Coordination Schemes Are Capital Investments

by Raemon10 min read6th Sep 202124 comments


Coordination / CooperationWorld Optimization

Second post in the Coordination Frontier sequence. Intro is here.

Once upon a time, we didn’t have money, or auctions, or markets. We didn’t have the concept of voting (let alone distinctions between first-past-the-post, or approval voting). We didn’t have banks or stock trading. We didn’t have Kickstarter.

We had to haggle over everything, which cost time. And it fundamentally limited the scale at which we could accomplish things. If a transaction cost exceeds the value of a trade, that trade can’t happen. [edit: arguably false, see discussion in comments, but fortunately not that cruxy for the rest of the post]

Once upon a time, someone had to be the first person to invent each of these concepts. The people around them were probably vaguely annoyed at having to learn a new thing. Many of the concepts required multiple building blocks and took thousands of years to reach saturation.

If you are happen to be around a lot of people around who:

  1. geek out about coordination schemes, and
  2. have an entrepreneurial bent

Then a really valuable thing you can do is explore new coordination schemes, while making an effort to distill them down into something the general population might actually be able to use.

This is hard. There are lots of failure modes. But, coordination schemes are a key thing that makes humanity powerful. Pushing the state-of-the-art forward is valuable.

With that in mind, let’s explore some examples and concepts. 

Examples: Negotiation Technology

Second Price Auctions

I remember the first time I got the results of a second-price auction. It felt like magic. Instead of arguing for hours about who got which room in a new apartment, I just wrote down my true preference for how much I was willing to pay for each room. Then I automagically got assigned a room that was cheaper than I had been willing to pay for it. (In a second price auction, each participant submits a sealed bid, and then person who bid highest pays the second-highest bid)

At previous apartments, we had just eyeballed the size of each room, and then came up with rent-allocations for each room that had vaguely round, fair-sounding numbers, and then picked rooms. It was chaotic, and didn't have a good way to account for some people valuing particular rooms for subtle, personal reasons.

Second price auction was an important, conceptual advance. But it’s not very popular in broader society. Why?

At my first second-price auction, someone had to patiently explain it to me before I trusted it. I'm not sure I actually did trust it until after everything had been resolved. Beforehand, it felt overcomplicated and I was sort of annoyed that we weren’t just eyeballing the rooms and winging it with nice round numbers. The explanation process took awhile, and I’m not 100% sure the marginal improvements in fairness/efficiency were worth it.

Thinking about how much I valued each room was actually pretty hard. “How much do I value something” is a skill that I had never really used – I never needed to. I just needed to eyeball a price and think “Worth it? or not worth it.”

Still, I think this was clearly worthwhile. Because later, I got to use second price auctions in other situations. I went in having a clear understanding of how it worked. I had developed some skill of evaluating how much things were worth to me. I got all of the magical “it just works” feeling and none of the stress. 

Other Formats

More recently, I was having a different sort of negotiation (over how much to sell someone some used exercise equipment for). They proposed a negotiation scheme intended to sidestep the haggling. 

The premise was: we each privately note down our true value for the object, then reveal simultaneously. If their value was higher than mine, we’d make the trade for the midpoint between those two values. Otherwise, there is no deal.

At the time, I didn’t quite understand the principles that were underlying this. I was somewhat stressed out about unrelated things. I wasn’t sure what my true value was, or how to figure it out. I had a vague expectation that I was supposed to do something strategic but I wasn’t sure what. (It turns out this was in fact false)

I felt slightly annoyed at the proposed system, and thought to myself “Well, I’ll just try this out and if it doesn’t work we can haggle normally afterwards.” I didn’t put much effort into figuring out my true value of the item.

I ended up choosing a number that was too high.

I tried to haggle more, and then my partner said “no, sorry the whole point was to replace the haggling process.”

So, the deal didn’t go through. 

Shortly afterwards I realized that the whole scheme depended on precommitment not to make further deals, to incentivize us to give reasonable numbers. Alas, it was too late for realizing it to matter.

I later sold the item on craigslist, costing me a bit of time I might not have needed to spend.


Once, I moved into a house with three friends. We were all veterans of second-price auctions. But, in this case, it was overdetermined who would need each room:

  • One room was huge, and it only made sense for me and my girlfriend to live there.
  • One room was teeny, and it only made sense for the poorest roommate to live there.
  • One room was medium, and obviously made sense for our fourth roommate to live there.

I’m not sure if there was an ideal, mathematical solution here. But we spent a couple hours talking about what to do. It seemed to be a zero-sum-negotiation where everything was predetermined except exactly how much rent we could extract from each other, and we didn’t know how to do that fairly. We each cared about not screwing over each other, but didn’t know how.

After an hour of arguing, I proposed “Wait, can we each just write down on a napkin how much we think each room should cost, and then reveal simultaneously? Maybe we all just secretly agree on the obvious distribution of rent and we don’t have to debate”. 

People were skeptical. But then we did that. And then it turned out, everyone wrote a higher number for their room than the other three people had written. Meaning everyone got to pay a lower price for their room than they thought was fair. (And meanwhile, our numbers for each room were indeed pretty similar to each other)

I’m not 100% sure what happened next – probably we eyeballed the numbers and then picked some round numbers that were cheaper than everyone had been expecting to pay.

But it felt magical. It depended on a lot of trust in the group. I think it might also have somehow been subtly informed by our previous experience with second-price-auctions.

Life Lessons

There’s not a single-definitive-moral here. But, possible morals include:

  • Coordination schemes are capital investments. It’s more work to use a new one, but once everyone in your ingroup has paid the upfront cost of understanding it, you get to use it over and over again.
  • Being able to use a scheme, explain a scheme, and convince people of a scheme, are separate skills.
  • The Skill-Of-Understanding-New-Coordination Schemes is also a capital investment. If you get good at understanding new systems when they’re presented to you, you can reap the efficiency and fairness benefits immediately. (Corollary: gears-level understanding of coordination schemes are particularly valuable capital investments because they help you understand other nearby concepts)
  • Coordination is cognition-constrained. It’s often bottlenecked on the least sophisticated person. If one person doesn’t get it, and needs to have it patiently explained to them, the increased efficiency might not be worth it in that instance. (But it might be worth it for future iterations, if you expect to negotiate with those people again in the future).
  • Some skills generalize across coordination schemes. In particular, the skill of “figure out how much you actually value something so you can make intelligent choices about how to bid for it” is going to come up in lots of different negotiation systems.
  • If you’re a game theory nerd, you may continuously find yourself on the coordination frontier. You may be forever having to patiently explain to people why they are leaving value on the table (for them, and/or for you). Or, sadly watching that value disappear. Alas. I recommend learning to grieve for it and accept it to some degree. But, also:
  • The skill of Explaining and Leading new coordination schemes is a very valuable capital investment. If you constantly are noticing that there are more efficient ways people could be doing things, you should probably invest in communication skills and leadership skills to actually help enact those schemes. (Related: Coordination as a Scarce Resource)
  • Sometimes, you can get most of the value just by eyeballing it and going with your gut. Shrug. Knowing when to do that is one of the relevant coordination skills. If you’re trying to learn coordination-leadership, make sure not to lose sight of this. I’ve known many a nerd who was too attached to the dream of perfect efficiency and didn’t notice it was coming at the expense of actually being more efficient in the current case.

High Level Takeaways

I think this all distills into two high level lessons, for two different target audiences:

For people vaguely annoyed at new complicated-seeming coordination plans: 

Even if a scheme seems more complex-than-it’s-worth right now, consider whether you’ll get to keep using it again in the future. See if you can get excited about that future value, and channel that into willingness to learn.

The engineers pitching complicated schemes at you may not be very good at explaining them. Try to ask questions that help you understand the underlying principles, so you can learn similar systems in the future.

Perhaps also: try to ask questions that help the engineers get better at explaining things. (Especially if the engineer is your friend or colleague that you expect to work with for awhile)

For people who love designing coordination systems:

Be aware: you are not your audience. 

The curse of many-an-engineer is to have poor insight into what their users actually want. Engineers overestimate how much complexity they can handle. If you spend a lot of time thinking about coordination theory, you may take a lot of skills and concepts for granted that others barely understand. 

Your brilliant system might easily take more time to explain than it saves in generating value (in any given instance). So, if you want to actually be adding value to the world, it either needs to be the case that…

  1. You’re actually going to use it multiple times with a given group, and therefore it’s worth the upfront cost of everyone learning it.
  2. The current use-case is a pretty big deal. i.e. if thousands of dollars are at stake, it’s worth a few hours to optimize them. If a hundred dollars are at stake, it may not be.
  3. You’re going to use similar schemes with the same group, such that teaching people the underlying principles/skills is a worthwhile investment. (In this case, it’s maybe worth more time to distill out those key skills, or reflect on what the future situations might be like so you can optimize for something more generalizable)
  4. You personally (or, the broader world) will benefit from implementing similar schemes in the future. Or, you expect to learn from the process which results in better schemes in the future, even if this particular group isn’t going to use them. In this case, recognize that you’re expecting to gain value from other people’s efforts, and it might be worth thinking of it more as a trade (where you offer them something in return for them putting extra time into learning the system)

A particular failure mode to avoid is when a system is easier for you, but not actually easier for others, and to not mistake “it’s easier for me” with “actually net valuable for people who don’t have my background knowledge.”

A particular success mode to be in is to use novel coordination schemes as a way to build up the skills of people around you and make them better off longterm. But I think this requires a particular outlook, which doesn’t come automatically.


24 comments, sorted by Highlighting new comments since Today at 7:27 PM
New Comment

Alternative takeaway: Often people don't know how much we value things. Haggling is a tool for figuring out how much we value things.

I agree, though I feel this framework helps me put into words how it is both important and difficult to actually negotiate/haggle. In particular the advice for engineers.

Are there better tools for figuring out how much you value things?

There are alternative methods for figuring out how much you value things. Marie Kondo recommends holding an object in your hands and asking "does this thing spark joy?"

I had a vague expectation that I was supposed to do something strategic but I wasn’t sure what. (It turns out this was in fact false)


Actually, unlike a standard second price auction where it's game theory optimal to just bid your fair value, there is a strategic element to this game, since the price you bid or offer will affect the clearing price.  In this case, the price at which your P(trade_occurs) * expected_utility_gain_from_trading is maximized is probably not your true fair value.  To give one simple example, if you were 100% confident their bid would be higher than your fair, you would definitely want to offer above your fair so that the mid price is higher.

Came here to say this. It doesn't even depend on knowing the other player's value with certainty -- if you shift your submitted price by $1 in your favor, you might give up a trade worth <$0.5 (if the other player's price was between your true value and the new number), and you might improve your price by $0.5 (if a trade happens). Even if you don't know anything for sure, it seems much more likely that a trade happens than the other player's price being in exactly that dollar, so it's good for you to do the price shift.

Low-level specific recommendation: Here is a really great calculator for splitting rents for different rooms. You enter in some basic info (total rent, number of rooms), and it continuously adjust room rents and asks individuals what their preferred room would be at different rent splits until it finds a rent split at which everyone would prefer different rooms. You can keep running it a few rounds past that to refine the answer more too.

I used that tool for my current rent split and it worked alright, although I didn’t understand the tool well enough in advance to know that we should do more comparisons than it automatically suggests. As a result, when it proposed a distribution of rents I was in the awkward position of wishing to trade with two of my roommates; preferring their rent-room combinations to mine. The preference was weak enough (I would have paid about ~25$ per month to trade) that I expected to lose value in arguing for further work on this (my roommates were somewhat suspicious of the tool to begin with, so making changes at this stage would have damaged trust). Overall I expect I would have gotten slightly more value from starting with the “gut and negotiate” method. However everyone left this negotiation fairly content and with increased trust, which has reaped fairly good value as well.

In that context I was the coordination pioneer, and it was helpful to leverage the reputation of the nyt when proposing the scheme. I think most people are rightly suspicious of people who propose novel solutions to current coordination problems since there may be a trick to it that leaves them open to abuse; a known reputation (yours or not) is useful for soothing that concern.

Once upon a time, we didn’t have money, or auctions, or markets. We didn’t have the concept of voting (let alone distinctions between first-past-the-post, or approval voting). We didn’t have banks or stock trading. We didn’t have Kickstarter.

We had to haggle over everything, which cost time. And it fundamentally limited the scale at which we could accomplish things. If a transaction cost exceeds the value of a trade, that trade can’t happen.

The rest of the article seems quite interesting, no doubt! That being said, the starting part (that I'm quoting) perpetuates the myth of barter that economics-type textbooks and similar push, which has little-to-no basis in historical/anthropological findings. A highly recommended text that goes further into what I'm talking about, among other topics in the text, is David Graeber's Debt: The First 5,000 Years. Take care ⌃‿⌃ !!

Ray said haggle, not barter. The concept of fixed prices for literally everyone including strangers is quite recent.

Fair enough, no worries at all! Very interesting article, thank you for sharing! I'll be honest I still need to read more haha! Take care!

Yeah I've now heard this from a few people. 

Ideally the post would have whatever the true "thing we had before money" was, and whatever true coordination issues resulted from that. Meanwhile it might actually just work to remove the second paragraph.

If one person doesn’t get it, and needs to have it patiently explained to them, the increased efficiency might not be worth it in that instance.

Corollary: if you surround yourself with a group of fellow game theory nerds, you can do more frontier exploration.  But successfully developing/explaining/using new mechanisms within this group will then be less instructive about how easy it will be to export new mechanisms beyond the group.

Hanging out with quant finance people probably also helps with this. At least some of them have clever things like this that they play as games at the office.

Yeah, and it is hard to figure out how smart people have to be exactly. I tried such an approach at our office with tasks to be distributed and used one such tool. Unfortunately, not everyone got it and that resulted in very disappointing assignments and anger as a result.

Instead of arguing for hours about who got which room in a new apartment, I just wrote down my true preference for how much I was willing to pay for each room. Then I automagically got assigned a room that was cheaper than I had been willing to pay for it.

I'm confused by how this second price auction worked. If there was just one room, I see how you'd do a second price auction to figure out who wins and at which price they get it, but how does it work when there are multiple rooms, and each person purchases exactly one room?

There's a paired optimization problem, where you assign everyone to a room, and the constraint that this assignment be 'envy-free'; that is, no one looks at someone else's assignment/rent combo and says "I'd rather have that than my setup!". There was a calculator that I can't easily find now which tried to find the centroid of the envy-free region.

There are other approaches that work differently; this one, for example, tries to split surplus evenly between the participants, and shows the comparison to other options.

Once upon a time, someone had to be the first person to invent each of these concepts. 

Well, no.  They were likely invented dozens or hundreds of times before they worked well enough for some group or use case to become interesting.  And a LOT of participants needed to trust the proposer and organizers enough to make it work. 

I agree with the overall thrust of this post, but it feels like it's really understating the challenges and headwinds.  The trust required to make it work is important here.  Second-price auctions are brilliant for one-shot transactions, and for cases where nobody is likely to use the knowledge of your true price against you.   But imagine if it wasn't among friends, but strangers, and if it were repeated every month.  Now strategy again matters.  Or imagine the landlord sees your bids/negotiations, and uses it to set the overall rent next year.  

More generally, suspicion of clever-sounding but non-common mechanisms is a protection against being tricked.  In almost all cases, the proposer knows more about the implications of the scheme than any other participant, so unless there starts out a belief that this is in good faith, there's a very large hurdle to more effective but less trivial mechanisms.

Is there anyone on LessWrong or in academia etc. who is actually researching coordination schemes at different levels (e.g. interpersonal, employee/social group cluster, organizational, city, national, international)? This seems like a neglected and important research topic if not.

I guess one of my worries with developing new co-ordination schemes is that some of these may allow people to better co-ordinate againtst the public interest, so I'm not hugely confident on the sign of developing these.

Is that also a reason to oppose every other advance?

I'm not saying I oppose making advances here, I just want us to think a bit more carefully about what kind of advances are good and bad.

Did anything in particular motivate starting this sequence?

A bunch of things, many of which are somewhat embroiled in social conflict where it's kinda annoying to spell out the details (not impossible, but requires more care and effort than I had to spare)

This post in particular was downstream of... well, basically the examples listed in the middle section.

It also is largely downstream of coordination lessons from the pandemic, which I plan to write up a post about next.