Things tend to get worse as they scale. As startups turn to large corporations, they empty of meaning. The tight social bonds of a village fail in the big city. What goes wrong?

Common theories are about coordination failure, trust breaking down, lack of reputation enforcement, etc.

Here, I'll explain these dystopian aspects of scale in a new way. I believe we systematically measure the wrong things at large scales, and we get what we measure.

The systems that need to scale up can be broken into parts—parts I'll call funnels, tubes, and spaces. As systems scale, we fail to monitor the health of one part: the spaces. This leads to space decay.

The dystopian aspects of scale are well-explained by space decay. This means, by measuring space health, we could make big things less dystopian. That's the good news.

I'll discuss how to measure space health. I'll also look into why we don't measure it now. The upshot is that we don't realize how important spaces are. We measure what we respect, and we don't respect spaces enough. Luckily, there's a great deal to be gained, even in our personal lives, by respecting spaces more, and become more articulate about their success.

Funnels, Tubes, and Spaces

Almost anything can be classified as space, funnel, or tube—personal relationships, products, social activities, or even physical places like offices and living rooms.

Funnels are when you're part of a goal someone else invented. Note: It's a funnel even if you like the goal! So: when you find yourself pulled in by the smell of french fries outside McDonald's—that's a funnel. If you work for an org that gets malaria bed nets to as many people as possible, that's a funnel too. Were you part of a cult that tried to turn you into Elon Musk? Funnel!

What ties these things together is that someone (not you) designed something to push people towards a goal or vision.

Tubes are things which accelerate each person towards their own goal. When you look something up on Google Maps, catch a taxi, or negotiate a win-win contract, that's a tube. Online examples are Google search, Amazon, and various APIs.

Tubes get you to your own goal quickly.

Spaces aren't about getting a clear goal accomplished. A space is a place for non-goal directed, exploratory activity. Like jazz jam sessions, art studios, or conversations with friends. At larger scale, research labs, festivals, and universities count as spaces. Online games like Roblox and Minecraft are spaces.

What do I mean by non-goal directed activity? In spaces, instead of looking for actions that advance a goal, we follow paths of attention that lead somewhere interesting. In a conversation with friends, you might attend to what you could reveal about yourself. At a jazz jam, you might attend to dynamics contrasts. While dancing, you might explore being slightly off balance. You combine these paths of attention like a painter combines colors, to uncover possibilities.

A good sign something's a space is if you don't want it over quickly.

The Effects of Space Decay

Spaces are great. It'd be a damn shame, if something bad happened to them.

Here's some (admittedly indirect) evidence that something bad has happened to spaces.

If we imagine a society without spaces, only tubes and funnels, such a society would have problems. Are these problems familiar? Are our spaces are disappearing?

1

Ambitious goals—like curing all diseases or mapping the brain—require exploratory spaces. So, a society without spaces would be incapable of ambition.

More broadly, institutions like science and democracy would break down. Their funnels (for winning political campaigns, riling up voters, ramping citation counts, etc) would hypertrophy, taking up a ton of our attention, while the relevant spaces (where scientists and citizens explore values like civic responsibility, epistemic humility, and the passionate pursuit of the truth) would atrophy. We'd have degenerate, funnel-only versions of democracy and science. These would be terrible places: without values to keep them in check, perverse incentives would compound.

2

Social connections would unravel without spaces. Funnels and tubes tend towards relationships that are transactional, rather than deep or exploratory, because people want to achieve their goals quickly and without risk. Transactions (short-lived, predictable, and simple) are the most minimal kind of relationship, where both parties stay atomized, and there's no possibility of surprise. [fnTransaction]

Those limited, transactional relationships aren't enough. So a society without spaces would have less trust, less social cohesion. There would be spikes in drug addiction, suicide, etc.

3

There's another consequence of people's desire to achieve goals quickly and without risk: the elimination of surprise. Funnels and tubes emphasize efficacy and predictability, and these take precedence over other values like creativity, boldness, vulnerability, embodiment, etc.

We use a blanket word for these values: meaning. In a good space, we explore what's meaningful to us—not what's efficacious. Spaces are where you can live expressively, and treat yourself as a source of surprise. fnMeaning

So, without spaces, there'd be a "meaning crisis". People would grab at sources of promised meaning, like radical politics and get rich quick schemes, and they'd try to fill the holes, but that wouldn't work out.

4

Finally, a life without spaces would be exhausting. Every encounter would be about getting something done and moving on to the next thing to get done. Sometimes it's your own goals; sometimes, other people's. Completing one checkbox just brings you to the next, never to a space for relaxation or open-ended exploration.

People would race around, looking for spaces, but not finding them. This would make them exploitable: funnels would dangle the vague promise of a space in front of them: "buy this beer and be loved by friends", "take this online course, get rich, then you'll be able to relax and explore". But these would never deliver, because even the drunk, rich people would lack spaces. It'd all be a tremendous waste.

Measurement Error

In the rest of this post, I'll claim that space decay is happening, and that it's mostly due to measurement error, and I'll show a way to address it.

I'll begin by looking at a space I was in recently, and show why might it decay.

With my girlfriend, I went to the farmers market for wild mushrooms. We found an array of mushrooms there we’d never tried, and asked the seller about them. A 15 minute conversation ensued: about mushrooms, seasonality, local forests, taste profiles, and how the weather this year is different from last. Recipes and forest locations were shared in both directions. We were all energized.

While we talked, the seller didn't focus on completing the sale, and neither did we. She might even have lost business by having a long conversation. At the end, we did purchase some mushrooms, and the vendor bagged them up.

In that story, the market's a tube, because it routes visitors to different stalls based on their goals^2. The conversation's a space where we explored our mutual interests in culinary mycology and nearby forests; and the purchase is a funnel.

Now,

Imagine you administrate the market. You want to gauge its success, so you can improve things. You’ll have clear evidence of tube success: how many people buy things? There's also clear evidence of the vendor's funnel success: her purchase and customer counts.

But space success is harder to measure. An interviewer or journalist could see it: in the energized feelings of the conversations; in whether the vendor enjoys her job. An anthropologist could discover out growing list of shared recipes.

But if you ran 1,000 farmer's markets, you couldn't visit them all, learn about the recipes, or see the energized feelings. You'd only have records of funnel and tube success.

Your trouble with measuring the success of your markets has big consequences:

• Your funders will need metrics, to justify funding. Whether they're governments, foundations, or investors, they have an obligation to allocate resources accountably and well. If space success cannot be documented, they cannot responsibly allocate resources to spaces, and those resources will go towards funnel and tube success instead.
• You won't be able to recognize which markets to learn from. Technical knowledge and supportive tools evolve only when success is recognized. Successful funnel- and tube-makers are recognized, best practices are extracted, and playbooks made. But, building and scaling spaces remains a shot in the dark.
• You'll struggle to attract distant participants. 'Top 10' and 'trending' lists—and even older discovery mechanisms, like "word of mouth"—work less well when successes are hard to talk about.

If you're a thoughtful administrator, you might notice this problem and try to measure the space aspects of your markets. Let's consider some ways to do that:

You put cameras all over the farmers market, connected to a machine learning system which detects smiles and laughter, and you attach funding to the smile count.

Uh oh. Now the vendors have an additional goal: to drive smiles by any means necessary. They develop all sorts of tricks, to get people to smile.

This would be a mistake because of Goodhart's Law. In general, putting a behavioral metric (like measuring smiles) on a space turns it into a funnel.

What else could you try?

You survey market-goers about the experience, and give each market and each vendor a customer satisfaction score.

Surprise! This doesn't lead to spaces either!

Again, we have Goodhart's Law. Reputation pressure (like on Uber) tends not to drive, spontaneity and deep relationality: when someone's reputation is on the line, they play it safe, and behave according to scripts and norms as much as possible. So, satisfaction scores drive polite, servile, harmless interactions—the interactions which are unlikely to get a negative score. The space turns into a funnel about getting through with an ok score.

Articulate Signals of Space Success

Perhaps I've convinced you that measuring the success of spaces is hard. But I haven't really got to the heart of the matter.

To do so, I must introduce another distinction—between articulate and inarticulate signals of success.

• Inarticulate signals. With smiles, purchases, and customer satisfaction scores, the customer doesn't have a chance to articulate reasons for something to be considered a success. They either smile or do not, purchase or decline to, give a five star rating or a negative one.
• Articulate signals. In contrast, when people recommend restaurants or chefs, they often provide lots of info about why a particular chef or dish was exciting for them. Success in cooking is driven by an ecosystem of articulate signals: magazines, specialized language, detailed reviews, etc.

To measure space-success well, I believe you need articulate signals. I'll give an example of the kind of signal I mean, then say why I think articulate signals are key.

So, put yourself back in the shoes of the market administrator.

You run surveys and interviews and many markets, collect meaningful experiences, and crystalize them in to cards like this:

Then, in each city, you ask the locals which cards they value, and where they experience them most, and see how your markets rank.

I think these cards are promising. But we can go further.

Instead of asking where locals have an experience, you can ask where that experience is easiest or best supported, and get into the details of what it means to support a value well.

You break down the "hard steps" of each card above: things that need to be possible in the environment for people to live by that value if they want to.

You then ask the locals where those hard steps are most possible, or best supported.

I see three reasons why articulate signals do better on Goodhart's Law.

First, articulacy gets you closer to what matters. Any measurement involves an inferential distance, between what's recorded and what it's supposed to mean. Articulate signals get closer to what matters, and that leaves less room for divergence between the proxy measure and the meaning. We see this in the example above: since the card text is strongly suggests a space, and the particular type of space the local wants, they would be unlikely to misidentify a funnel or tube as a good place for the cards.

Secondly, articulate signals create polycultures, not monocultures. Think what food would be like, if customers could never express articulate signals about food? Let's imagine if cooking used an inarticulate measure like smiles or satisfaction scores:

An app called "Meals" is on every phone. It shows all meals in the world, ranked by the number of people who've eaten that meal. For 13 years, the Big Mac tops the chart with the Whopper close behind. Thanks to this global ranking of meals, it is finally clear what people want from food. The people voted, with their wallets and their time, and showed us which meals matter most. We realize now that earlier trends like arepas, ceviche, farm-to-table, and molecular gastronomy were just mistakes due to lack of data.

Monocultures always feature inarticulate signals. Articulacy is what separates indie film from Hollywood blockbusters (where success is measured by box office gross), and what separates gastronomy from fast food ("meals served").

The reason is because, with articulate measures, there are many ways to succeed. Polycultures emerge where different suppliers specialize in different things, rather than collapsing success into a battle for one rating, which takes place in the crudest vocabulary.

In the values cards example above, different vendors could specialize in different values. This removes a conformist pressure we see with inarticulate measures, and creates a polyculture.

Finally, articulacy protects against social performance. Social norms can be engineered to get people to do all sorts of things—for instance, to smile forcedly in a customer interaction, or to leave a positive review in an app. An articulate measure is not proof against this: a discussion about mushroom recipes can be driven by social norms, and we can easily imagine a vendor feeling pressured to have a "Land-to-Table Living" discussion with each customer.

But articulacy does provide some protection. First, it's harder to fake a review if you must describe advantages in detail, and relate them to your own life and values. More importantly, articulate measures can get us from the realm of social actions (which can always be forced) to more private assessments (which are harder to fake). In the "hard steps" example above, I suggest asking locals where it's easiest to "discover people with shared interests" or "build relationships". It's hard to imagine social pressure leading customers to get these assessments wrong.fnCapabilityApproach

Conclusion

The main thing that's required in all this is articulacy about spaces—about the different things people get out of them, and about what makes them work well. This makes me think what's necessary isn't just new measurement techniques, but a community—a community that talks about spaces like chefs talk about food, or like funnel-making entrepreneurs talk about funnels.

People in such a community would want to gather articulate signals and study them. They'd want to allocate funds and publicity based on those signals.

Such a community would be of interest to farmer's market administrators, of course, and to people making all sorts of exploratory spaces: research labs, art collaborations, games, "tools for thought", etc.

And—at the risk of sounding hokey—I think that community would also be of interest to another group. Lovers. Many problems in relationships (bad bosses, bad dates, bad lovers, and bad friends) come from treating time together as a funnel (towards sex, towards marriage) or tube ("I care for the kids; you make the money we need"), rather than a space.

So let's form a community of space-makers—for everyone who wants to sit with friends, find something that's meaningful to them, and make space to explore it together. And then make spaces at larger scales.

This book is dedicated to all space-makers.

🗄️

To add