Wednesday, June 13, 2007

Causal Loop Diagrams, stories, and macrobes

One standard tool of Systems Dynamics is the Causal Loop Diagram. This tool is explained at great length in MIT Professor John Sterman's text "Business Dynamics", but a short explanation is given by Daniel Kim in "Guidelines for Drawing Causal Loop Diagrams."
(John Sterman had a paper in the March, 2006 issue of AJPH on "Learning From Evidence in a Complex World", so he's finally been given "judicial notice" by Public Health. Always a good start.)

Kim begins:

The old adage "if the only tool you have is a hammer, every-thing begins to look like a nail" can also apply to language. If our language is linear and static, we will tend to view and interact with our world as if it were linear and static. Taking a complex, dynamic, and circular world and linearizing it into a set of snapshots may make things seem simpler, but we may totally misread the very reality we were seeking to understand. ...

Articulating Reality
Causal loop diagrams provide a language for articulating our understanding of the dynamic, interconnected nature of our world. We can think of them as sentences which are constructed by linking together key variables and indicating the causal relationships between them. By stringing together several loops, we can create a coherent story about a particular problem or issue. [emphasis added]
I haven't been able to get away for a few weeks for intensive training in Vensim or Causal loop diagrams, but they are certainly referred to in the professional literature as being a strong basis around which to bring many different interest groups together and reach a better common undertanding than would be possible without even turning on the simulator.

Still, it appears to me, a relative newbie, that Causal Loop diagrams still suffer from the concept that feedback comes in only two flavors - "positive" and "negative", not the full multidimensional spectrum I described in recent posts for "self-aware, goal-seeking, feedback control loops." Thus, on web sites such as Pegasus Communications, we see the classic "two" kinds of loops, those labeled with an "R" for "REINFORCING" and those labeled with a "B" for "BALANCING" (or "negative" feedback reducing difference from some fixed goal state.) See also Mindtools' description of CLD's with somewhat clearer diagrams.

Here, I fear, the power of the ability to turn on the computer and have it crunch through ranges of estimated parameters short-circuits the process I would recommend -- namely, putting the CLD up on the wall, standing back a few paces, and looking amid all the N-factorial combinations of N "loops" for a few "self-aware, self-protective, self-repairing, goal-seeking, feedback-mediated control loops."

At the risk of hitting a lot of hot-buttons, let me say that these, in my mind, constitute a kind of proto-life, which is to say that they are active agents that "might as well be alive" because they satisfy the usual definitions of "life", which is to say that they:

  • Consume energy
  • are self-repairing
  • adapt to their environments
  • are self-aware
  • seek something akin to homeostasis when disturbed
  • resist being shut down or shut off
  • are capable of learning and becoming smarter
BUT, because the entities I'm describing are creatures of the "control" domain, what flows in their equivalent of "veins" or "neurons" is control information and in particular real-time, real-world phase-lock signals. And, as I've emphasized, control information can easily jump from one medium to another, so it's tricky to track it down and "see" it the first time, although once you "see" it, like most visual patterns, you can keep on seeing it.

So, once again, I bring in the hand and water faucet picture, and the person-driving car picture, to illustrate the number of different stages that a single "control loop" can pass through and draw together into synchronous action.

Aside: The fact that the action is synchronized is all the difference in the world, as that is the difference between a laser-beam and incoherent light, where one cuts steel and the other is a little bright to look at. Most of our real-world measurements, unless we are into Very Long Baseline Interferometry (VLBI), discard "phase" information and absolute time as "irrelevant." VLBI can work even if the telescopes in different countries are not connected physically, if a very accurate record is kept of each signal and the records synthetically reconnected in virtual space inside a computer. But it does require recording not just "amplitude" or "power" but also the phase component of the signal at each antenna -- information we normally discard.

The big Y-shaped array of "dishes" that Jodie Foster was using to listen to the stars with in the movie "Contact" was a VLBI, where the spacing between the dishes, which are on railroad tracks, could be altered to focus on different wavelengths of incoming signals.
Another example with some smaller feedback loops that compete for our attention is the "story telling feedback loop" picture I put up yesterday, again repeated below. Don't try to dig into the details. Just notice that there is a big loop that covers most of the diagram, surrounding the light blue bar -- and that is the main, persistent, "I am a person" kind of loop. Then there is also a smaller loop with a shorter lifetime managing incoming visual input at the lower right, and two competing permanent-fixture loops at the left -- One driven by higher levels reaching downward and trying to raise this person's goals; the other driven by the person's frustration limit and protection against overheating basically, which tries to lower the goals again until they are achievable.

It requires reflection, and explicitly asking the question, to realize which loops are self-aware and self-repairing if damaged.

Consciousness certainly keeps shutting down every night, but it recovers the next morning, usually. The visual system has many small loops that leap into action when triggered, then go back to sleep. If we lost them we'd be essentially blind in a sea of unfiltered noise.
Aside: I'm not sure about the loop I drew in the upper right, where the person's actions are echoed back to them, with various lagtimes, by different parts of the environment. Maybe that's just the classic, passive, "environment" that's envisioned by epidemiology -- with the same intelligence and adaptiveness as a canyon's walls. Or, maybe the world develops its own set of ruts and habits around reacting to you, as an irritant is surrounded by pearl inside an oyster, and those prove to be so useful that they are endowed with self-aware, self-sustaining, independent status to keep an eye on you and provide very fast feedback, as if you'd touched a stove, when you try to harm the world. That's a sort of meta-sociological question touching on guardian processes, for some other day.

Cutting to the chase: Hypothesis: Because self-aware, self-repairing agents survive noise and damage that will disrupt other, dumb, passive, accidental "loops", they will tend to end up dominating the landscape -- even if they don't reproduce or form support alliances. But, they will tend to form support alliances too.

Similar hypothesis I put forward a month or so ago: Because organizations tend to find and fix small-scale, non-complex problems, if we assume problems arise due to noise at every level in some equal amount, then the large-scale, "complex" ones will end up dominating the landscape, because those are the ones that keep getting put off and not addressed or fixed.

Synthesis of those two: In any long-lived multi-level complex adaptive system, large-scale, complex, active, self-aware, self-repairing control loops will end up dominating the landscape and being the primary shaping force.

And, sad corollary: Until we build scientific tools that can glance at a picture like M.C. Escher's Waterfall, and "see" at a glance "where" it is "broken", we will continue to be plagued by these large scale active agents.

We are, in fact,most likely swimming in a sea of semi-alive "macrobes" -- a concept probably as distressing as Pasteur's "germ theory" that had a sea of "microbes" swimming inside us. They would certainly be as "alive" and as annoying as viruses, and if they were not well, we would feel it, being, as it were, inside the "whale".

Of course, before going into anaphylactic shock at the idea of macrobes, I should point out that you already are familiar with some of them, as a big "yawn." Those would include persistent, self-aware, self-repairing, energy consuming, possibly self-extending macro-agents known as "families", "corporations", "cultures", "religions", and "nation-states." If the Gaia theory is correct, it would also include the Earth as a whole. If religions are even partly correct about some big issues, it continues at scales much larger than the Earth. However, the larger such an agent would be, the more slowly changing it would be, and at some point we could locally treat it as "fixed" or a "constant" for planning daily activities.

So, if sociologists, and even untrained civilians recognize that corporations and countries exist, what's the big deal here? What contribution to our collective wisdom am I suggesting this framework brings to the table?

Again, the most important point I'm making has three parts:

Hey everyone strugling with methodologies for feedback and multilevel systems in Public Health! Control System Engineering already solved that! Read the Literature!


Hey everyone in Control System Engineering! You have some potential new clients over here in Public Health!

Finally this one: Children! Stop fighting!

Public HealtH? stop picking on corporations -- the healthy ones hold your planet together right now. And the diseased ones need your insight and techniques to be healed -- once you master multi-level organism healing techniques. And, Hey, CEO's? :Please stop kicking Public Health in the shins -- they're trying to keep your workforce alive and healthy and productive, and besides they're closer than anyone to understanding The Toyota Way in terms of a health multilevel organism. Religion please stop picking on Science, and vice versa!

And, everyone, there's a big qualitative difference between a "distal factor" and the big toe on your other foot, so before you bite down .... oh, never mind. You'll find out soon enough!

By this model, there really only is one multi-level life form occupying this planet, and while it is the job of clinical medicine to heal people at the 1-body level, it is the larger distinct job of "Public Health" to deal with disharmony at any level -- between cells and cells, people and people, cultures and cultures, nations and nations, corporations vs. corporations, departments vs. departments, silo versus silo within hospitals, etc.

Because all that will persist is actually connected through all those loose-couplings (amplified by compounding feedback loops over long times), in a "control" or "regulatory system" sense, it's all only ONE body. We share parts of it, or levels of it. But it's hard to have your own foot have gangrene and not be affected by it, sooner or later.

The biggest problem right now is that the healers of society cannot easily see the feedback loop connections and evaluate the strength of each link and of phase-locked groups of links. That's the missing toolset. And that already exists, but indexed in a different literature where public health seldom treads. Now, with the new competency (2006) for MPH students from the ASPH, the focus on "systems thinking" will lead us there. The March 2006 AJPH is a start, but our work is cut out for us.

It puts a kind of different light and torque on things if we assume there is only one Body here, with many pieces and parts, that we're trying to heal and make right. It won't do to fix most of the body but leave a tooth or limb infected -- that'll turn around and bite "us".

If every level and pair of levels had different rules, this would be a huge problem, probably intractable. BUT, if every level and pair of levels has the SAME set of rules in "control space", then instead of many levels being harder to "solve", suddenly many levels becomes more hints and easier to fix. We have one equation and one unknown and 50 clues, not 50 equations with 50 unknowns and no clue.

That's WAY BETTER. Just align the fragmentary knowledge of the control structures of each level on a mental transparency, then put them on top of each other at the same scale and orientation, and look through the whole set, and all the clues will line up and reveal the full picture that applies at every level, even though we only have a little bit of it right now on each of those levels.

The prospect is compelling. It's a win-win-win solution, and we might just be able to get every field to give up 1% of its budget to work on this single problem that is relevant to working out more details in that field, for every field. It could be politically acceptable. It might fly.

Sphere: Related Content

1 comment:

Ryan Lanham said...

Reblogged...let me know if you'd like me to remove...

Thanks for the interesting post...