Ah yes, the genetric algorithm (GA). Well, in short, scheduling personnel is a very visible, policital, and important job. It affects people's lives, and has to be done farily and correctly. I did my best (by hand) for a few years (scheduling 50ish professors into 200 class sections), and I knew I was doing a bad job, but had no choice. It would take 2 weeks to do, because I could only mentally handle working on such a job for so much in a given day before I had to set it aside. So I came up with a preference form, that each could fill out dealing with things like preferred time of the day to teach, daily schedule compactness, preferred class levels, etc. These preferences could also be weighted (0..1) by importance to each person. So the GA starts out with a long, two-part inital random sequence that consists IDs to all professors followed by IDs to all possible classes. Using crossover and mutation (the usual GA steps), the order of the professor IDs vs the order of the classes (prof 0 gets class, prof 1 gets class 1, etc.) are shuffled by the GA. (There are a lot of strange constraints to this you'd only see in academia, related to pay, and contractual teaching requirements.) A fitness is calculated for each professor's latest schedule, by computing a weighted sum of their preferences, based on the assigned classes pulled from the long GA sequence. It's incredible how well it works. It is done in under an hour, people seem happy, everyone's schedule fitness is minimzed, and each's schedule fitness is explained to them (i.e. you got +3 points because you wanted an 8am class, and got one). Everyone can also see everyone else's fitness so it's a "share the pain" kind of resulting schedule. At times when we go in to tweak things by hand, it is clear how tightly the GA optimized things, as any human intervention we try usually and obviously makes things worse. Bottom line, the GA generates the CSV, which Selenium types into PS. After using the GA, I'm convinced that neither humans (or spreadsheets) should be invovled in scheduling at all. I also think GAs are also pretty incredible.
