functional determinism or overfitting to chaos

by Sebastian Benthall

It’s been a long time since I read any Foucault.

The last time I tried, I believe the writing made me angry. He jumps around between anecdotes, draws spurious conclusions. At the time I was much sharper and more demanding and would not tolerate a fallacious logical inference.

It’s years later and I am softer and more flexible. I’m finding myself liking Foucault more, even compelled by his arguments. But I think I was just able to catch myself believing something I shouldn’t have, and needed to make a note.

Foucault brilliantly takes a complex phenomenon–like a prison and the society around it–and traces how its rhetoric, its social effects, etc. all reinforce each other. He describes a complex, and convinces the reader that the complex is a stable unit is society. Delinquency is not the failure of prison, it is the success of prison, because it is a useful category of illegality made possible by the prison. Etc.

I believe this qualifies as “rich qualitative analysis.” Qualitative work has lately been lauded for its “richness”, which is an interesting term. I’m thinking for example for the Human Centered Data Science CfP for CSCW 2016.

With this kind of work–is Foucault a historian? a theorist?–there is always the question of generalizability. What makes Foucault’s account of prisons compelling to me today is that it matches my conception of how prisons still work. I have heard a lot about prisons. I watched The Wire. I know about the cradle-to-prison system.

No doubt these narratives were partly inspired, enabled, by Foucault. I believe them, not having any particular expertise in crime, because I have absorbed an ideology that sees the systemic links between these social forces.

Here is my doubt: what if there are even more factors in play than have been captured by Foucault or a prevailing ideology of crime? What is prisons both, paradoxically, create delinquency and also reform criminals? What if social reality is not merely poststructural, but unstructured, and the narratives we bring to bear on it in order to understand it are rich because they leave out complexity, not because they bring more of it in?

Another example: the ubiquitous discourse on privilege and its systemic effect of reproducing inequality. We are told to believe in systems of privilege–whiteness, wealth, masculinity, and so on. I will confess: I am one of the Most Privileged Men, and so I can see how these forms of privilege reinforce each other (or not). But I can also see variations to this simplistic schema, alterations, exceptions.

And so I have my suspicions. Inequality is reproduced; we know this because the numbers (about income, for example), are distributed in bizarre proportions. 1% owns 99%! It must be because of systemic effects.

But we know now that many of the distributions we once believed were power law distributions created by generative processes such as preferential attachment are really log normal distributions, which are quite different. This is an empirically detectable difference whose implications are quite profound.


Because a log normal distribution is created not by any precise “rich get rich” dynamic, but rather by any process according to which random variables are multiplied together. As a result, you get extreme inequality in a distribution simply by virtue of how various random factors contributing towards it are mathematically combined (multiplicatively), as opposed to any precise determination of the factors upon each other.

The implication of this is that no particular reform is going to remove the skew from the distribution as long as people are not prevented from efficiently using their advantage–whatever it is–to get more advantage. Rather, reforms that are not on the extreme end (such as reparations or land reform) are unlikely to change the equity outcome except from the politically motivated perspective of an interest group.

I was pretty surprised when I figured this out! The implication is that a lot of things that look very socially structured are actually explained by basic mathematical principles. I’m not sure what the theoretical implications of this are but I think there’s going to be a chapter in my dissertation about it.