Social scientists are generally reluctant to predict civil strife. There are several reasons for this hesitancy: 1) the open system within which outcomes are derived, has been considered far too complex to predict anything more than the direction of trends or the increased likelihood of an event occurring, 2) the data on civil strife had to be counted and then published for typical researchers to use, which a created a huge time lag between data collection and their analyses. As a result, the efforts of social scientists have largely been confined to the understanding of the determinant variables using various types of causal or time-series models. These models have provided very good fit for the sample data for various types of conflicts and social strife, but have been rarely used for out-of-sample forecasting. Recent rise of social media is witnessing the creation of a brand new multi-disciplinary methodology for forecasting short term, measured in days or weeks. For the past three years, the author has been a Co-Principal Investigator of a multi-university team developing models for forecasting incidents of civil strife in Latin America, the Middle East and China, funded by IARPA (Intelligence Advanced Research Project Administration). Our forecasts are evaluated for accuracy by a third party for accuracy. In this paper, I will present the methodology, results of forecasting, and discuss the ethics of such an endeavor.