Mutual learning in networks: Building theory by piecing together puzzling facts

Abstract

Organizational learning research is based on the idea that individuals can learn more together than alone. Network connections between members of an organization allow them to engage in a mutual learning process whereby they share what they discover and potentially learn and improve their performance at a faster rate. Existing research highlights the importance of network structure in facilitating mutual learning, with a centralized network expected to improve learning and performance when the unfamiliar assignment is complex. An example of an unfamiliar complex assignment is new product development. The features of a new product and how those features should be combined are unknown. In a centralized network, members are connected to a focal individual but disconnected from each other. The disconnects facilitate individual search and experimentation, increasing the odds that a superior solution will be discovered. To understand how network centralization affects mutual learning and performance when an unfamiliar assignment is complex, this chapter offers a theory-building exercise. The exercise is motivated by a conflicting empirical result. Recent research indicates that for a complex task, individuals learning in a decentralized network, a network that contains a relatively large number of direct and indirect relationships, can outperform individuals learning in a centralized network. The exercise amounts to puzzling through the empirical results. Putting the pieces together suggests that the requirements for mutual learning can be met in either a centralized or decentralized network. Moreover, communication timing appears to be critical. When communication across relationships is delayed, individuals working in a decentralized network can explore a diverse set of ideas while maintaining the ability to exploit a proven idea. When communication across connections is not delayed, a centralized network produces better outcomes. The results of the theory-building exercise suggest a contingency: the ideal network for a complex assignment could depend on the rate of communication across network connections.