For control signals it is largely academic. Main use is for i/o, as many cpu's either float or go high on the pins during reset, it is not a great idea to have peripherals briefly activate during this time, so active low is the preferred option. On the input side it seems that historically active low was the preferred option. This of course would allow a pull down to a common logic ground from different votage sources. Also if it is in relation to cpu's, usually you would have internal pull ups you could enable, this will save a few components, so is a little simpler and cheaper to manufacture. May also be better for noise consideration. Another reason is that many older processors could sink a lot more than they could source, useful for output, but not so relevent these days.