To satisfy this need, a model system is studied for the collective solution of a sequential problem using self-organizing dynamics. The model system is a large number of non-interacting agents that solve a maze using simple rules of movement. The collective solution exhibits a rich set of properties associated with complex adaptive systems, including emergent properties, redundant subsystems, robust performance, persistent disequilibrium, information condensation, and functionality of the whole greater than in individual. The richness of the results of a model problem which is so simple suggests that much of the complex behavior and function that we associate with social systems is due to the collective interaction of simple processes, and not from the complex processing of a few processors.
Because of the nature of the problem domain, the simulations can evaluate the effect on the collective solution of modifying an individual's contribution. The simulations demonstrate that under ideal conditions large numbers of individuals can solve problems far better than an average individual. The ideal conditions are when the full diversity of the population contributes to the collective decision. This diversity includes breath and depth of the learning experience and differences in performance (yes, even poor performers are needed). The collective performance was observed to be much worse than the average individual in only a few circumstances: when the individual learning experience was based on a random-walk solution (an individual has no problem-solving capability), when the individual was indecisive (all opinions are of equal importance) and when a leader was selected randomly from the individuals in the collective. Each of these has relevance to failures in our existing problem solving approaches. The results of the simuations provide guidance on the most efficient way to capture the contributions of individuals without sacrificing the functioning of the self-organizing, collective system.
The conclusions of the study also have implications beyond the Symbiotic Intelligence Project. This study provides insight into the question: What individual capabilities are necessary (and therefore have biologically evolved) which enables social evolution to function?