Directors: Jessica Flack & David Krakauer
Conflict, collective behavior, and the emergence of multi-scale, hierarchical organization in adaptive systems
An outstanding question in evolutionary theory is why life has evolved to be hierarchically organized. From cells, to organisms, to societies, evolution generates structures nested in space and time. Organisms are made of molecules within cells within organ systems within bodies. Brains are composed of neurons organized into networks, which are organized into interacting brain regions. Societies are individuals interacting in nested social networks, out of which persistent social structures emerge.
We are exploring the possibility that new levels of organization arise as asymmetries resulting from noise and competition among lower-level components become amplified and consolidated through a process of symmetry breaking. The emerging higher-level structure is essentially a coarse-grained representation – an average is a simple example – of the conflict dynamics at the component or individual level. As they interact, the components use the slowly changing coarse-grained statistics to make decisions, as these averages are more predictive of the future state of the system than the constantly in-flux events at the lower level. The in-flux events, on the other hand, allow the system to closely track environmental changes, preparing it to adapt. In this way, biological and social systems are sometimes able to balance tradeoffs between robustness--a system's ability to survive a range of threats, and evolvability--a system's ability to change in response to changing environments.
A computational approach: Mapping component behavior to aggregate properties through the construction of adaptive, causal circuits
We propose that aggregate spatial and temporal patterns, represented as coarse-grained variables, arise through an optimization process, in which components compete for locally-defined rewards, leading them to adopt strategies that are either transiently or long-term stable. As the environment changes, the optimization process generating the distribution of strategies and consequently the coarse-grained representations, can change as well, allowing for the construction of a distributed, adaptive circuit that connects component behavior (in brain--neural firing patterns, in social systems--strategic decision-making patterns) to aggregate patterns (e.g. in brain--characteristic frequencies and spatial coherence, and in social systems--distributions of fight sizes, power distributions, etc.). We are pursuing the possibility that it is through a process of collective social computation that new levels of spatial and temporal organization arise.
Endogenous coarse-graining and component cognition
If the coarse-grained variables created through a process of collective computation are to feed-down to influence component behavior, they must be perceptible by the components. This requires that they be regular and relatively slowly changing. We study the heuristics components--whether cells or individuals--use to estimate these statistics and how errors in their estimates affect the predictive utility of the coarse-grained variables as well as system robustness.
Model systems
We study these issues empirically and computationally in animal society model systems (the macaque genus) and brain (the visual system). In modeling work, we compare dynamics of social, neural, developmental, and immune systems.
Photo credits: Chimp pant-grunt: Frans de Waal, Macaque Social Group: R. Fontaine, Bee Policing: K. Foster