Enabling policy would cost same as usual.
Disabling policy would cost percent of max cost depending on intensity:
If you had policy X, that has cancellation cost Y and had intensity lets say 0.5, then actual cancellation cost should be 0.5*X
Similarly decreasing/increasing it would cost depending on slider decrease/increase.
If policy X is set to a (between 0 and 1), and had decrease cost D and increase cost I, then decreasing it would cost (a - final intensity)*D, and increasing it would cost (final intensity - a)*D
This way you could choosing between changing 20 policies by tiny bit or 5 policies by larger part.
If policy is set in such way that least controversial level is 0.5 then it should be changed accordingly.
More complicated suggestion:
Also Increase costs, Decrease costs, Cancellation costs and Activation costs themselves should be influenced (or capped) by voter support.
That is specific policies would be cheaper/more expensive depending on voter group membership and voter support.
Try to add Abortion laws in country full of religious conservatives.
If voters are happy, then you could get away with such policies.
Also this would probably mean some policies would be activatable only if you set slider on certain positions.