Consider Herman, a person.  Considered as an individual Herman is a very complex thing.  He is a product of a myriad of interactions between his genes and his physical and social environments through time.  He—like all of us—has a brain with billions of neural connections, a “machine” that is as or more complex than the universe.

Scientifically speaking we understand very little about Herman.  We certainly cannot accurately predict what he will do over a long stretch of time and cannot totally understand what he did after he did it, since there are so many interacting variables at play in Herman’s brain, genes, environments, and history.  Each individual is a complex system (in the technical sense).

If we want to “understand” Herman, we have to consider him as member of a group and then study the group (with statistics, for example).  But, then, we have lost Herman and are making assessments about average members of the group, not about Herman.

In fact, we often cannot be sure that we have the right group for Herman or that Herman is not actually a special case (perhaps a member of a special sub-group of the larger group).  In either case, our statistics are not going to tell us much—at least not much that is correct—about Herman.  And we are going to miss lots of things that we could have studied had we got our categories right (though we cannot usually be sure they are right and that there are not better ways to carve things up).

People do not fall “naturally” into most categories.  We have to argue that we have the right people in the right categories, which is more like herding cats than we think.  We are now well aware of this when it comes to gender, but it is true of many other socially significant categories.  How much “blood” do you need to be Native American or “black”?  How little money or education do you need to be poor?  What do you need to like or hate to be a terrorist?  Are grades the right measure of who is a “failure” at school?  Where do we put people who play video games regularly but do not call themselves “gamers”?  Is Herman a “poor reader” if he reads two grades below his grade, but reads World of Warcraft manuals several reading levels above his grade?

Do people who do not approve of abortion but do approve of capital punishment belong in the same category (“pro-life”) as someone who abhors both?  Do people who believe in “choice” in regard to abortion but not for gay marriage belong in the same category as someone who approves both?   Is someone a “libertarian” when they want the government out of everything but other people’s bedrooms?

It turns out that a great many of our judgments about Herman based on measuring him as a member of a group are wrong.  Work in education and medicine claims to tell us “what works”, but often it doesn’t work for Herman (for all sorts of different reasons, but the basic one is that Herman is a complex system and controlled studies do not apply to complex systems).

So, what to do with Herman?  What to do with Herman the child in school or Herman the patient in a hospital or Herman the prisoner in prison? Well, as I have said Herman is a complex system and we would have to study him the way we do complex systems in physics or meteorology.

We would have to watch Herman closely over time, collect all sorts of Herman interactional data over time, iterate Herman (that is, watch him do similar things over time), perhaps build Herman simulations, and for the results of all our Herman watching and Herman simulations, we would have to engage in post-hoc explanatory theory building always open to revision.  This is what some people call “qualitative research”, though, ironically, it is standard in the “hardest” science, physics.

Why would we do this?  If we do it, the predictions we make about Herman will probably be no better than weather predictions (which are better than nothing).  But we will be able to make and revise better and better judgements about how to help Herman in school, in the hospital, in prison, and in society.  Here, for sure, “better than nothing” is better and surely better than bad ideas based on groups to which Herman does not really belong or belong in any way useful for what we need to do for Herman here and now.

Of course, we will always need categories and statistics (though the statistics we usually use in our journals stink).  There are too many humans to study all of them individually.  But let’s have the humility to admit that our statistics do not capture Herman. And let’s have the intelligence and morality to study Herman when he is in our care, and not just groups, even if that means telling the quantitative scientists to “stand back” (are you listening teachers?).