Moving Together: Understanding Collective Ant Behavior through an Agent-Based Model of Pheromone Dynamics

Abstract

Collective behaviors, wherein individuals interacting according to simple rules produce com- plex emergent behaviors with no central control, has captivated scholars for decades. The study of these behaviors has the potential to both clarify our understanding of fundamental questions across multiple disciplines, as well as provide insights for applications in human technologies and organization. Ants, as one of the most prolific social insects, have been a subject of intense study for the last half century due to the large number of collective behaviors they engage in. In this thesis, we examine two surprising behavioral phenomena observed in videos of the clonal raider ant under laboratory conditions: Ostwald-like clus- tering and temperature-induced de-aggregation. Given past research, we hypothesized that these behaviors could perhaps be explained as results of pheromone dynamics that ants are known to engage in as part of well-understood adaptive functions, such as nest construction and foraging. In order to address the mechanistic questions of interest, we developed a novel agent- based model of pheromone-ant dynamics, to understand how much of the observed behaviors could actually be explained by simple pheromone dynamics. Many scholars have constructed mathematical models as part of their research. However, most of these have involved a high level of abstraction to replicate collective behaviors without focusing on the mechanisms involved. Our model therefore went beyond the traditional level of complexity seen in most of the literature in the discipline. Through Monte Carlo simulations, we carried out quantitative parameter analyses and qualitative examinations of the model. The results showed that the model was able to fully replicate one of the two behaviors of interest (Ostwald-like clustering), while partially replicating the other (temperature-induced de-aggregation). Path-following, a phenomenon seen in most ant species, was absent in model simulations, partially explaining the model’s shortcomings. This represents the most promising approach to improve the model in future research. The results of the study thus showcased the potential of fine- grained agent-based models coupled with computer simulations to generate insights about mechanistic questions in biology and other disciplines.