The simple approach is to treat each degree of freedom (DOF) as a damped spring. The elastic forces look like this for each DOF:
4 values per DOF, in radians: the min and max for range, and the min and max for the relaxed region. Some joints don't have much of a relaxed region at all - they act like simple springs where any displacement produced a proportional response up the limit possible. Others definitely exhibit a wide range that doesn't cause any spring-back.
Such a simple approach certainly has its drawbacks - for instance, it does a lousy job of handling the irregular range of joints such as the shoulder. However, it's probably valuable to see how far this approach will take us, before moving up to a more complicated scheme. It's probably not a coincidence that other v-human systems, notably Marilyn and DK-96, show significant, visible flaws in their shoulder rotation and placement of upper arm. It's a difficult joint!