Evolution of self-definition

With current immune system models, the definition of "self" is usually concerned with patterns associated with normal usage. However, this has the disadvantage that the discrimination process itself may be disabled by a virus and there is no way to detect this because the algorithms controlling the pattern recognition are not included in the self-definition. To avoid an infinite regress of increasingly higher levels of reflection, we propose a model of mutual reflection based on a multi-agent network where each agent monitors and protects a subset of other agents and is itself monitored and protected by them. The whole network is then the self-definition. The paper presents a conceptual framework for the evolution of algorithms to enable agents in the network to become mutually predictive. If there is no critical dependence on a global management component, this property of symbiosis can lead to a more robust form of distributed self/non-self distinction.