Dynamics on and of Networks
Jean-Louis Giavitto, IRCAM, France
Ensemble music is the result of a choreography of events and expectations in time. And the capacity for real-time synchronization and coordination is a common ability among trained musicians performing together a music score. But is a computer able to understand the dynamic of playing together?
Parts of the answer are provided by Antescofo, a system that attempts to enable computer-human musical interactions in the context of mixed music, i.e. when humans and computers are performing together. Antescofo provides an abstract programmer's model for an artificial musician in an ensemble with musical real-time constraints. Antescofo ability to interact musically relies on a dedicated heterogenous model of time, that encompasses event-driven and time-driven specifications, absolute and relative time, and subjective and social time. This model of time makes possible the common understanding of the coordination required to take part in musical interactions, a first step in the development of a musical companionship and a practical experience in the possible sharing of time between man and computers.
James Marshall, Sheffield, UK.
‘Optimal’ decisions on graphs:
Optimality theory for decision-making exists at two levels, optimal decisions by individuals, and optimal decisions by groups. Although these levels can be closely related, they are not often studied in conjunction. In this talk I will explain how, on complete graphs, optimal individual decisions can be optimally combined by quorum detection. I will then consider what may happen when we move to incomplete graphs, and how naive-Bayes agents can synchronously integrate social information very simply; in doing so, however, non-optimal information cascades occur within the group. The problem is resolved by moving to the more realistic case of asynchronous decisions, in which social information must be appropriately integrated with private evidence accumulation.
Soil webs, sea mats and scaling the turtle: Making sense of the multitude of feedbaks in flow networks
It is easy to get overwhelmed by the complexity of the dynamic systems we are surrounded by and are part of. So many interactions, generating many more feedback loops, making it impossible to predict what will happen next.
Theoretical ecology has a long tradition of studying interaction networks. It has identified simple mathematical relations between their structure and stability. But when real data are used to parameterise the interaction strengths in classic theoretical models, these relations no longer hold. For example, the complexity of a network (defined as the product of system size and its connectance, the proportion of possible interactions that are realised) is traditionally linked with instability, but observations show that it has no relation with stability at all. The theoretical relation with instability is an artefact of the assumption of random interaction strengths. To explain the stability of naturally organised systems, we need to quantify the interactions, go beyond the pairs, and quantify longer feedback loops.
With examples from soil food webs, above-and-belowground Antarctic ecosystems and bryozoan competition networks on the sea floor, I hope to show that with this quantitative loop analysis we can take a step towards a general stability theory grounded in observation. I will address questions like:
How can we compare systems encompassing many different time scales?
How far do we need to go beyond pairwise interactions?
Why do natural systems become more complex over time, or How can we marry stability with development?