Some thoughts on Scaling

post by DavidXYu · 2022-05-10T20:14:00.181Z · EA · GW · 2 comments

Contents

  Scaling-As-A-Goal 
  Scaling-As-A-Whole
  Alright, so why don't we just do Scaling-As-A-Whole then?
  In Summary
    Scaling of a part without respect to a whole often distorts organizational coherence and scaling of a whole with respect to the parts can become qualitative growth in capacity.
None
2 comments

"This is a fundamental view of the world. It says that when you build a thing you cannot merely build that thing in isolation, but must also repair the world around it, and within it, so that the larger world at that one place becomes more coherent, and more whole; and the thing which you make takes its place in the web of nature, as you make it."

Christopher Alexander [1]

 

"Scaling" is everywhere lately. We use concepts like Blitz Scaling, read books like the High Growth Handbook[2], and hear about EA and the current funding situation [EA · GW].

As scaling becomes increasingly saturated into our everyday thinking, I want to think together, especially within EA, to consider more deeply our notions of scaling, and how we tend to implement it.

In particular, I want to differentiate between two ways of thinking about scaling that I commonly see. The first is Scaling-As-A-Goal and the second is Scaling-As-A-Whole. The terms themselves are somewhat arbitrary, and hopefully I can make them clearer by elaborating. My aim is for us to have more useful ways to differentiate our ideas.

 

Scaling-As-A-Goal 

This type of scaling is probably the most often experienced around us. It fractally expresses itself from the society level to the organizational level to the personal level (e.g. GDP should hit 3% this year, our Q2 revenue has to hit 50% growth from last year, or ‘I have to make sure I'm working out more than 3x a week’). 

For those familiar with "GoodHart's Law", you probably already see where I'm going with this. 

Often, arguably always, scaling by a particular goal leads to distortions and pathology downstream. That is, any externalities not accounted for by the goal usually decrease in quality, occasionally stay the same, and only rarely do they increase alongside the desired scaling. (e.g. GDP not accounting for human experiential wellbeing or carbon emissions leads to decreases in both).

 

Scaling-As-A-Whole

This type of scaling is what we most often wish for or intend but have a hard time communicating about.  It's often less fit for quantitative measurements, and is captured more by prose and poetry. Examples of this might sound like "we wish to be able to increase the flourishing of all humans existing now and in the future" or "I want to support my family's happiness".

Scaling-As-A-Whole requires proportional scaling across an ecological whole. This often means when Part A scales (e.g. monthly active users are increasing rapidly), Part B may "check" the growth until it's ready to correspondingly scale (e.g. our backend can't handle this much traffic without disrupting user experience). With effective feedback, the organization creates a homeostatic function for slowing down or speeding up, when either is called for.

Challenges arise, though, when the organization/organism's executive function, say the CEO, believes that User Growth is “The Metric” that will be legible to investors. The CEO thus ignores the feedback from the engineers that the rate of increase can't be handled well. This then sows the seeds for makeshift solutions that creates technical debt, and worse, reduces the capacity for future feedback to flow well. 

Crucially, sometimes the executive function needs to ignore certain feedback for the good of the whole. Consider your body suggesting "sleep is great" at the final stages of hypothermia, when the only thing preventing death might be your executive function denying that suggestion.

 

I notice we tend to interchangeably use words like "scaling" and "growth" for both notions of Scaling, and that seems to mix up our thinking in a confusing and ultimately unproductive way. For example, when scaling a community, it’s often the organizers’ intention to build a vibrant ecosystem, but by the time implementation comes around, it’s largely translated as “how many new meetups did we have this month” – usually also accompanying the loss of spirit and motivation from the organizers.

Good illustrations of the difference between these two scaling dynamics are sometimes more obvious in the biological realm than the social realm.

Using a human as an example, there's a wild difference between the rapid growth of a particular part, AKA cancer vs. the holistic growth of the human from childhood to adulthood. The former not only eats up resources disproportionate to the rest of the organism, it also diminishes the host's capacities and ultimately kills them over a few years. The latter creates an actual human, functional and creative, and usually takes a few decades. 

Finally, a bit of a caricature, but, while we might think of scaling a community by rapidly growing its membership or grow a startup by raising a few million and then hiring rapidly, we would not think about growing a tree by growing one particular branch first.
 

Alright, so why don't we just do Scaling-As-A-Whole then?

Unfortunately, we have to navigate by our maps, and the Map is not the Territory. No matter how we represent the whole, it will still be an incomplete representation – a map. Without relying on metrics, whether it’s KPIs or GDP or the number of steps I took today, we lose the ability to even monitor the parts. We need a meaningful union between the two meanings of “Scaling”.

The key difference in practice, I believe, is shifting away from:

Scaling-As-A-Goal as *against* Scaling-As-A-Whole

and instead move towards:

Scaling-As-A-Goal as *a function of* Scaling-As-A-Whole.[3]

This means that when we are scaling a particular part of a company or community, we are actively looking for information/feedback from other parts to understand its externalities. (e.g. Is the metric performing well from different points of view? Is the metric even a good representation of what we want?) It also means that when the executive function overrides feedback, it's with as much clarity as it can convey on why, and if it can not do so, err on the side of communicating it well first.

In this sense, we can perhaps come up with a lightly held, metric/representation of how well can the whole scale. That is: The capacity for parts to dialogue amongst each other within a whole is an indicator of the room for the whole to scale sustainably.

Thinking about operationalizing this well is surely its own post, as it involves how well we can even trust our own metrics. Most organization’s feedback initiatives internal or external are notoriously noisy.

However, to the extent the ideas themselves are clarifying, we can be invited to both measure how we are doing with our interlocking spreadsheets and do meditations on what the spreadsheets can never cover.
 

In Summary

Scaling of a part without respect to a whole often distorts organizational coherence and scaling of a whole with respect to the parts can become qualitative growth in capacity.

I think the opening Christopher Alexander quote is exceptionally pithy, and I will include it at the end again:

"This is a fundamental view of the world. It says that when you build a thing you cannot merely build that thing in isolation, but must also repair the world around it, and within it, so that the larger world at that one place becomes more coherent, and more whole; and the thing which you make takes its place in the web of nature, as you make it."

 

  1. ^

    A Pattern Language, Christopher Alexander

  2. ^

    Great book btw, would recommend

  3. ^

    For those familiar with Iain McGilchrist’s work, particularly The Master and His Emissary, it is the most fitting frame for what I’m trying to get at with the relationship between the two scaling dynamics.

2 comments

Comments sorted by top scores.

comment by Vaidehi Agarwalla (vaidehi_agarwalla) · 2022-05-15T02:41:16.906Z · EA(p) · GW(p)

I really love that main quote, am curious how you came across it. I'd also love to know if you would recommend any summaries or sections of the books you recommend in the footnotes. 

Some specific areas I liked:

1) I think the phrase "scaling-as-a-whole" is very useful and gets to the crux of the problem with the current situation, I also think I resonate somewhat with this sentence in particular: 

Scaling of a part without respect to a whole often distorts organizational coherence 


Although I'd quibble over what organizational coherence means exactly, I might say something more vague. (I'd be curious how you define the term) 

Here's an alternative "scaling a part without respect to the whole leads to another part being a limiting factor"

  • But this is not the full picture - e.g. if funding is abundant talent becomes the limiting factor, but there's other externalities that don't get captured by saying "talent is a limiting factor"

Another: ""scaling a part without respect to the whole leads to (negative) externalities that disrupt the balance of the existing system"

  • Not a huge fan of this one either (assumes a lot - e.g. that the current system has a good balance, that balance is desirable etc.), but I think they are getting somewhere. 

2) I like this proposed change in framing to consider the EA movement more holistically. 

The key difference in practice, I believe, is shifting away from:

Scaling-As-A-Goal as *against* Scaling-As-A-Whole

and instead move towards:

Scaling-As-A-Goal as *a function of* Scaling-As-A-Whole.[3] [EA(p) · GW(p)]

This means that when we are scaling a particular part of a company or community, we are actively looking for information/feedback from other parts to understand its externalities.

Replies from: DavidXYu
comment by DavidXYu · 2022-05-21T14:40:40.844Z · EA(p) · GW(p)

The quote is from Christopher Alexander's book A Pattern Language. It's an introduction + 200 or so "patterns" (design principles from different levels of architecture, from cities to chairs), where the reader can consult patterns that feels relevant to them. He recommends A Timeless Way of Building for more of the theory, and A Pattern Language for a more applied representation.

re 1) I have thought of it as imagining a child whose arms has chosen to grow faster by becoming an adult arm before the rest of the child body. The decoherence in the child is represented by the fact that not only is the arm putting strain on the child's body, the child likely awkwardly uses the adult arms in exceptionally sub-optimal ways to the dissatisfaction of the arm and the body. In addition, from the arm's "perspective", it's likely to believe it's altruistically helping the body by growing faster to reach higher and can feel defensive when feedback from the body is asking it to slow down. Hope this help illuminate the meaning of coherence here!

re 2) My understanding would be that this is extraordinarily difficult. Though as EA is a movement that puts good epistemics as a fundamental value, I also think we have one of the best shot at doing this well and learn from that process.

I really appreciate you engaging with the post the way you did!