Most of the papers above look at emotion as a necessary switching mechanism that would have to be properly integrated with a general-purpose planner, diverting its focus when necessary. It's a nice idea, and may help us in building better programs. But --- considering that the physiological linkages on which emotion signals are based came long before rational thought --- is it biologically plausible?
Consider what I asked you to read in the Unconventional Approaches section of the planning tutorial. The approach there is to build up through the evolutionary tree, making artificial organisms that mimic the competences of real animals. These animals do not represent goals, subgoals, tasks, plans, symbolically in the same way that conventional planners do. Consequently, it would be impossible to build emotional switching on to them. Yet, if they're based with sufficient accuracy on real animals, they'll embody some mechanism corresponding to emotion. Where does this leave Oatley, Johnson-Laird and Sloman? No references yet, but think about it.