The goal of this project is to create a cognitive agent simulation framework for studying social behavior in online information environments. We are developing a scalable, virtual laboratory, calibrated on real-world data, for studying dynamics of online social phenomena and information diffusion at different temporal resolutions and at multiple scales, from individual to community to global collective behavior.

Individual agent models within COSINE will be based on first-principles of human behavior uncovered through empirical analysis of the vast troves of online behavioral data. These models will incorporate bounded rationality and cognitive biases within models of attention. In addition, COSINE’s multi-resolution, scalable framework will enable time-resolved, massive simulations of dynamic information environments.

Key to COSINE’s success will be its ability to model complex phenomena arising in multiplex networked information environments. To achieve this, COSINE will incorporate networks into interactions between agents, thereby enabling the study of the interplay between individual behaviors and network structure, including (a) how individual traits affect where in the network individuals position themselves, (b) the information environment these positions produce, and (c) the impact this has on individual behavior. Finally, the system will be calibrated on real-word data collected from a plethora of online platforms.

Team

COSINE is a collaboration with colleagues at the USC Information Sciences Institute (Emilio Ferrara, PI; Kristina Lerman and Jim Blythe, co-PIs) and the University of Notre Dame (Tim Weninger, co-PI). The IU team:

Former Team Members

Support

This project is supported by the DARPA SocialSim program under contract W911NF-17-C-0094.