(cross-posted from my Substack)
Learn It by Heart?
Thoughts after reading Stafford Beer’s The Heart of Enterprise
I left my previous piece on Stafford Beer in an ambivalent place, and I decided to follow up by reading The Heart of Enterprise. I’m glad I did! Although I’m still not completely sold on the utility of Beer’s viable system model, I feel like I have a much better understanding of it from the way it was presented here and more sympathy for the argument that it can make sense as a general model.
Meditations on Variety
Before I get into reflecting on the book, part of the benefit of reading these books was being exposed to Beer’s “variety engineering” perspective, and the way he thinks about “variety” and information in general. It helped me develop some of my own thoughts on the topic, although I think they’re still developing. One point Beer makes is that all systems have “transducers” when going from one domain to another (as a simple example, your ears and sensory cortex are transducing from vibrations in the air to what you hear), and often this process involves attenuating the variety that is represented. Humans are great at being able to zoom in and out, both visually and conceptually, but we can’t manage infinite gradations, so the range of variation on a property can influence how much variety we perceive it to have.
And we’re often not trying to perceive properties in isolation, I suspect there may be some interesting things going on when we try to think about properties in combination and one of them has a stark attention-grabbing aspect to the variety on display while the other is more subtle – it may be hard for our minds to independently zoom in or out. This may be why it seems beneficial to organize things so they can be engaged with at different scales.
One reason I’ve liked Beer’s way of talking about variety is that it seems relevant to some of my subjective experiences. When I write I have to work past my anxiety about judgment or other negative response, but it isn’t that there’s some specific criticism I’m dreading it’s that in the absence of anything else to work with it feels like any possible judgment could be incoming. That tracks for me in that not knowing what someone might say means they have greater variety than if they’ve said a specific thing (Just like “maybe” has greater variety than “yes” or “no”). But Beer also says that variety will tend to increase to match across interfaces, and isn’t being creatively blocked a low-variety state where you do nothing? I’m wondering if I was seeing it backwards – in nature a prey animal that senses danger will stop what it’s doing in order to prepare to fight any threat or flee in any direction, maybe the “frozen in anxiety” state is actually high-variety that only looks low-variety from certain perspectives.
From a game design point of view I think that speaks to things like “analysis paralysis” and the “paradox of choice” problem – if someone is presented with a huge range of possible options they may find it hard to select any (especially if they have a hard time evaluating the choices, such as when a player is first learning a game). It’s also why I’m skeptical of people leaning heavily into pitching “you can do anything you want!” when it comes to tabletop RPGs, I think that a comprehensible environment is more conducive to eliciting action than trying to amp up the perception of variety.
One of the examples of a “variety attenuator” is the way that an insurance company will map any particular person to the categories that their insurance model understands: you’re a unique individual, but to the insurance company you’re an entity in some health-category, some income-category, some age-category, etc. If the company can only think in terms of those restricted categories it limits the variety it can engage in while dealing with you. This variety attenuation phenomenon might be why I’m instinctively skeptical of what I think of as the “taxonomy first” model of RPG Theory – asserting that we need to be able to classify different types of games so that we can study how the different types work. I think a more fruitful approach may be what I think of as “anatomy first” – figure out how parts of particular games work and interact without an urgency to impose a category constraint.
More Beer? Bottoms up!
In contrast to Brain of the Firm, in The Heart of Enterprise Stafford Beer starts at the bottom with some basic building blocks and builds up to his full model. I found this to be far more comprehensible than the other book because it gets him to talk about the components functionally, rather than the model appearing fully formed and being explained by analogy. He starts by trying to make the case that a two-level controller, basically a controller that interacts with the environment and a meta-controller that operates on a different timescale and tweaks the controller’s parameters, is capable of handling a greater variety in its environment than a single monolithic controller would be. I found this appealing because it reminded me of the Reasonable Mind / Emotion Mind model in DBT, where the balanced engagement of both types of cognition enables you to handle things better than letting one or the other run the show. The differing timeframes aspect reminded me of the argument that emotions are strategic evolutionary adaptations that work on a different timescale than the immediate – for example, in the short term the “smart move” is usually to give in to a bully because whatever they’re demanding is probably less bad than the consequences they can inflict on you, but if you have to always give in to every demand that would be bad in the long term, so we have anger to sometimes influence our short-term “smart” decision-making to potentially push things to a better place long term. That seems to map very well to an emotional meta-controller influencing the settings on a rational controller.
An awful lot of oscillations
Beer calls a collection of viable systems that are grouped together as part of a larger whole System One, those are the systems that are directly interacting with the environment. And to help them, they need an “anti-oscillatory” System Two. Although the idea of wanting to damp out excessive oscillations made sense to me from an analogy to electric circuits, in this book he gave some business-world examples that made the idea more concrete and coherent for me. Imagine a company where the Parts Division makes parts that are used by the Final Assembly Division to make a product. If operating completely independently, normal variations could end up producing bursty feast-or-famine cycles even if the average level out output from the Parts Division matches the needed average level of input by the Final Assembly Division. Imagine that the Parts Division shuts down its production line for a little while to run some maintenance. Meanwhile, the Final Assembly Division is running faster than usual so they can make time for a teambuilding event. Suddenly they find that the warehouse is critically low on parts since they’ve been using them up and the Parts Division hasn’t been delivering new ones. The Final Assembly Division starts freaking out about the crisis. Hearing that there’s a crisis in Final Assembly related to a lack of parts, the Parts manager cancels all breaks and tells the division to work overtime. After a little while they have a different problem: the warehouse filled up and they’ve got nowhere to put the extra parts coming off the production line. The coordination between the divisions that should keep that kind of thing from happening is System Two.
All services, surprisingly, at first, inhibit variety to some extent. If I have my hair cut, for example, to a particular style, then several other styles that were immanent in my unkempt hair have been lost: variety has been reduced. I do not however consider that I am being oppressed by my barber, because I gladly expected that he would reduce my hair’s variety in this way. Equally, the viable system engages the services of System Two to cut down the variety of its operational interactions insofar as they are inherently oscillatory – and only to that extent.
Showtime, Synergy
An independent company has a lot of degrees of freedom for how to operate. Imagine a tire company: they might specialize in rugged off-road tires, high performance racing tires, low-cost economy tires, etc. Any of those could be perfectly reasonable choices. But if it’s not an independent company but a division of a company that makes vehicles, in order to have synergy they need to be making the kind of tires that work well with the kind of vehicle you get by assembling the parts from all the other divisions. If you’re trying to put tractor tires on a motorcycle the company (including the tire division) is probably going to go out of business. System Three is the management process that guides the Systems One toward choices that work synergistically rather than at cross purposes with each other. As a business example, Beer treats accounting that keeps divisions legally compliant as a System Three activity – if a single division is cooking the books and gets caught that can end up hurting all the other divisions that were trying to play it straight.
I was amused that the book used “driving on one side of the road” as example of non-oppressive variety constraint since I used that in a previous post. Here’s Beer’s version:
Most of us are motorists. If it were not for an instruction – a metasystemic intervention – which says: ‘drive on this side of the road’, the traffic system would surely, and finally, stop. The driver continually applies this rule, which is a low variety statement, and thereby amplifies it infinitely (or for as long as he continues to drive). Thus it is true that his liberty to drive wherever he likes is massively impaired. Even so, he has no sense of oppression resulting from an overloading of his channels to System Three. Now suppose that no-one had been able to think out this low-variety rule. How could System Three possibly achieve its purpose of keeping the traffic moving? It would have to ask a driver to submit a plan of his route in advance; it would have to plot a route on a map showing the driver’s movements on this side of the road for every inch of the journey; and the driver would need to carry a passenger to read him the continuously unfolding instructions. The resulting loss of freedom would be identical; but the sense of loss would surely lead either to emigration or to walking.
This distinction is by no means trivial. Part of the autonomy which the elemental operation must feel that it exerts resides in the knowledge that it is trusted. It is trusted to invent amplifiers that enable it to handle its own horizontal problems. It is also trusted to apply simple rules, that are handed down in the interests of either legality or synergy, in an appropriate way. It is its own variety amplifier, despite the simple rules that largely constrain its variety deployment once they are amplified. Yet the reasons for this are understood, and there is no sense of oppression. The sense of oppression arises when there is no trust, and therefore every move is predetermined from above.
Because System Three is where all the activities of the Systems One “come together” (in a lower-detail, abstracted way), it’s kind of the hub of “internal operations” of the overall system. Beer often refers to Three-Two-One as three systems that naturally clump together.
There’s An Outside Chance
While System Three is focused on what’s happening inside right now, System Four is about looking outside and into the future. It’s what is finding threats or opportunities that Three-Two-One might need to deal with. In the book there’s a case study of an insurance company that Stafford Beer was consulting with, and he was urging them to consider that they didn’t have a System Four at the corporate level since all the market development was happening at the product-line level. He was telling them they weren’t spending enough time and energy thinking about things like what isn’t typically insured now but could be helped by insurance or what financial products can we offer that leverage our expertise with risk that don’t look like conventional insurance?
Number Five Is Alive
Beer is big on the idea that “variety can absorb variety”, so you design the systems such that they interact and achieve a sort of balance. System Five is the thing that coordinates the balance of what’s happening inside (Three) with what’s happening outside (Four) and prevents them from getting into oscillation. Thus this Three-Four-Five clump is the other natural grouping (with Three conspicuously being part of both). While I was happier with how the other four Systems were explained here, System Five is still a bit mysterious, although maybe that’s the best that could be done. Sometimes System Five is the senior management trying to guide the business, sometimes it’s the collective will of the people in a country, sometimes it’s just “conscience”. I think there’s supposed to be a zen-like quality to it, which in some sense sits well with me, but in other ways is a bit too ephemeral to be confident I’m understanding the model correctly.
Yo dawg, I heard you liked this model so we put this model in your model
Having built up Systems Two through Five, the book circles back and says, “Hey, wait a minute, haven’t we just described a system that is composed of an environment, something that interacts with that environment, and some parts that function as a meta-controller? Why that’s what System One is! In a series of increasingly hard-to-parse diagrams the book presents the argument that the model is recursive up and down, that any viable system contains and is contained in a viable system. Because this version of the model is based on elements that are more plausibly universal (the parts need to work together, what’s inside needs to deal with what’s outside) I find this argument more palatable than the one presented in Brain of the Firm. Like a theory that asserts that all stories have a beginning, middle, and end it’s easier to concede that there will always be some way to map particular things to this model.
The recursion element gets to Beer’s somewhat spiritual take on how System Five is supposed to work – it should be in tune with the infinite recursion. This somewhat resonated with me in the sense that it seems to map to other ideas that make sense, such as that it’s good if you’re able to take pride in your work and feel like you’re making positive contributions to the world and not feel like you’re just exploiting or extracting.
A stylistic note
The book engages in a convention of having between-chapter interludes where a group of fictional characters who have been reading the book and just finished the same chapter you have meet in a bar and talk about it. While initially it struck me as hokey, over the long run I found it surprisingly effective to have a little virtual book-club built into the book itself. Maybe a technique worth stealing?
Should I open my brain to the psychic maelstrom?
One way to see if a tool is useful is to try to use it and see what happens. An idea I started seriously contemplating while reading this book is to see whether mapping a TTRPG’s system to it would make sense, and Apocalypse World seems like an obvious candidate. While I respect Apocalypse World’s game design I wouldn’t exactly call myself a fan. But over the years a lot of people have seemed to misunderstand things about how the game works (both fans and detractors, and in my opinion even the designers), and I had tendency to slip into “someone is wrong on the internet” mode in reaction. One of my back-burner ideas for a post is to try to do a deep dive into explaining how I think Apocalypse World works. So maybe a post that tries to map Apocalypse World to this viable system model would be a good way to do that. Or maybe that’s trying to do two ambitious things at once to the detriment of both. Let me know if this is something you’d like to see me do or if you have other thoughts on the matter.