I was working in retail, which means we were trying to forecast human behavior—how much of a given item people would want to buy at some time—but we only cared

I was working in retail, which means we were trying to forecast human behavior—how much of a given item people would want to buy at some time—but we only cared about that behavior in aggregate, not at the level of individual people. That definitely makes it qualitatively different to the sorts of prediction and optimization that the paper talks about.