Abstract of Dissertation Work

Emerging cyber-infrastructure tools are enabling scientists to transparently co-develop, share, and communicate in real-time diverse forms of knowledge artifacts. In my dissertation work, these collaborative environments are modeled as complex adaptive systems using collective action theory as a basis. Communication preferences of scientists are posited as an important factor affecting innovation capacity and resilience of social and knowledge network structures. Using agent-based modeling, I developed a complex adaptive social communication network model. By examining the Open Biomedical Ontologies (OBO) Foundry data and drawing conclusions from observing the Open Source Software communities, I presented a conceptually grounded model mimicking the dynamics in what is called Global Participatory Science (GPS). Social network metrics and knowledge production patterns are used as proxy metrics to infer innovation potential of emergent knowledge and collaboration networks. Robust communication strategies with regard to innovation potential are questioned by exploring different parameter and mechanism configurations. My objective is to further our understanding of the dynamics in GPS and facilitate developing informed policies fostering innovation capacity.

Research Approach and Interests

I have good level of skills on programming in Java, Python, R, and C++. I usually use RePast (Recursive Porous Agent Simulation Toolkit) to program simulation environments in my research. I am zestful to study on Complex Adaptive Systems. It all started with the special topics class on Complex Adaptive Systems (CAS) that I have taken at Auburn. I developed an agent based knowledge spillovers model as a project assignment in the class. I am also involved with the project funded by NSF, which is named “Dynamics of innovation and creativity in virtual scientific commons.” The project is focused on development of computational models to study complex dynamics in knowledge creation and innovation in GPS. In Auburn Modeling and Simulation lab, we host OBO database and there are other tools developed for network visualization and data visualization. I conducted user tests and reused the tools to generate graphs of my simulation outputs. I also use D3 libraries, JSON data-interchange format, and JavaScript to visualize force-directed graphs of social networks generated from output data of simulations.

I personally think that CAS approach is plausible and essential to further our understanding on complex systems consisting of heterogeneous agents that have non-linear interactions. Emergent behavior is not easy to predict by static engineering methods’ predictions. We need to foster complex adaptive computational studies on individual or ecosystem level dynamics of today’ s complex problems. CAS approach should be involved more in our curricula supported with its successful applications. As an opinion, the importance of discovery and improving understanding in complex systems should get more attention/credit along with the prediction aspects. But while we are creating computational models to develop understanding, we need to provide detailed documentation regarding reproducibility concerns. Since reproducibility is essential in science, use of UML diagrams, documentation templates (ODD template etc.), and description of the model and the experimental design in a markup language (ModelML, SED-ML etc.) should be encouraged and the models should be shared openly. It would promote re-use of different models and increase the speed of development in computational science by building our work on top of others’.

The opportunities to collect and parse mass amount of data empirically support mechanisms/operating principles in our models. I worked as a customer intelligence analyst querying the transaction data of millions of customers in retail industries, in Turkey. Data mining is a useful method to discern plausible mechanisms to be implemented in our simulation models. We need to grow the environment bottom-up in order to explain it. We can determine initial conditions, build metaphors and analogies examining the real world data and discern emerging patterns to seek in our simulation outcome. Data is essential for validation purposes and to increase model credibility. But, simulation models are not the exact replication of reality. Along with conceptual (theory base and interpretation of the theories) and operational validity (in different scales) of simulation models, I believe that considering the randomness and uncertainty in the complex environments and the lack of real-world data, we must use meta-heuristics to explore and exploit the parameter and scenario space, so that we can find more robust system designs as opposed to find an optimal system design in a scenario. It would also increase generalization ability of our models.

Additionally, during my teaching duties in Industrial and Systems Engineering department at Auburn, I did pedagogical studies to design more flexible project assignments that can fit into different aspects of the curricula. I build analogies between non-traditional systems (non-textbook examples) and the core subject matter to develop “fun to solve” assignments and aim to improve the synthesis ability of the students. I was involved with design of a project assignment in Stochastic OR class to teach students Markov chains by a Golf Match-Play tournament problem – ( I am also interested in adaptive optimization and the use of it to solve real world complex problems. In my dissertation work, I implement genetic algorithm to seek for more robust communication landscapes in GPS.

As a researcher, I want to raise awareness on CAS and further the use of it in policy studies. I have also interest on social problems we face in our daily lives (related with healthcare system, education system, epidemics, smarter urban area designs, and social networks etc.) and want to ask more questions about policy of the systems, not just about optimization of the way the system is. Hence, Verification and Validation studies gain importance to inform policy-makers on the fitness for use of the models. In that sense, reproducibility of multi-agent simulations is another area of interest for me. I aim to get involved with more research focusing on what-if scenarios and getting from policy to practice. I aim to define the systems with operating mechanisms that are grounded conceptually, empirically, and theoretically, then to design experiments to develop understanding of emergent patterns that would lead us to better normative designs.