Standard Diffusion in Networks: A Set of Models to Analyze IT Infrastructure Path Dependence

Research output: Book/ReportPh.D. thesis

Abstract

Standards enable effective collaboration among individuals, organizations, and systems of all sorts in our increasingly digitized society. Self-reinforcement in standard diffusion processes creates installed base advantages for adopters, credible standards, and fosters a variety of complementary products and services. Self-reinforcement, however, breeds path dependence and lock-in. Newly introduced, more efficient standards are often disadvantaged because they have smaller networks. The examples of global airline distribution and organizational IT infrastructures suggest that work focusing primarily on the network size – in the tradition of Arthur’s path dependence model – has difficulties explaining how inertia actually builds up and how important standardization patterns such as islands of shared technologies can arise. As the notion of path dependence has come to impact research and managerial thinking, I believe it is important to relax restrictive boundary conditions of its conceptual core. I contend that path dependence theory must account for a broader range of interaction patterns and growth logics. I view path dependence as a problem of standard diffusion in networks. Consistent with this view, I first suggest a model of standard diffusion in growing networks. The model reproduces Arthur’s path dependence model and a Polya Process as special cases and allows testing of the effect of different growth parameters on path building. Agent-based simulations show that network effects – formed as a function of a growing network size – and spillover effects – contingent on the degree to which an agent’s partners adopt – are usefully distinguished in growing networks as having different, non-monotonic effects on diversity. Network effects foster one standard’s dominance due to increasing network influences. Spillover effects, in contrast, limit influences from growing network sizes: segregated regimes can come to settle as new agents are less dependent on the total number of adopters. A case study of a recycling company demonstrates the model’s usefulness for understanding the evolution of organizational IT landscapes. In addition, a method is introduced to IT managers and architects that identifies critical IT systems with respect to their architectural embeddedness and links a system’s network position with continuance inertia. Based on problems of path creation in global airline distribution IT, I then suggest a second model conceptualizing the diffusion of a new standard as a contagious process that spills over from one organization to another. I operationalize codeshare linkages among airlines as a network and perform a network analysis. External shocks potentially trigger domino effects that cascade through the network. I test scenarios with respect to varying adoption thresholds that enable me to examine when and where a new standard diffuses to a nontrivial fraction of agents. I introduce a group detection algorithm – switching maximum cliques of players – to demonstrate the effectiveness of targeted compared to random network interventions. A two- step procedure for path breaking is thus suggested that identifies a set of key players and switches them collectively. Viewed together, these results demonstrate the value of a network perspective to understand better path dependencies in complex (inter-)organizational IT infrastructures.
Original languageEnglish
Place of PublicationBerlin
PublisherFreie Universität Berlin
Number of pages234
DOIs
Publication statusPublished - 2014
Externally publishedYes

Keywords

  • Organizational IT architectures
  • Path dependence
  • Standard Diffusion
  • Network analysis
  • Agent-based modeling
  • Simulation

Cite this