We report on the ongoing development of a research framework for dynamic integration of information from hard (electronic) and soft (human) sensors. We describe this framework, which includes representation of 2nd order uncertainty. We outline current and planned human-in-the-loop experiments in which an "ad hoc community of human observers" provides input reports via mobile phones and PDAs. Our overall approach is based on three pillars: traditional sensing resources ("S-space"), dynamic communities of human observers ("H-space" and resources such as archived sensor data, blogs, reports, dynamic news reports from citizen reporters via the Internet ("I-space"). The sensors in all three of these pillars need to be characterized and calibrated. In Hspace and I-space, calibration issues related to motivation, truthfulness, etc. must be considered in addition to the standard physical characterization and calibration issues that need to be considered in S-space.