[The Panther]

I just conducted a quick test. My initial hypothesis is as follows:
Positive dominion = reduce unrest by a random integer from 0 to 2.

It did not seem to matter what the actual dominion level happened to be. It is possible that a higher dominion value made the unrest decrease by 2 more often than 0, but perhaps not. I certainly did not run enough tests to make a claim one way or the other.

The results of my unscientific test:
0 unrest reduction - 26 times
1 unrest reduction - 25 times
2 unrest reduction - 31 times

This small sample certainly supports the random theory. And yes, I did have a province with only 1 dominion reduce unrest by 2. I also had a 10 dominion province with a 0 result.

Using this untested theory, you can therefore assume that unrest will reduce by about 1 per turn per province in positive dominion.

Using Alneyan's opposite formula hypothesis, you can perhaps also assume that enemy dominion increases unrest by 0-2 with an average of 1 per turn.

Cheers

Edit - Oh, I used order 3, which may also have an effect.