The Fuzzy Line Between Structure and Agency: A Fuzzy Inference Approach to Modeling Agent Response to Governing Institutions

Abstract

The representative realism of agent behaviors and emergent system phenomena in agent-based models (ABMs) plays a crucial role in computational social science and the study of social systems. As ABMs become more prevalent in mainstream use, informing the design of public policy decisions and other societal governance functions, the importance of ensuring credible simulated agent decisions becomes increasingly critical. In pursuit of this realism goal, ABM researchers have explored computational frameworks for both sides of a classical sociological theory debate concerning the primacy of structure or agency in shaping individual behavior, as well as integrative theories which challenges the rigid dichotomy between structure and agency. Integrative theories reject the notion of structure as purely deterministic and agency as completely unconstrained, emphasizing the active role of individuals in shaping and being shaped by social contexts. While sociological theory-based ABMs are richly represented in the literature, there is limited computational research focusing on operationalizing these integrative theories that bridge structure and agency. To address this gap, I propose fuzzy agent-based institutional modeling (fABIM) as an approach to emphasize social institutions and their influence on the reflexivity of agents—their adaptive capacity for considering strategies, rules, norms, or obligations in action decisions that comply with a dynamic internal mental model of social standards. My approach seeks a novel method for capturing the interdependent dynamics between institutional governance (structure) and deontic reasoning of individuals (agency). This work presents three contributions aimed at bridging the gap between theory and representative realism in computational modeling. Firstly, I build upon existing work for computationally characterizing structure within societies by incorporating the ADICO grammar of institutions by Crawford and Ostrom (1994) into ABMs. I introduce a sub-grammar in context of ABMs for each ADICO element, allowing for parameterization of agent reasoning and consideration of institutional tradeoffs. Secondly, fuzzy logic is implemented to connect individual context and interpretation of structure, to reasoning and subsequent action. This is achieved through the introduction of fuzzy logic functions that represent diverse and imprecise mental models of individuals, which capture varying degrees of support for governing institutions in different social contexts. These functions are integrated into a fuzzy inference system, serving as a heuristic construct for both framing and modeling the decision-making process of computational agents. Finally, I implemented and operationalized the first two contributions to demonstrate fABIM as an approach to apply within a computational framework, the theoretical foundations and dynamics between structure and agency. To examine and validate these advancements, a simple model of the commons governed by social institutions is used to examine each contribution. The research concludes with a discussion on the potential utility and application of the proposed fABIM concept to more representative real-world scenarios, specifically those characterized by highly divisive social contexts that are concerned about complicity of populations to imposed top-down structure (e.g., mandates). Such scenarios, for example, the inconsistent response of the US population to adopting nonpharmaceutical interventions during the COVID-19 pandemic, highlight the significant impact that voluntaristic action can have on emergent societal outcomes that depend upon complicity with social institutions. Toward this, the central thesis of this research proposes that enhancing realism in models that capture adaptive individual agency in response to institutional governance holds broad applications, particularly for evidence-based policy development and decision-support.