Leading large: emergent learning and adaptation in complex social networks

The mechanism by which adaptation in large organisations emerges from localised bottom-up processes remains largely unexplored. This paper describes the emergence of a learning algorithm in organisations which crosses levels of analysis. It posits that what is essentially a neural network arises naturally in organisations with individuals as nodes, interactions as edges, and influence relationships among them performing a function that is analogous to synaptic weights. This network structure enables organisations to adapt through a process of influence process structural learning that is analogous to back propagation learning in traditional neural networks. The model describes leadership within top management as expressing the organisation’s response to environmental stimuli about which top managers have little direct knowledge. Leadership acts to change influence relationships among managers by altering their relative status and reputations. The theory implies that influence relationships exhibit a power-law distribution, a potential marker of emergent collective agency.