🚀 go-pugleaf

     I just submitted a paper which touches on this subject to the
1990 Artificial Life conference.  

     One key idea: some emotions are spatial.  This applies most directly
to fear.  Fear comes from a direction.  It is not a scalar quantity.
A scalar field, however, is a useful model of fear.  Such a model can
be used to motivate behavior directly, using Khatib-type fields or
Kass-type optimization.

     Being paralyzed by fear occurs when the organism is caught in a
local minimum in fear space.  Panic is a method for breaking out of
minima by injecting noise, as in simulated annealing.

     It helps if you try to think about this in terms of lower animals,
at, say, the mouse level.  The minds are simpler and the philosophical
issues less annoying.