Surveillance in Counter-Terrorism and Counter-Insurgency type environments requires exploitation of all possible types of available information. U.S. experience in Iraq and Afghanistan has shown the advantages of incorporating surveillance from both trained and inexperienced human observers, and the importance of accounting for certain types of Contextual Information to include cultural factors and non-physical data such as data regarding a political situation (such data now labelled as “soft” or unstructured), along with integration of the data from traditional electronic sensors (labelled as “hard”). Because much of these soft data cannot be well-qualified or calibrated, the evidential basis for decision-making is one of only partially-known uncertainty, creating a requirement to consider new decision-making paradigms for decision-making under extreme uncertainty and unknown/partially-known uncertainty, and for learning strategies based on optimized query-response techniques. This paper will address the design issues and approaches for prototyping new Information Fusion processes that are able to combine all of the data types mentioned above, linked to a decision-support framework employing these new decision-making paradigms.