Information, Interactions and Behavior

Banner Image: A group of flamingos can be seen here. Research shows that they are very social animals and are also known to communicate with each other using non-verbal cues (in addition to verbal ones).All images on this website are by Anshuman Swain, unless stated otherwise. Please seek permission before use.

1. Group size and decision making in guppies

Project members: Anshuman Swain and William F Fagan

Status: Published; Swain, A. and Fagan, W.F., 2019. Group size and decision making: experimental evidence for minority games in fish behaviour. Animal Behaviour, 155, pp.9-19. https://doi.org/10.1016/j.anbehav.2019.05.017

Animals tend to learn and make decisions inductively, and simple, individual-level behavioral decisions can scale up to yield interesting emergent properties at the population level. The minority game is a theoretical formulation based on the principle of inductive learning, wherein a group of individuals, each facing two equivalent choices, self-organize to achieve maximum coordination. Coordination increases with memory length up to a certain threshold, and thereafter, at very high memory lengths, decisions resemble a random choice game. We invoked and observed minority games in guppies, Poecilia reticulata, by forcing them to choose between two symmetric chambers in a series of repeated trials. At an intermediate timescale, guppies self-organized into a globally efficient state featuring maximum coordination. After a large number of trials, the guppies approached a steady state in which they behaved as if they were randomly choosing between the two chambers. Intriguingly, both the time taken to reach this globally efficient state (i.e. learning time) and the grouping behavior for decision making depended on the sexual composition of the fish populations. We identified a positive correlation between group size and learning time in the experiments, which we further explored using simulations to uncover the form of decision-making framework at play, testing between a leader-based framework and a consensus-based one. Our simulations supported the presence of a consensus-based decision-making process in the system. This work also provides an unexplored general framework to further investigate and understand simple decision-making dynamics and structure in animal groups.

Overview of the experimental trial. Fish were kept in an aquarium (a), where they were given a mechanical disturbance in the central part of the aquarium (b), which made the fish choose a side, after which a mechanical bifurcation was added (c) so as to make choices irreversible. The water level on the majority side was lowered (d) to give a negative feedback to the majority. The number of fish on each side was noted and the whole cycle was repeated.

Variance in attendance versus information using the classic minority game simulations. This figure serves as a reference for understanding the parameters that we are seeking in the experimental data: (1) equilibrium value of variance, which occurs at high values of levels of information; the average group size is one-fourth of this equilibrium variance (2) learning time, the information level at which the variance reaches its minimum; (3) magnitude of that minimum variance; and (4) reduction in the standard error of the variance as the amount of information increases after the minimum is reached.We see a similar behavior in the guppy experiment shown on the side.

2. Emergent behavior in ant interactions

Project members: Anshuman Swain*, Sara Williams*, Louisa Jane Di Felice*, and Elizabeth A. Hobson

*contributed equally

Status: Published in Animal Behaviour; formed as a part of a working group at Complex Networks Winter Workshop (CNWW) 2019, Quebec City, Canada

Ant colonies display high levels of organizational complexity requiring no centralized control. Local interactions among ants, and between ants and their environment, are crucial in allowing for species with limited cognitive capacities to generate complex behavior at the colony level. However, little is known about the specific role played by local interactions and communication between ants, as collecting extended data-sets on ant behavior is an intensive and inexact process. Over the last years, the development of tracking systems has opened up the possibility of obtaining rich data-sets of ant-to-ant interactions without the need of manually observing and tracking ant behavior. In this work, we are analyzing the data-set collected by Mersch et al. 2013, providing spatial and interaction data of six Camponotus fellah colonies over 41 days, with a total of 985 ants. Using methods of network analysis, we measure emergent properties of the ant interaction network and analyze the role played by interactions in task allocation and switching behavior.

Depiction of proportion of ants from a single colony, having different roles (N: Nurse, C:Cleaner, F: Forager) and relative switching behavior. Behavior was constant more or less over a ten day time period. Each ant was tagged and individually surveyed. Missing ants refer to the ants that died or lost their tags.

3. Exploring microbial community dynamics as a species continuum

Project members: Anshuman Swain*, Levi Fussell* and William F Fagan

*contributed equally

Status: Published in PNAS

Understanding the origin and maintenance of microbial diversity is a major challenge. How does a community that contains a mixture of toxin-producing, toxin-sensitive, and toxin-resistant species remain stable? Previous theoretical studies have uncovered salient processes that stabilize diversity in such communities, but have only done so for communities involving few discrete species. Here we explore a microbial community model where species are defined on a continuous trait dimension, which theoretically allows for communities with an unlimited number of species. The continuous spectrum generalizes intra-species relationships of toxin (or antibiotic) production, sensitivity, and resistance that arise randomly in populations through mutation as well as interspecies relationships arising due to genetic diversity. Indirect (tripartite) interactions shape species coexistence as toxin-resistant (antibiotic-degrading) species modify the toxin/antibiotic interactions between two other species. Each model species has three attribute properties (species it can kill, species it is resistant to, and species by which it can be killed), all of which are defined in terms of “relational bands” on the trait axis. Species are drawn at random from the trait axis to populate a habitat (e.g., agar plate) where they can interact with other species located within a specific “interaction radius.” Adjustable model parameters are the width of the relational bands, the size of the interaction radii, growth rate, and mutation rate.

The interplay between the size of the relational bands for killing and inhibitory interactions leads to spontaneous chaotic and cyclic dynamics. Moreover, the size of the killing relational band dictates the chaotic dynamics, independent of the killing interaction radius. Extreme mutation rates direct the community into chaotic or stable states, but intermediate mutation rates lead to a dynamic stability where species’ population sizes are constant but interactions occur at the spatial boundaries, independent of inhibition. Intermediate-sized relational bands also engender dynamic stability, but with cyclic species dominance. Further, having both inhibition and killing capabilities in the model, coupled with mutation, consistently causes strong cyclic patterns in species dominance, cycles that are highly nonlinear with varying periodicity, amplitudes, and plateaus. Collectively, these results demonstrate the rich spatiotemporal dynamics that are possible when large numbers of microbial species with restricted but heterogeneous rules for aggressive, inhibitory, and resistant interactions live in a common spatial arena. These findings of rich dynamics may also be relevant for coral communities and other spatially structured systems featuring diverse types of interspecific interactions.

Example videos of the population dynamics that takes place in such a system. Colors denote species.

Page updated

Google Sites

Report abuse