simulates the emergence of a network out of a decentralized autonomous interaction ; Section 3 illustrates a reverse engineering technique — from data to model — that we are starting to develop and its importance for policymaking ; Section 4 takes a broader perspective on an ABM / NA policy and discusses how it can overcame some limitations of the current approach . Section 5 concludes with some remarks .
2 . Agent-based Modeling and NA : the Benefits of Cross-fertilization
The very definition of a complex system
involves structure and patterns emerging from a decentralized autonomous interaction . The exploration of this micro – macro mapping is well suited to ABMs , but what if the emerging structure is a network ?
To put it differently , social , economic , and technological networks in the real world are generated through contacts made by individuals pursuing their
78