Thinking Back and Forth: Uncovering Mechanisms of Advice-Seeking Network Formation through Empirical Agent-Based Models

Advice-seeking networks are frequently sampled and studied within competitive environments, such as law firms or financial institutions. In these contexts, social exchange theories provide valuable insights into the network formation mechanisms that underlie observed structural patterns. By fitting simulated network data generated from mechanisms inspired by social exchange theories to empirical networks, researchers can typically isolate the effects of attractiveness toward specific advisors who possess relevant resources, such as skills and expertise, while also controlling for the negative status costs associated with seeking advice. However, these theories offer limited guidance in understanding the emergence of advice networks in non-competitive, supportive contexts, where different dynamics, such as reciprocation and transitive closure, become more prominent. In this study, we propose an agent-based modeling (ABM) framework to explore advice-seeking in non-competitive settings. We define various underlying mechanisms that contribute to network formation by unpacking different reciprocity-based theories. Our aim is to test whether mechanisms that assume the existence of a reciprocity norm are more effective in explaining emergent advice network patterns than those that view agents as more calculative. These calculative agents decide to seek advice either from a given advisor, expecting future benefits from the same individual (direct reciprocity), or from another advisor of their current advisor (generalized reciprocity). To quantify the effects of these different causal mechanisms, we employ Approximate Bayesian Computation (ABC). We then perform Bayesian model selection on an empirically observed advice-seeking network within a coworking space to identify the most plausible drivers of network formation.