Propinquity Cities So Far

Finding alternatives to war can save a lot of money


In any dense city, lots of people will be struggling to occupy the same set of spaces. To function, cities need to have some systematic way of resolving those positioning conflicts, a method for deciding who gets to go where. The methods we use now for resolving positioning conflicts (land markets and rent) have a lot of overhead that is both very obviously overhead and also overlooked as inevitable. I talk about that extensively, and some of its unexamined costs. I present an outline of what looks like a better method, Propinquity Optimization (proq), which resolves positioning conflicts at minimal cost, enabling a much higher maximum quality of life in dense cities.

It feels urgently needed, to me. I am not sure whether it is the most urgently needed thing that I can be working on (I’m also responsible for this humanization of recommender systems/​harmonization of global discourses and.. some other stuff). There’s some discussion of its global importance in the Longtermist Significance section.

In the course of this, I also discuss quite a lot of the problems in applied preference aggregation and some potentially novel ways to resolve some of them. Even if you aren’t interested in building better cities, you might want to read it just to see an instance of applied utilitarianism as a legal mechanism. I think that aspect of it is really pretty neat.

A Propinquity City assigns services and residents to whatever proposed locations optimize an aggregation of the preference expressions of the residents.


In dense cities, even once housing supply has exceeded demand, most city dwellers will still have rent extracted from them to a significant proportion of their income: Dense living means that demand within the urban center doesn’t ever drop. You can maybe get arbitrarily cheap housing in some incredibly sparsely populated outskirts, but not in the dense part. There seem to be levels of affordability that are firmly unreachable under the kinds of land allocation methods we use now.
When we notice that there seem to be firm and significant limits on how cheaply a technology can ever come to operate, even at its peak, smart buyers start looking for alternatives.

Price competition happens when competitors with high prices can be, in some way, outrun, by competitors with low prices. That just doesn’t happen to land traders in high demand areas. If one land trader offers drastically lower rents that others can’t match (maybe because they have entered an already high mortgage), those other land traders still sell all their units and stay in full business. They make their money, same as always. They don’t get outcompeted. Cost-efficacy, beyond a point, is not rewarded with any increase in market share. The unfit are not selected out.
The result is, reliably, the costs of operating in dense cities will always be high enough to significantly reduce mean quality of life. As long as most land within the city is privately held, this will not be solvable.

There is also a commons problem in urban economies: When rents are raised on a beloved shared service (a restaurant, a teahouse, a bookstore, anything that provides a lot of value to the locals), forcing it to raise prices and cut quality, the losses in (real) property value in the surrounding neighborhood (which now no longer has that service as it was) exceed the individual gains to the service’s owner’s owner. The city suffers more than we can know under this dynamic. A participant in this economy—even one who expected a fair chance of getting to be the land owner—would prefer that this dynamic couldn’t play out, such is the extent of the value being removed.
There would be no central park under private ownership.
We must imagine the many central parks that never were, and never can be.

You could probably solve part of this problem with a type of city where land is owned and developed by a non-profit, or local government, where land rent goes to improving quality of life in the city (public spaces, libraries, meeting rooms, schools, etc). But, why draw taxes through those wars of rent? Why take taxes in proportion to rent conflict? Is that really a good way to resolve the positioning conflict, or to take taxes? Taxes disincent things, why disincent dense living? If we were designing something explicitly to do those things well, in correct proportion, I don’t think it would do it like that.

Land pricing provides shockingly few economic functions

I wasn’t expecting to be able to come up with so many points in support of this. Here, I’ll be going through a checklist of things that useful market systems generally do, and I find that land markets generally don’t do them well, if at all:

  • Land is not efficiently priced (Land price is adjusted quickly upwards, but not downwards; there is no way to short most of it, so the price is usually not an accurate reflection of demand (page 15 of Inadequate Equilibria))

  • Increases in land price tend to be captured by people who didn’t create much of that value, and can’t really create much more of it: The value of a piece of land comes mostly from what it is next to, and the most impactful things that can be done to affect land value like making public goods like schools or parks or transit routes or even just food courts aren’t in control of the land owners who benefit. Also, please contemplate the georgist meme.
    The prospect of increases in land price incent few of the causes of increases in land price.

  • Increases in land price tend to punish a lot of the people who did contribute to creating it, by living and working here, paying rent, and patronizing the local businesses, we are rewarded by having to pay even more rent and having the businesses we have grown to love increase their prices or cut quality or close down and be replaced with something premium mediocre or just mediocre.

  • Price signals from land markets often set density in proportion to demand, but it’s not clear to me that land markets are actually particularly good at this. Would anyone argue that the city’s architects couldn’t set density levels well enough themselves? It seems to me like that’s a pretty simple problem and that urban planners are mostly already solving correctly?

    • Don’t basically all cities control density pretty tightly? I know that a lot of density restriction is just nimbies defending housing scarcity, but it can’t all be that, can it?

  • Land is inelastic. Price signals conveying increased demand for land don’t lead to the creation of more land, because it’s not possible to create more land.

    • We can imagine multi-story cities like the shimizu megacity pyramid, which do effectively create more land, but no non-governmental process has ever created one of those, nor perhaps ever will.

    • We could frame ordinary multistory buildings as an increase in supply of land, but, per the previous point, I don’t think we need land markets to help us to decide when and where taller buildings are needed, it’s not that hard to get it pretty much right.

    • We can also imagine seasteads, which could perhaps part in the middle thus “creating new land” in a quite real sense, but they will need a pretty complicated process for governing the insertion of new land into the middle of things and it’s not clear to me that this process will be distinct from the process I am going to propose.

  • You could argue that the auctioning (literal, implicit or historical) of land ensures that any given parcel will go to the person who wants it the most. That’s one way of doing that. Another way of making sure things go to the person who wants it the most is by having applicants physically fight over it until each side is brought too close to death to justify continuing and the last one standing is declared renter. We tend to agree that it’s good to avoid physically fighting over anything, because fighting imposes a great cost on its participants. Similarly, I contend that rent/​bidding is actually not all that different in that respect. As a method of apportioning resources fairly according to conflicting wills, it costs about as much as possible.
    And the ones who own the city benefit from having that process of occupancy conflict resolution being as expensive as possible, but the people who live and work in them mostly don’t. They would like to go somewhere where it’s handled differently.

In summary, land markets are not very good at what they do. They provide less functionality than we might have imagined.

To completely replicate their functionality, we will need lots of information about the housing stock, and some (perhaps democratic) negotiation tool for deciding who and what gets to go where. Since those things would be useful to have anyway, that’s what I’m trying to develop here. That’s what proq would be. Tools for pooling information and negotiating with minimal overhead so that we do not need to burn so much money fighting each other for space.

What it is

Residents (who have bought shares, funding their part of the construction, who pay rates) in a Propinquity City provide the city with a pretty complete expression of their needs and preferences about their housing and their neighborhood. The city defines a mathematical function that represents how well those expressed desires are being satisfied, given everyone’s locations. Solvers try to find ways of positioning residents and services that will make that number go as high as possible. Whichever location solution resolves with the highest number, is instated.

More Concisely: Every month, a Propinquity City positions services and residents according to whatever proposed location solution optimizes an aggregation of the expressed preferences of the residents.

Residents end up in the presence of the people they want and are wanted by. Services are allocated space according to expressed public will for them, rather than how much rent they can pay. We solve the occupancy conflict problem via cheap, efficient, negotiation towards an optimization criterion, instead of through a bedlam of costly bidding wars.

The metric (or, the expression language) representing an individual resident’s desires, focuses mainly on these features:

  • Adjacency desires: How near the resident wishes to be to specific people, types of people, services or types of services

    • The system only recognizes an adjacency desire to the extent that it is reciprocated by an adjacency desire of the other party. This limits nonconsensual interaction.

      • Consider, for instance, a celebrity, wanted by millions. The ones who want them the most are often not the ones they would want to live beside. The ones they want in return would be lost in the crowd. This prevents that.

      • Services generally automatically reciprocate resident adjacency desires (which for now would be implemented with a default maximum desire to be adjacent to anyone (or their preferred resident type))

  • Requirements about their housing, things like “must be on the ground floor” or “must have blackout curtains” or “must face east”.
    To support this, it would be a good idea to try to develop an open process for adding new qualities (and dropping disused ones?) to the checklist, that the city couldn’t anticipate. I really don’t like the idea of relying on a single bureaucracy to decide which qualities of a piece of housing might be worth keeping rows for.

    • This system should probably play a role in measuring needs to decide what sorts of new housing is built.

  • Whether they’re willing to be moved, and how important it is to them that they not be. Generally, the aggregation gets a bonus for not moving a person (unless they communicate that they want to be moved). Having to move is annoying and it should only be done when it would raise utility further than the threshold.

    • It may be possible to plan cities in such a way that moving wouldn’t be annoying at all. See Eliezer’s Movable Housing for Scalable Cities. Look at how nice Kasitas were going to be.
      I think proq housing should at least consistently make apartment doors wide enough for a forklift to drive through.
      I’m concerned that going for fully modular relocatable housing would make this concept much harder to realize, but hm, (thinks of Elon Musk) maybe sometimes, dreaming freely and going after unprecedented things increases the probability of success, so long as you dream with an engineer’s clarity.

    • There are lots of questions I have about this that I think we will almost certainly find answers to when we run the test games

      • Whether even very small movement penalties are totally good enough to keep proq from moving people whenever reasonable

      • Whether we can just promise not to move a resident unless their utility would be raised by it. To me, this seems likely to result in commons problems that would lock the city dead, but who knows.

      • Whether we can simply trust residents to just honestly report how annoying it would be for them to be moved, or whether we have to restrict that input to ensure that the configuration will be able to improve when it should.

    • Shouldn’t people have a right to stand still?
      It’s not clear that they should. Consider how we don’t give people a right to stand still in a road. Living, dynamic communities are at least a little bit like traffic. They want to shift as communities shift, as new people and services come into existence, as movements split and scenes evolve.
      You should usually be able to stand still most of the time, but we can’t make a firm guarantee right now. Maybe the test games will reveal that a universal right to stand still wouldn’t cause the city to clog at all, but I definitely wouldn’t bet on it.

  • I suppose the system will need to recognize legal restraining orders and not violate them.

    • Unsure how to implement in a utilitarian frame. Restraining orders are typically between a victim and a perpetrator of some heinous act. Restraining orders should not harm the victim; the burden of creating distance should be placed mostly on the perpetrator, but I’m not sure how to get the math to work in proq so that this rigidity wont mildly punish victim and perpetrator equally as often. An egalitarian altruistic city cares just as much about the needs of the perpetrator as the needs of the victims...

      • It’s conceivable that there wont be much of a burden in the creation of distance; that the next best neighborhood for each person will be about as good as wherever they were. Testing needed.
        If the act was bad enough, whatever second degree social bonds were still holding these two in adjacency may sever and create a distance without any intervention from the solver, in which case it really wouldn’t be a big deal.

    • Reciprocal desires to be separate (in response to, say, a breakup) would be supported without any special legal order.

      • There would be something beautiful about having an incentive to revisit your distancing orders against your exes every couple of months and maybe declaring that you are at peace with them now, and finding, one cycle, that they reciprocate. (Oh. This might need an additional little UI/​system to be good. Otherwise they’d have to like explicitly arrange forgiveness by talking to each other which most people wont do, or decide unilaterally, which is a mess.)

  • Etcetera. The preference expression format will be about as broad and varied as the needs of people themselves.

Then those measures of the quality of each resident’s situation are aggregated in some way (added together, for instance).

That aggregation of the preferences of the residents is the metric that the propinquity city is legally obligated to optimize.

I’m not exactly sure what operation should be used for aggregation. It may need to vary between different proq cities, depending on their population’s preferred variant of utilitarianism. Candidates include:

  • Simple addition. Traditional and decisive.

  • Addition of the sqrt of each individual utility (or some other function that has diminishing returns; logarithms, raising to the negative power), which will make the system try to help the people who have the least before it gives more to people who have a lot. That isn’t my metaethics, but it’s practical. It’s there to reduce the number of people who end up feeling cheated out of what they were promised, or to allow the involvement of people who might have otherwise anticipated that they would be made sacrifice to the greater good due to some personal predisposition (I don’t think I can currently imagine what sorts of characteristics would make an IRL utilitarian more inclined to sacrifice a person’s wellbeing for others, but a picture would probably surface over time and people would react to that)

    • On the other hand, some parts of the individual utility function would already have decreasing returns (in the same way that money only buys decreasing marginal happiness, your 10th neighborhood friend is not half as important to you as your first). So I’m not sure this would really be needed.

  • The min of the utility functions: The score of the solution is the lowest score of any resident, it is all we should care about.

    • The firm anti-Omelas.

    • I’m going to be straight up: I don’t think this is the correct aggregation function. The system should care, at least a little bit, about opportunities to make the vast majority of the populations’ lives better, even if those interventions will not help some people.

  • Is there a function with an additional parameter that would naturally lerp between those? Idk. Just curious.

    • Maybe a weighted sum of the mean of the poorest i residents for all i in n

      • I’m starting to think this might be reaching for a kind of mathematical elegance that the moral principle of equality isn’t going to turn out to have, though. There are a number of reasons equality reliably emerges as a value. Again, one is just pragmatism; if people can tell that they’re guaranteed to receive a bad deal, they wont buy an apartment and the city will not get to welcome them in. Another purpose is keeping truces, by keeping power ratios between factions entrained with their sizes. Those are completely different sorts of objectives, and these aggregation rules don’t much resemble either of them.

        • Huh. What if those things are implementable though. What if you could solve utilitarianism’s sacrifice problem.

          • A crude solution to the predictably sacraficial deal problem would be to assess applicants’ other options and make sure they’re guaranteed utility greater than those (and maybe factor that into the price of their share) so that even if the optimizer would have otherwise disfavored them in some way it’s still going to be worth buying in.

          • Solutions to the inter-faction instability problem could probably be cobbled together from measures of kinds of access to capital and then having the optimizer try to keep those fixed… but… negotiating an agreeable design for this would be uh, challenging.

Proq resembles utilitarianism, but utilitarianism couldn’t really be implemented, even if the political will were there (or if there were some rawlsian veil that evened out the expected payoffs and guaranteed that the deal would be worth it for everyone). We can’t read peoples’ actual utility functions. We can ask them to describe their utility function for us, but they will not generally answer with the true utility function, we will receive a speech act that has been carefully, strategically shaped to benefit them more than the truth would have.

Anything implementable is always going to be more like a voting system than a metaphysical ideal.

To argue for any voting system, we need to be able to argue that the dishonest individually rational voting strategies that people will inevitably discover and deploy will tend to add up to acceptable outcomes. If we can get people to tell the truth about their preferences, we can just measure the solution in terms of those, but it is difficult to incent people to tell the truth. In many cases, it’s provably impossible.

I don’t know what strategic voting would look like under proq, I don’t think we’ll know until the system exists and we can play around with it and see which strategies thrive, but I know that there will be some analogue to strategic voting, there always is.

I’m thinking about making a game version of life in a proq city and getting adversarial economist types to all try to “win” at it. (though it’s important to emphasize here that in a eusocial game, winning doesn’t tend to look like domination. It will tend to look like trading beneficially with others as a side effect of pursuing whatever your goal is. One of my projects in game design is addressing the alarming ubiquity of contrived zero sum multiplayer games. Every time I read the rules of an otherwise peaceful eurogame and wind up meeting again the phrase “whoever gets the most victory points Wins The Game” I groan a little louder. Soon you will be able to hear my groan from the mainland. One person’s gain should not be presumed to be another’s loss. Life isn’t like that. Humans aren’t like that. Long ago we entered a pact that bound our fates together.)

There, in those games, we’ll get a glimpse of this political ecology’s future, and we’ll see if the system continues optimizing utility under strain.

I thought I’d need a pretty decent prototype propinquity optimizer algorithm for that, but I’m starting to think it might be a lot easier, and maybe a lot more fun to do a thing where every resident is able to submit hand-authored incremental improvements to the position solution.

In the long term, offering a prize to whoever can optimize the allocation solution’s aggregate utility the most might elicit near optimal solutions from specialists in location solving, who I’d anticipate would make use of some fancy algorithms, but it’s conceivable that a series of incremental improvements from individuals and volunteers might turn out to do well enough in the beginning, as well as fostering enfranchisement.

But anyway, in the least, whatever process optimizes the aggregate utility, it does not have to be a ministry of the city. The great thing about defining an objective, easily computable measure of solution quality is that it means we can cheaply consider allocation suggestions from whoever will offer one. If their solution scores the best, then it pretty much must be the best and that is the one we will pay for.

So, one of my current considerations is this: How much could individual people use their understanding of their propinquity locality to incrementally improve the solution for themselves until we arrive at a solution that will be pretty close to ideal. I don’t know. But I’d like to experiment. Playing propinquity optimizer seems fun to me. Making an app for editing location solutions and measuring their total propinquity also sounds like a great starting point for designing location optimization algorithms, if those later turn out to be necessary.

If we do let residents submit incremental manual edits, a naive implementation would have some difficulties

  • What if multiple edits are coming in every minute but it takes the average person over a minute to compose, consider, and upload an edit. How do they get through.

    • I guess there’d need to be some fairly sophisticated operation that tries to apply edits that might be slightly out of date. It would ignore what they say about residents who are no longer where the patch thought they were, for instance. Sometimes it would have to report that the difference is too great and the edit is no longer applicable, or that some change in another part of the solution made the edit a negative change and so it can’t legally be applied. Hopefully this would not happen too often.

  • quibbles, probably not important

    • Possible Exploit: A strategy where a solver superpower might hold back their best solution and permute the solution slowly but steadily upwards around drastic shifts to prevent those with local knowledge from being able to get their edit through, to ensure that they will not lose the prize to some rando who uses local knowledge to improve on their best solution.
      Unlikely, as this would be both hard to make and pretty evil. It’s just unlikely that anyone would want to do the work. More likely, if local knowledge ever took the lead from a professional solver, they would lobby the city and we’ll figure out what to do with that later.

    • Would there be a possibility of edit wars? … Generally.. no, actually! Edit wars seem to be impossible, or in another sense, desirable. If edit A is legal, it must be increasing the utility (could ban edits that have zero effect on the utility to ensure this). The edit that negates A then, harms the solution, and would not be allowed to go through. To effectively reverse A, the adversary would have to find some way to at least mildly benefit the solution in some other respect to bring it higher than it previously had been, which would be a positive externality and should be allowed.

    • Should residents be allowed to change their adjacency preferences during the incremental edit period?

Longtermist Significance

I found proq by following the anguished cries of the present. In these cases it’s good to step back and remember the endgame and ask if it still makes sense in light of that.

I find that I have more questions than answers.

A Propinquity City would support extents of quality of life that I’ve argued aren’t possible under the current paradigm. It would be nice to have.

Ultimately, though, dense cities will not be as important in the near future, given remote work (which I expect to be irresistable once VR headsets with foveated rendering reach retinal pixel densities) and dirt cheap automated delivery systems. Propinquity is good adjacency, but adjacency wont matter as much.
Proq arose from a concern that a dense city could not ever be made cheap. I do still believe that, but I’m not sure we sorely need dense cities to be cheap. Might we live almost just as well in sparser, broader land markets where not all units sell, where it is theoretically possible for land prices to compete down to negligibility, where there aren’t far more buyers than sellers.

Proq offers us a future with at least one dense yet affordable megacity in the western world, a lively intersecting patchwork of emergent communities growing somewhere in the middle of the continent. The future without proq still offers us tesla-quickened land markets, expensive in urban centers but perhaps decent work will be available from any small town grouphouse with an internet connection. We may want to scatter broadly if we want to live on a non-profit’s wages, but we will be able to live well enough. It’s conceivable that this difference in living situations will have some predictable effects on cultural evolution. Anthropological forecasts on this would be deeply interesting.

Proq will invite anyone who knows about it to contemplate legal systems that constitute from the optimization of a utility function. I wonder if experiencing the results of that might make the alignment problem more broadly obvious. For better, or for worse?

I wish we had a clearer picture.

Practical Concerns of Deploying it in Reality

I really hope we wouldn’t need to convert any pre-existing city that already has high land prices. That just looks like an impenetrable, unscalable political wall to me. I am not planning for that.

We might have to kind of start from scratch. This isn’t necessarily as depressing as it sounds. There are places in the world where construction progresses very quickly. Perhaps one day those businesses processes will make it to the west.

If it does take a while for a city to grow, oh well. Personally, I think I would love living in a tiny fetal city. Maybe it could be like arcosanti. I can dream, at least.

Acquiring land

I would like if we could make this deal with the regional council: Once the city needs land, it could buy it at triple the price that rural land would have been expected to command had the city never been built.

This prevents land-owners from holding the city’s growth hostage with their newfound land wealth that the city, by being adjacent to them, created. It is just. It is profitable for the rural land-owners, their land-value still goes up significantly, and they still benefit from adjacency to the city in other ways. A rational citizenry, knowing that this city could not grow otherwise, would accept the deal.

In Case no Rational Citizenries can be Found

It’s conceivable that there are places where the the land owned by the city on its outskirts could be scaled up faster than the city grows, meaning that by the time the region’s landowners believe there’s a thing of value here to exploit, when the city does start to press up against the edges of its domain, it would have enough residents to vote for fair prices for further expansion.

Seeding the City’s Economy

A delightful puzzle. Finding a series of productive yet crazy organizations, each wanting to be near some of the ones before them, progressively becoming less and less crazy, until reasonable people start to get it. I can see some pieces of the solution; first businesses that don’t mind solitude, then businesses that can operate with just an internet connection, and by then we will have more pieces to work with that I can’t anticipate right now from here.

What this needs from you

We need to get generally better at designing less costly ways to credibly signal will: Develop voting theory, maybe develop some auction theory for cases where cost-free outcomes are not possible, but where very low-cost outcomes might be possible:

  • An aside: we (Colton Dillion and I, and some passing contributions from others who were around) have been exploring a few low-cost auction designs, the general theme is that a cost is imposed for bidding and losing, so money is only burned when participants in the auction can’t cheaply reach agreement about an advanced prediction of who was going to win. These sorts of auctions would make the most sense for dividing assets that have no owner; unclaimed territories, radio bandwidth, perhaps, occasionally, network traffic or road space? (There is only a tenuous sense in which urban land could be included in this category, I can’t yet see how to apply these systems well to negotiating out of the wars of rent, but I’m going to keep looking.)

    • An auction where all bids made have to get paid to the house, even losing bids. Brutal but simple.

    • Regular auction, but each bid has a bidding fee. Failed bids are, thus, discouraged. Recognising and ceding to the strength of will of your opponents without bidding is thus incented.

    • Colton proposed this (and did some analysis): Everyone stakes an amount of money that is supposed to be proportionate to their will. This remains fixed. The stakes are revealed, and there is a withdrawal period where people can leave the pot at no penalty (surrender). For those who remain, the war begins. Each dollar from the largest bid is matched at random to a dollar from the remaining contenders and burned, steadily over time until one bidder is left remaining via sortition. The owner of the last remaining bid wins.

      • I like this mechanism. It lets you call peoples’ bluffs by staying in and warring with them. It lets costly evidence of will increase gradually, only as much as is necessary for the less willing to be convinced to step aside.

    • I propose a continuous version of the above, where essentially the dollar unit of matching approaches the limit zero, which makes it non-random. Although this would not make auction outcomes predictable, as we cannot know when other bidders will pull out, it makes it more predictable, which should increase players inclination to surrender.

Expressions of interest in buying a share—a permanent entitlement to a propinquitously located apartment in a Propinquity City, cost of about 40,000USD plus a yearly rates fee (covering maintenance and governance) - would be pleasant to receive, though not actionable at this stage.

I may develop a small game for examining and maybe demonstrating propinquity optimization. If anyone else would be interested in developing a game about propinquity optimization, I’d contribute heavily. I really do feel like there’s a fun game to be found in there.
Designing a proq game with a score criterion that accurately reflects of a propinquity city’s optimization criterion would be a really cool challenge. There is a chance, small, that it would help billions of people by helping to speed proq into reality. So for a game designer, it’s very much worth thinking about.

  • I was delighted to learn that there is a game, Islanders, about a type of propinquity optimization. I find it pretty inspiring and it has renewed my energy.

If anything here seems insufficiently well justified, or questionable, I encourage you to please ask about it. Chances are decent that I will have thought pretty deeply about it and I will have lots to say about why it was unavoidable, and I just wasn’t able to fit it in anywhere.