E-MEME: Epidemiological Modeling of the Evolution of Messages

We are now fighting"wars of perceptions,"according to General Stanley McChrystal,"in the minds of the participants."Missions operate at multiple geographic and temporal scales villages to nations, hours to months. Planning, execution, and assessment of counter insurgency and related missions require understanding the effects of our (and others") actions on those"perceptions."In Phase I of E-MEME, we successfully applied epidemiological models to reproduce empirically observed diffusion of ideas"memes"across electronic media, using real data from the Muslim world. We applied statistical language processing methods to extract memes, parameterized simple epidemiological models based on the observed patterns, and ran those models to reproduce the patterns. Details of our results strongly suggest that we modeled signal rather than noise. In Phase II, we will: add more data sources, including fast-paced social media like Twitter; make some straightforward engineering enhancements; and extend our innovative methods in both meme extraction and epidemiological modeling. The resulting technologies will be embedded in a collection of web services, making them readily usable in various applications. This work will be of direct operational relevance for intelligence analysts, watch officers, and for military operations more broadly at tactical, operational, and strategic scales.