Technological Base and the Evolution of Complex Learning Networks: An Agent-Based Simulation

Adam Tatarynowicz, Department of Business Administration, University of St. Gallen

Empirical studies of cooperative agreements between firms contend that interfirm networks frequently take on characteristics consistent with the graph theoretic notion of a small world: they are characterized by uneven structures composed of several clusters that are internally dense yet have relatively few links leading from one cluster to another. While this revelation is nothing new in management science, scholars are only now beginning to realize that this complex network topology is not equally prevalent in all industrial settings. Some industries are apparently more “small-worldly” than others, and it remains largely unclear as to what particular elements of the industry produce these differences.
In this paper I focus on the technological base of an industry as one possible determinant of the observed network structures. Defining technological base as the number of different knowledge types required for production and, hence, successful learning (i.e., knowledge breadth) as well as the extent to which each knowledge type can be maximally utilized in the given sector (i.e., knowledge depth), I propose an agent-based model for the emergence of complex interfirm networks in different industrial environments which I simulate as a simple combination of these two parameters. Learning takes place when two autonomous agents (i.e., organizations or firms) form a joint alliance in order to recombine their existing informational endowments or when a single agent decides to innovate independently in the absence of appropriate partners. Three considerations drive the partner search process: complementarity of the knowledge resources held by both potential allies, the history of their previous collaborations, and the information about one another gained via common past and present contacts. As alliance formation continues and subsequent partnering decisions become increasingly embedded in the surrounding social structure, a large network of agents evolves whose properties are studied along with the properties of the learning processes occurring on that network. My results show that the topology of networks generated by these mechanisms depends heavily on the underlying technological base – with small worlds observed only within a limited region of the parameter space – and that there is a strong link between the resultant network structures and the agents’ learning performance.

Date: 08 March-00 0000