Egypt I, Grand Staircase/Escalante NM, Utah, May 2018 — Braden taking a look

I was recently prompted to read by my friend, Hanzi Freinacht, about the theory called a Model of Hierarchical Complexity (MHC), developed by Michael Commons , in the ’80s, and expounded on and patented for use in Artificial Intelligence (AI).  It’s billed as an information-theoretic (meaning things end up in bits and bytes) methodology for measuring complexity of thoughts and actions.  As such, it ends up being divorced from culture, social structure, and human behavior, which makes it appealing as an uber-algorithmic be-all, end-all way of describing how humans act.  Efforts have been made (relatively successfully, I might add!) to map these to various developmental stage theorists, such as Jean Piaget.  That’s all well and good.

The way that all scoring schemes have to work, though, is through what I call an integral, or perhaps integrative approach, and I don’t mean in either the typical socio-psychological senses.  I mean it in the math sense, and what THAT means is that, if you remember your first-year calculus, the definition of a definite integral is that you take some function (a drawn-out wiggly line) and over some range, collapse it down to a number.  For those that are now indulging in an age-induced headache, that’s the whole “area under a curve” thing.

You might remember counting the little squares or something.  The bottom line is that you take a complex sequence of information, and come up with a measure/scalar value for A.  From: https://www.mathsisfun.com/calculus/integration-definite.html

Needless to say, Commons’ scheme is FAR more complex than that, with lots of bits in different locations that represent different types of things — namely that complexity builds on complexity, and in order to hop up different levels — very meta-meta-linear!, you first have to demonstrate mastery of a lower level.  For example, you have to know how to add and subtract real numbers before you can figure out how to add and subtract real variables.  For those that this stuff is some kind of mental Adderall, I highly recommend surfing through the Wikipedia table on MHC (the link is the reference given on Wikipedia.  Not surprisingly, it’s algorithmic, and as such, poses as objective.  And it pops right out of the scientists’ social structure.  Agency appears explicitly not at all.

As such, people are using this for calibrating AI (that’s what the patent’s all about) and that’s all fine as well.  But when you collapse the generative dynamics out of the picture, you lose intent, empathetic development, and most importantly, individual agency.

I’ve done a lot of pondering on the complexity of knowledge myself.  For example, you may have highly empathetically evolved, reflective societies or cultures, like some aspects of Tibetan Buddhism, that in the end do a pretty miserable job of providing for the lower-level v-Meme needs of their core constituencies.  Zen monks may do an excellent job of understanding deeper mysteries, of the universe, but they might not do so hot on growing tomatoes.  For me, a better view of complexity gets laid out on a 2-D plot, shown below;

Evolution vs. Sophistication

Where something like MHC would come in is distilling a location in this 2-D space, (for you math people, might be more like a vector) into a single (scalar) value  (kind of like a meta-vector magnitude. )  All my math friends can fuss at me in the comments.

Of course, MHC collapsing that down to a value — that’s its purpose, after all.  So a society could value everyone’s opinion very highly, yet be far less complex than a low-empathy one that used sophisticated authorities and algorithms.  Chinese society is a great example of this — they got stuck with narcissistic authoritarianism about 2500 years ago — and made a very sophisticated society indeed.  But because of their lack of empathetic evolution (look up a deconstruction of the 36 Stratagems here — it’s also on Wikipedia) increasing sophistication led to diminishing returns, and they were easily captured by higher v-Meme, yet psychopathic foreign powers. Those powers took advantage of chemical substances that addressed the pathologies of the depression/low performance that the Authoritarian v-Meme counts on to establish control.  I’m referring to smoking opium, of course. At the same time, that fundamental culture, when mixed in with a little bit of Performance/Goal-Oriented v-Meme evolution stuff (Deng Xiaoping’s ‘black cat, white cat, doesn’t matter as long as it catches mice’) will, in the span of only about 40 years, made 76% of Chinese members of the middle class.

And, of course, we have our own issues in contemporary society with neurobiological hacks (sugar anyone?) that may yet bring us down.  Do the big comparison — in the US, now, only 50% fall under that measure of being in the middle class. More empathy and evolved individuality? Maybe.  More prosperity?  We’re on the backside.

Regarding evolution/sophistication — what would that mean as far as MHC?  MHC might record increasing complexity with an increase in bits, but what do bits in various places actually mean?  It’s a good question, and I’d argue that you could look at a very complex society with a lot of sophistication, and a lot more bits, yet still not understand why it might not be doing well, because the dynamic of creation of those bits would be poorly understood.

The nice thing about empathy and Conway’s Law is we have some deeper insight into the larger ‘Why’ of information creation, and can more positively construct social systems that give us the goods.

When it comes to AI, there’s also nothing wrong with coming up with an MHC score for predicting the potential development time for a particular AI algorithm.  But MHC still leverages the reinforcement/supervised/unsupervised learning paradigm that dominates thinking in AI.  Contrast that to the insight that knowledge structures give.  Algorithms of increasing complexity?  We’ve got that (well, sorta.)  Making the jump to developing strategies that capture individual experiences and the outcomes of independent agency, as well as complex heuristics?  Eh, not so much.  When I can call Apple Help and get the natural language processor to understand how I say the serial number off my Airport router, then I’ll start becoming more interested.

BTW, a tip of the hat to my friend, Hanzi Freinacht, high up in the Swiss Alps, writing books like The Listening Society, that got me started thinking about this.  Hanzi, there must be something in that goat’s milk you drink that makes you so smart in such an isolated environment.  😉

PS:  Thought some folks would find it funny that I’ve been calling a ‘definite integral’ a ‘direct integral‘ for some time now.  Please don’t send me back to Calc I!  Or make me plow through that Wikipedia post!