Challenge Propagation: Towards a theory of distributed intelligence and the global brain

Abstract

We sketch a foundation for a new theory of distributed intelligence, based on the process of challenge propagation, which extends the mechanism of spreading activation in neural networks to the collective intelligence emerging from a network of interacting agents. Challenge propagation is a form of self-organizing, distributed processing that allows agents to collectively tackle challenges too complex for a single agent, and that can be mathematically and computationally modelled. The basic idea is to combine the notion of “challenge”, which is defined as a phenomenon that elicits action from an agent, with the notion of “propagation”, which denotes the process by which such phenomenon is iteratively transmitted from agent to agent. A challenge is a generalization of the notions of problem, opportunity and activation. It can be characterized by valence (positive or negative), prospect, mystery and difficulty. An agent’s action on a challenge will typically “relax” the challenge, but not resolve it altogether, so that some degree of challenge remains for further agents to act upon. Propagation occurs either via a shared medium in which challenging traces are left for others (stigmergy), or via a network of agent-to-agent links learned through reinforcement of successful transmissions. Introduction