See also: Wikipedia's article on the human molecule.
Humans are kind of like molecules, even with a sprinkling of quantum metaphors. In massive numbers, atoms and people are generally predictable, but, when examined closely, individual cases disclose a limited amount of unpredictability. Free will is real, even if most people don't exercise it most of the time. Let's say it takes a free will decision to choose to do something new, while doing something habitually is just mechanical. And people are truly creatures of habit.
One can browse the mostly accurate historical record and perceive a kind of evolution of human society, or perhaps technology. The day when "high technology" is cheaply and ubiquitously available will be drastically different from today, but certain social adjustments will need to be continuously maintained no matter what kind of technology is deployed.
And let's also say that a complete model of global human behavior could be developed, along the same lines as the elusive climate model. One could input a bunch of variables, run the model, and see how groups of people form and splinter in a plethora of manners. One could add variables, representing state sponsored programs or potential free will actions that "catch on" and form fads and trends.
There's an inherent limitation, however, when dealing with free will (akin to peering beyond a black hole's event horizon). To compensate, chaos can be introduced into the model, but this would necessarily result in wildly divergent outcomes. To compensate for that, a model should be developed that emulates known history, with known chaotic factors (which would be unknown when running the model into the future). Naming these chaotic factors would be the same as identifying most of the notable figures from the history books, though accuracy will always be in question.