Collaboration in scientific digital ecosystems: A socio-technical network analysis
This dissertation seeks to understand the formation, operation, organizational (collaboration) and the effect of scientific digital ecosystems that connect several online community networks in a single platform. The formation, mechanism and processes of online networks that influence members output is limited and contradictory. The dissertation is comprised of three papers that are guided by the following research questions: How does online community member’s productivity (or success) depend upon their ‘position’ in the digital networks? What are the network formation mechanism, structures and characteristics of an online community? How do scientific innovations traverse (diffuse) amongst users in online communities? A combination of exploratory, inductive and deductive research designs is applied sequentially but in a non-linear manner to address research question. The dissertation contributes to the literature on scientific collaboration, digital communities of creation, social network modelling and diffusion of innovation.^ The first paper applies network theory and spatial probit autocorrelative modelling technique to evaluate how member developer’s positioning in digital community correlate with his/her productivity. The second paper looks at the dynamics of developer’s participation in online developers’ network for a period spanning 7-years using exponential random graph models (ERGM). This paper applies theory of network (network science) to model network formation patterns in developer community. The third paper, like the first, applies network theory and to understand user network characteristics and communication channels which influence diffusion of scientific innovations. Bass and spatial probit autocorrelative models are applied for this analysis. ^ Data from this study was mined from developers, authors and user communities of nanoHUB.org cyberinfrastructure platform. NanoHUB.org is a science and engineering online ecosystem comprising self-organized researchers, educators, and professional communities in eight member institutions that collaborate, share resources and solve nanotechnology related problems including development and usage of tools (scientific innovation). Data from collaboration and information sharing activities was used to create the developers, authors and user networks that were used for analysis.^ Results of the first paper show that the spatial autocorrelation parameter of the spatial probit model is negative and statistically different from zero. The negative spatial spillover effect in the developer network imply that developers that are embedded in the network have a lower probability of getting more output. The structural network characteristics of eigen vector centrality had statistically significant effects on probability of being more productive. Developers who are also authors were found to be more productive than those in one network. The implications of these findings is that developers will benefit from being in multiple network spaces and by associating with more accomplished developers. The autocorrelative and interaction models also reveal various new modelling approach of accounting for network autocorrelation effects to online member.^ Results of the second paper show that developers form in a manner that follow a pure uniform random distribution. Results also show that developer’s collaborative mechanisms are characterized by low tendencies to reciprocate and form homophiles (tendency of developers to associate with similar peers) but high tendency to form clusters. The implications of network formation mechanism and processes are that developers are forming in a purely random and self-organized manner and minimum efforts should be applied in trying to organize and influence the community organization. The results also reveal that a simple link to link ERGM and stochastic dominance criteria can be combined to characterize the network formation characteristics just like the ERG(p*) model but have an advantage of overcoming degeneracy challenges associated with ERG(p*) models.^ Results of the third paper show that bass model is a good predictor for diffusion of scientific innovations (tools) in online community setting. Results also show different innovations have varying levels and rates of adoption and these were influenced by both external and internal factors. Results of the micro-based model found degrees and betweeness centrality as some of the internal variables that have positive influence on the adoption of innovation while centrality measures of power or leadership were found to have negative influence of adoption process. The relative time taken to run a simulation (measured as job usage time) was also found to be negatively influencing diffusion. The implication of the study results is that bass model is a good fit for evaluating and forecasting adoption of innovation in online communities. Moreover, network structural characteristics are responsible for adoption of innovation adoption and policy making should consider tool adoption enhancing ones. Additionally, researchers could further explore the network structural characteristics that are driving diffusion of innovation.^
Sabine Brunswicker, Purdue University.