Agent-Based Institutional Modelling: Novel Techniques for Deriving Structure from Behaviour

Abstract

Institutions, as manifestations of social behaviour, are crucial elements of any society's structure and are widely studied in the area of New Institutional Economics as a determinant of a society's development and prosperity. Although taking promising steps by considering a wider array of behavioural influence mechanisms, to date analytical approaches primarily examine institutions from the perspective of stable equilibria, which are assumed as indicative for the existence of institutions. In this work we shift the focus from equilibria to an interactionist perspective, and apply Agent-Based Modelling (ABM) for the purpose of institutional analysis.

ABM offers the ability to represent institutions in great detail by using modelling metaphors that are in close alignment with real observable social structures and interaction patterns, as opposed to concentrating on rational strategy choices of structurally uniform individuals. With ABM we can leverage a sociological perspective and "grow" institutional structures from the bottom up, instead of just interpreting those from a bird's eye perspective.

This work not only applies ABM to institutional analysis, but also develops it further as a tool for the representation and systematic analysis of institutions. Essential contributions pointing in this direction include a refined generalised institution representation (nADICO) that captures the complexity of institutions. We further propose a continuous norm concept (Dynamic Deontics) that models the formation and change of normative understanding based on environmental stimuli. A final extension is the analysis of normative understanding on different sociological levels using Interval Type-2 Fuzzy Sets. These contributions enrich the computational social scientists' toolbox with modelling and analysis mechanisms that are a) sociologically grounded, b) facilitate a fine-grained explicit representation of institutions, and c) permit the multi-level analysis of social systems.

To showcase ABM's applicability for institutional analysis, in the first part of this thesis we review a seminal case from the area of comparative economics, Greif's Maghribi Traders Coalition, a medieval trader collective that operated based on informal means without relying on contractual enforcement. Based on the detailed analysis of recent literature, we offer a refined interpretation of the coalition. We further introduce the Genoese society that Greif contrasted to the Maghribis, since it primarily relied on formal contracts to govern cooperation.

With both scenarios as a basis, we review selected abstractions Greif puts forth for his original equilibrium-based model. Those include the assumed inherent secrecy among Genoese, in contrast to the openly sharing Maghribis. We further relax the assumption of a closed trader coalition for the Maghribi side. A third aspect involves the neglected significantly contrasting role conceptualisations in both historical societies. To deepen our understanding of those aspects, we develop dedicated agent-based models to explore alternative explanations.

In the second part of the thesis, we move towards the greater goal of growing institutional structures from a behavioural, sociologically motivated, perspective. We introduce the aforementioned generalisable representations for institutions and continuous conception of normative understanding. We then apply those to explore the third aspect of the historical scenario and analyse how far the different societal role stratification could have influenced normative understanding. Using fuzzy concepts, we conclude with a multi-level analysis of normative alignment for both trader societies.

This work thus both applies Agent-Based Modelling, but also enhances it for the purpose of Institutional Modelling and Analysis.