COLLECTIVE COMPUTATION ACROSS SCALES OF BIOLOGICAL ORGANISATION

Abstract

Biological systems—from cells to tissues to individuals to societies—are hierarchically organised, which suggests there are multiple natural scale at which these systems should be studied. Across disciplines (e.g. neurobiology, genetics, evolutionary biology), biologists often focus on isolating a single scale and understanding the patterns observed at this level of the hierarchy. While the task of understanding observed patterns in nature requires revealing the mechanisms that underlie them, these mechanisms generally operate at a different scale than the observed pattern itself. In many cases, the patterns must be studied as emerging from interactions among a collection of smaller units. However, in other cases, they are imposed upon the system by larger scale constraints. For example, a compass-like representation of landmarks in the fruit fly brain is only observable at a larger scale by coarse-graining the neural firing rates; gradient tracking in fish schools emerges from individuals responding to local resource levels; natural selection on phenotypic traits is constrained by the species’ evolutionary history and other environmental factors. In all examples, spatially and temporally distributed information is captured at the system level and translated to some behavioural output. In this thesis, I explore the link between the different scales of biological organisation, and the idea that hierarchical organisation allows these systems to maximise information extraction across space and time. The four chapters of my thesis draw links across these scales of biological organisation—from neural ensembles to individuals and from individuals to groups. In Chapter 1, I reveal the mechanistic and functional underpinnings of spatial decision-making across taxa and contexts. This work naturally draws a link between mechanisms at the cellular level (neural ensembles) and behavioural output (decision-making) that arises at the organismal level. Chapter 2 presents a method for automated tracking of animals’ movements from laboratory-based videos, which provides an important step in quantifying the invisible networks of social interactions—a key component for exploring the link between individuals and animal collectives. These tracking methods are then applied in Chapter 3 and Chapter 4 to expound the link between consistent individual differences in various kinematic traits, and how these traits affect group structure, leadership, movement dynamics, and foraging across social, spatial and temporal scales. In Chapter 3, I analyse trajectory data of fish schools and identify representational and kinematic properties that predict leadership events before they occur. I then further explore this line of inquiry in Chapter 4 by revealing how consistent between-individual ii differences in these kinematic properties translate to consistent differences in group functioning across ecological contexts. This work serves to highlight the important evolutionary consequences of the social environment in which individuals are embedded throughout their lives. Together, these four chapters propose general mechanisms underlying the computational capabilities of groups (neural ensembles and animal collectives) by revealing how they are able to access information at spatial and temporal scales that are unavailable to the individuals that compose them. My work supports the idea that biological systems are inherently computational; that they are composed of ensembles of interacting units capable of making predictions about their environment—thereby reducing uncertainty about the future. If this is true, then the ubiquitous complexity and multi-scale structure observed in biological systems is an inevitable consequence of evolution, allowing these systems to discretise space and time, and facilitating efficient extraction of information and energy across a range of spatial and temporal scales.