The idea is a bit of a guess. But with so many parameters in music software I thought why not write out my hunch. The idea is to calculate a reduced set of new ranges for the parameters you have currently available to you. The logic is. There should be values in those ranges that when combined with other parameters make no sense. They make the music sound bad. On the other side. There should be combinations of parameter values that are popular. If you collect this data it should then be possible to present this to the music creator in the form of reduced parameter settings. Here I guess you can group the whole range to from bad, possible and better. Further I assume the parameters are likely to be dependent so when enough parameters have been initially chosen. The rest should have a probable range within they could vary.
In the image below. The different range groups are shown. Green shows the range which is calculated to have the best sound and the red range should be avoided.