WED 2:00 PM Discussant-type Session -  Diffusion Dynamics

Full Report:

Local Analysis of Individual-based Viral Dynamic Models with Eigenspace and Eigenvalue Elasticity Analysis
This presentation started with the definition of the system to be modeled and the most significant factors that influence the process of spread of an infectious disease. Then the Eigenvalue elasticity method was discussed and the limitations of using this approach were presented. Using this approach, it is possible to compare the importance of parameters, and to control the disease spread, to some degree, in the stage of disease outbreak. The co-effects of eigenvalues, eigenvectors and coefficients make obstacles to analyze models with a large number of state variables.
Further work is required to find a way to refine eigenspace methods to understand the impact of parameter changes on the long term behavior of the model. But, in conclusion, it is difficult to apply eigenvalue analysis to an Individual Based Viral Model.

Paper #2
The Impact of Aggregation Assumptions and Social Network Structure on Diffusion Dynamics
The presenter started by describing the model of information flow, two groups of actors, potential adopters and adopters. The basic assumptions and the dynamic mechanisms used in the model were also described. The causal loop diagrams and the stock flow diagram were described and it was explained how the scenarios were constructed. The author argues that the way information flows in a network is significantly important on diffusion processes. Usually it can be assumed the network is perfect, but it is not always true.
Agent based models were created to compare the results being obtained from the system dynamics model. There is a significant convergence behavior; as density of the network increases the diffusion should match the dynamic systems model. A convergent behavior as a function of network density was observed. As a conclusion, there is a relationship between the adoption level and network density.


Discussion ( relevant questions)
Impact of network density is relevant, but what is the impact of locality? How strong are local connections?
The author did not consider the locality, he is did not include that in the network.

Is this approach worthwhile to pursue?  The eigenvalue method is used a lot and most of the control literature is focused on analysis of the model, but in this context things are more complex. IBM is worth to pursue? This technique is worth to research and pursue or is a no go direction?
This research is focused on whether or not the method is useful for IBM, and so far it shows maybe is not.

Angelica Burbano