I am going to present a paper about the concept of modeling in a PhD colloquium and award at European Meetings on Cybernetics and Systems Research EMCSR 2014, 22-25 April 2014 in Vienna, which is a biennial from 1972 . This program that I have been selected to present in is a competition among only 6 selected PhD students. I am looking forward to it and I hope it will be a place for deep discussions, unlike majority of conferences that you go and see several meetings in parallel and no major questions in any of the talks.
The topic that I am writing about, quickly becomes abstract, since its goal is to analyze the foundations of current concepts of modeling in scientific disciplines.
I argue that the majority of existing modeling methods including numerical and analytical are under an umbrella of the same paradigm, which can be called rational modeling.
In this paper, I am trying to attack the fundamental underlying assumptions of rational modeling regarding their capacities in dealing with complex systems.
Further, based on some ideas in category theory, we propose a conceptual inversion ind the paradigm of modeling, which proposes a shift from explicit representation of objects of inquiry (set theoretic approach to modeling) to implicit representation of objects (category theoretic approach).
Further, we think this paradigm shift! is only possible along with an inversion in the concept of observation. By the change in the concept of observation, I mean what is happening now a days, is that data is no longer experimentally designed and collected, but data in a form of continuous online streams is being emitted from everywhere as kind of by-products of our daily digital activities. Therefore, in a way data is no longer just the support of our empirical studies, but rather it is everywhere, overtime in different modes. It is a new infrastructure itself.
Therefore, based on this conceptual shift that we propose plus the availability of large data sets (I intentionally avoid the term Big Data), we need new computational modeling toolsets. Among them, probabilistic network models such as Markov Networks and data driven modeling techniques such as Self Organizing Maps have the highest level of alignments with this conceptual shift.
You can find an extended abstract of this paper here.