DATE: Wednesday 10th October 2012

TITLE: Acting on their own behalf: Norm-generating and norm-following mechanisms in simulated agents

SPEAKER: 
Matthew Egbert email  Informatics and biology departments at 
the University of Sussex web 

ABSTRACT: 
One of the fundamental aspects that distinguishes acts from mere events is that actions 
are subject to a normative (good/bad) dimension that is absent from other types of interaction: 
natural agents behave according to intrinsic norms that determine their adaptive or maladaptive 
nature.

In this talk I will discuss our recent paper in which we present a minimal model of a metabolism 
that is coupled to a gradient climbing chemotaxis mechanism. We use this model to show how the 
processes that determine the viability limits of an organism (the conditions in which it will live 
rather than die) can also influence behaviour and how, therefore, agents can act "on their own 
behalf", i.e., to satisfy their own intrinsic needs.

Dynamical analysis of our minimal model reveals an emergent viable region and a precarious region 
where the system tends to die if environmental conditions remain the same. We introduce the concept
of normative field as the change of environmental conditions required to bring the system back to 
its viable region. Norm-following, or normative action, is defined as the course of behaviour whose
effect is positively correlated with the normative field.  We thereby make progress on two problems
of contemporary modelling approaches to viability and normative behaviour: 1) how to model the 
topology of the viability space beyond the pre-definition of normatively-rigid boundaries, thereby 
allowing the possibility of reversible failure; and 2) how to relate, in models of natural agency, 
both the processes that establish norms and those that result in norm-following behaviour.

The work presented is an extension of a paper that was selected as one of the best submissions to 
the European Conference on Artificial Life in Paris 2012, and should be of interest to a wide 
audience, from philosophers of biology to protocell researchers.

BRIEF-BIO:  Matthew Egbert is a Research and Tutorial Fellow with the Evolutionary and Adaptive 
Systems Group at the Centre for Computational Neuroscience and Robotics, an interdisciplinary lab 
that is associated with both the informatics and biology departments at the University of Sussex. 
In this role, he teaches graduate and undergraduate level courses on artificial life and adaptive 
behaviour while carrying out research into the relationship between metabolism, adaptive behaviour 
and evolution.  This research typically involves the development and analysis of minimalistic, 
dynamical computational models of biological phenomena.