The mechanisms of variation, selection and inheritance, on which evolution by natural selection depends, are not fixed over evolutionary time. Current evolutionary biology is increasingly focussed on understanding how the evolution of developmental organisations modifies the distribution of phenotypic variation, the evolution of ecological relationships modifies the selective environment, and the evolution of reproductive relationships modifies the heritability of the evolutionary unit. The major transitions in evolution, in particular, involve radical changes in developmental, ecological and reproductive organisations that instantiate variation, selection and inheritance at a higher level of biological organisation. However, current evolutionary theory is poorly equipped to describe how these organisations change over evolutionary time and especially how that results in adaptive complexes at successive scales of organisation (the key problem is that evolution is self-referential, i.e. the products of evolution change the parameters of the evolutionary process). Here we first reinterpret the central open questions in these domains from a perspective that emphasises the common underlying themes. We then synthesise the findings from a developing body of work that is building a new theoretical approach to these questions by converting well-understood theory and results from models of cognitive learning. Specifically, connectionist models of memory and learning demonstrate how simple incremental mechanisms, adjusting the relationships between individually-simple components, can produce organisations that exhibit complex system-level behaviours and improve the adaptive capabilities of the system. We use the term “evolutionary connectionism” to recognise that, by functionally equivalent processes, natural selection acting on the relationships within and between evolutionary entities can result in organisations that produce complex system-level behaviours in evolutionary systems and modify the adaptive capabilities of natural selection over time. We review the evidence supporting the functional equivalences between the domains of learning and of evolution, and discuss the potential for this to resolve conceptual problems in our understanding of the evolution of developmental, ecological and reproductive organisations and, in particular, the major evolutionary transitions.

Simple distributed strategies that modify the behaviour of selfish individuals in a manner that enhances cooperation or global efficiency have proved difficult to identify. We consider a network of selfish agents who each optimise their individual utilities by coordinating (or anti-coordinating) with their neighbours, to maximise the pay-offs from randomly weighted pair-wise games. In general, agents will opt for the behaviour that is the best compromise (for them) of the many conflicting constraints created by their neighbours, but the attractors of the system as a whole will not maximise total utility. We then consider agents that act as 'creatures of habit' by increasing their preference to coordinate (anti-coordinate) with whichever neighbours they are coordinated (anti-coordinated) with at the present moment. These preferences change slowly while the system is repeatedly perturbed such that it settles to many different local attractors. We find that under these conditions, with each perturbation there is a progressively higher chance of the system settling to a configuration with high total utility. Eventually, only one attractor remains, and that attractor is very likely to maximise (or almost maximise) global utility. This counterintutitve result can be understood using theory from computational neuroscience; we show that this simple form of habituation is equivalent to Hebbian learning, and the improved optimisation of global utility that is observed results from wellknown generalisation capabilities of associative memory acting at the network scale. This causes the system of selfish agents, each acting individually but habitually, to collectively identify configurations that maximise total utility. Selfish Agents and Total UtilityThis paper investigates the effect of a simple distributed strategy for increasing total utility in systems of selfishly optimising individuals. The broader topic concerns many different types of systems. For example, in technological systems, it is often convenient or necessary to devolve control to numerous autonomous components or agents that each, in a fairly simple manner, acts to optimise a global performance criterion: e.g. communications routing agents act to minimise calls dropped, or processing nodes in a grid computing system each act to maximise the number of jobs processed (1,2). However, since each component in the network acts individually, i.e., using only local information, constraints between individuals can remain unsatisfied, resulting in poorly optimised global performance. In an engineered system one could, in principle, mandate that all nodes act in accord with the globally optimal configuration of behaviours (assuming one knew what that was) -but this would defeat the scalability and robustness aims of complex adaptive systems. The question for engineered complex adaptive systems then, is the question of how to cause simple autonomous agents to act 'smarter' in a fully distributed manner such that they better satisfy constraints between agents an...