Simulations, situations, voter happiness can range between 0 and 1.
Those values should be reachable in infinity like logistic function:
If negatives and positives balance each other, then function would end up very close to 0.5 after lets say 10 turns.
To reach <0.01 or >0.99 you would have to make sure that pushing in one direction is way stronger than pulling in other direction, essentially tug of war.
Some things would have higher inertia and others could have lower inertia - that is slowly/quickly reacting to sudden changes to input.
That is things should move like in molasses not like in vacuum.
Simplest way to do so would be adding 50 or so hidden simulations.
For example one affecting health simulation:
Cause: Health,0+(1*x) - reacting linearly to this simulation.
Effect: Health,(1-x)^N-(x^N) - strongly pushing it up when close to 0 and pulling it down when close to 1.
Higher N would reduce effective range closer to borders.
So N = 1 would mean proportional effect, while
N = 20 would mean effect appearing only if simulation is very close to 0 or 1.