I currently gather my own daily day - manually. I am also involved with the team that is creating SkedPal and I am recognizing that there are opportunities to gather all sorts of new data via this programme and others like it called auto-schedulers.
Because it's a programme that produces a calendar run daily, in the future it should offer rich opportunities to gather the kind of data mentioned by justintimmer.
At the moment, Google Goals and Todoist Smart Schedule promise to build a schedule for you that's based on your personal history, plus that of others. However, I haven't seen too many actual success stories, which makes me think that they are trying to do too much at once.
In my opinion, in the beginning, users just want a way to auto-schedule themselves accurately and the best information to use isn't based on machine learning, but an understanding of their own behaviour. However, as time goes on, an app could certainly include this kind of information.
I think it would supplement user input on each task, rather than replace it. If the kind of data that justintimmer describes were available, then users could build higher quality schedules because they are using better-quality feedback.
I hope this helps,