Circles with sbt-ethereum I: Understanding Circles

When I first met my friend Martin Koeppelmann in 2015, he was grappling with the problem of how to organize a UBI (Universal Basic Income) on a blockchain. I was skeptical then, and I remain skeptical. But Martin kept on, and the result is “Circles“.

Although Circles bills itself as, well, “A Basic Income on the Blockchain”, I don’t think that’s the best way to understand it. I think a better way to understand the project is as a form of self-organzing mutual insurance.

In this post, Part I, I’ll describe how I think Circles works, and why it’s best understood as a form of self-organizing mutual insurance. In Part II I’ll show how to interact directly with Circles smart contracts using sbt-ethereum to create your own Circles ERC-20 token on the xDAI network. (See here and here first for getting started on that network with sbt-ethereum.)

Note: My identity on Circles is 0x647E68F4BCBC843F39c80bb02da96dD13308f657, which I suppose via the dApp UI would correspond to:

Feel free to trust me. (Bwahahahahaha!)

Me coins

It’s technically trivial to do. Suppose you define a token you can issue at will. There’s no fixed 21 million coin limit. You can manufacture these coins out of thin air, as many as you want. Would there be any point to that?

If you tried to just sell them for cash to buy a Lambourghini, probably not. It probably wouldn’t work. But what if you treated each coin as an IOU, as in “I owe you one.” Then people who trust you — your goodwill, how reliably you live up to commitments — might accept your coins in exchange for favors.

“Help me out here, I need an oil change, you’re good with cars.” He does, so I issue him a me-coin. The one coin he now has means I owe him one. But maybe he doesn’t need so many favors from me. He needs favors from someone else though, who could use a favor from me. So he trades my indebtedness to him, the “one” I owe him in the form of a coin, to that someone else for a favor. And so on. Eventually it may come back to me — someone who receives my me-coin from a third party can redeem it to me, ask for a favor from me in exchange for my accepting back (and conceptually “burning”) my own coin.

Such a “me-coin” has a very clear basis of value. It will be at least as valuable as whatever favors I am willing to do to redeem my own coins. If I am a little shit about it, and basically don’t do anything of value when someone comes to me with a me-coin to redeem, then my me-coins might have little value. If I am always willing to do a favor worth at least $10 (which might include just paying up $10, it doesn’t have to be a chore), then each me-coin will be worth at least $10, and it can circulate as a kind of money on that basis, so long as my promises remain credible.

There will be a tension between the quantity of me-coins I have issued and the credibility of my promises. If I’m trying to deliver at least $10 of value for each me-coin redeemed — but I’ve issued a million of them, and I don’t have ten million dollars to surrender or some special talent that makes my favors with millions — then my promise will not be credible, and the value of my me-coins may decline. (“May” is not “will” — I can try to profit by issuing a lot of coins for valuable favors or purchases, hoping recipients don’t carefully track the limited credibility of my promises. Just like old-time banks sometimes “succeeded” at issuing deposits while squandering or absconding with the assets that might back them, a me-coin issuer might spend coins into existence under an expectation of value she won’t be able to support, and just stiff holders if they eventually come ‘round to redeem in quantity.)

So me-coin value is fundamentally based on trust. The me-coin issuer “owes one” to the initial recipient of the coin. Because the coins are transferable, the me-coin issuer ultimately owes as many “ones” as coins issued to the anonymous community that comes to own them. The value of the coin can be whatever, but the floor of that value will be the value of the favors the issuer is willing to do to redeem coins, even under conditions where she us called upon to redeem them all. If she can convince people to “hold a float”, to keep some of their wealth in the form of her coin rather than redeem them quickly, the practical value of the coin will be higher than that floor, the value of the favors the issuer is willing to do in exchange for redemption when redemptions trickle in at an ordinary rate, rather than happening all at once. If the issuer is trying her best to retain value, she will do a lot for occasional redeemers in hopes that will encourage others to hold their coins until they really need to call in a favor from her. In a community of me-coin issuers, if everybody trusts in one another’s goodwill, we might imagine that explicit redemptions would be rare. Alice would issue me-coins to whomever she accepted favors from, but when she does a favor back to Bob, she just accepts his me-coin rather than redeeming one of her own.

(If they ever really want to settle up, they could figure out how many of one another’s coins they have in common, and mutually redeem. Say Alice has 122 Bobcoins and Bob has 111 Alicecoins, then they have 111 in common. Bob could redeem 111 Bobcoins for favor of surrending to Alice 111 Alicecoins. Bob would be left owing Alice 11 net-favors, represented by the 11 Bobcoins she would not have surrendered. But this only “works” if both Alice and Bob are comparable with respect to the quality of favors they offer upon redemptiom, so they are willing to exchange and forego those favors on a one-to-one basis.)

Me-coins as mutual insurance

Suppose we have a group of people who trust one another a great deal. They agree in principle on a standard of value for me-coins. It need not be so explicit, but for clarity let’s say they agree that a me-coin should be redeemable for favors worth about $10 each. That needn’t mean redeemable for $10 directly — me-coin issuers aren’t intended to be full-reserve banks — but each issuer, if asked to redeem a coin, will strive to do some favor worth about that much value. Because members of the group all trust one another, and they’ve agreed on a standard of value, they accept one another’s coins on a one-to-one basis, like Alice and Bob above.

Now say also that people within this group agree that in hard times, members should be able to suspend redemption, but still issue coins tradable on a one-to-one basis for anyone else’s. Bob in hard times can still issue money, and spend it for favors from Alice, Carol, and David. But if Alice comes to Bob and asks to redeem a Bobcoin for a $10 favor, Bob can say no, he doesn’t have the time or money to redeem right now. If Alice, Bob, Carol, and David all do their best to redeem on-demand, and each face the same, uncorrelated risk, of having a spell of hard time, this is an excellent mutual insurance system. During each spell of hard time, the victim or victims can spend as usual, because the non-hard-time sufferers continue to accept their scrip for valuable favors. Bob in hard times spends a lot of Bobcoin. But he redeems none, and doesn’t receive any of his companion’s coins, since he is not doing favors in hard times. On net, he comes to owe the rest of the group lots of “ones”, for all of the unreciprocated favors. But once Bob is back on his feat and Alice is in trouble, Bob will end up doing unreciprocated favors to Alice, running down his net debt. As long as the hard-times are pretty uniformly distributed, over time it will all balance out. The arrangement is a mutual insurance scheme by which the group carries members through hard times in exchange for each member having the right to be carries through hard times herself.

What if occasionally permanent disability strikes? As long as the group agrees to cover it in advance, this scheme works for that too. Suppose that Alice, Bob, Carol, and David face a risk of losing their limbs in a horrible accident, and never being able to do a favor again. If the group agrees that come hell or high water they’ll accept one another’s coins one-for-one, then Carol and David are horribly disfigured, Carol and David will still be able to live, by spending Carolcoins and Davidcoins for favors from Alice and Bob, favors like feeding and taking care of them. Ex post this might seem unfair to Alice and Bob, but ex ante they faced the same risk of being disfigured. Just like there’s no “injustice” that ex post you’ll have paid premiums and gotten nothing back on your fire insurance if your house doesn’t burn down, there’s no “injustice” in the ones that got lucky having to care for those that did not. Had the disfigurement fallen on Alice and Bob, they’d have been taken care of instead. Ex ante, they each agreed that it was a smart choice to trade the promise of support to lose the risk of starving after an accident.

This really is a fine mutual insurance scheme. Ex ante the participants have to trust one another to live up to their commitments to accept one anothers’ coins on equal terms for valuable favors. And, for it to be fair, the participants must face approximately equal risks (of hard times or permanent disability), so that at the start it’s a good deal for everyone. It’s not perfect. Like any insurance scheme, it will fail in the face of systematic rather than idiosyncratic shocks. If Alice, Bob, Carol, and David have their hard-time spells simultaneously, or if they are all permanently disabled, no one will be able to do favors for anyone and they will all be screwed.

But there’s a vulnerability. What if Carol and David are permanently disabled, but then rather than issuing and spending “normal” amounts Carolcoin and Davidcoin in exchange for the favors they need to survive, they decide to take advantage. After all, following permanent disability, the coins they issue are costless to them. No one can redeem them for favors, they are permanently out of the favor-giving business!

Capping issuance to limit risk

We assumed, when developing “me-coins as mutual insurance”, a very high degree of trust between participants. If everybody is very trustworthy, mutually acceotable me-coins could be a pretty great mutual insurance scheme! However, we’ve also seen that if participants are not so trustworthy, there are hazards. In particular, there is the hazard that some participants might issue freely but shirk on quality or willingness to redeem their own coins. Participants like this poison the scheme. They issue scrip to get favors from the rest of the group, promising they owe the group “one” for each favor, but they don’t do (or do low quality) favors in return when called upon by other members of the group. Given less than saintly humans participating in the scheme, and the lack of some external coercive system to compel reciprocation of favors, our mutual me-coin scheme will be vulnerable to and harmed by this kind of behavior.

Our first line of defense against that is simply to choose wisely. Don’t allow into our, um, circle of mutual acceptance the kind of people who would issue IOUs for favors, then fail to make good on the IOUs, unless that’s a matter of serious need (so it’s an agreed-upon legitimate insurance benefit) and even then only in thrify quantities.

But, in real life, we don’t know our own, let alone one anothers’, hearts so perfectly, and we all change with time and circumstace. So how might we limit the risk that one of our collaborators will issue a lot of coin for favors they will never reciprocate? Well, one simple approach is to cap how much each participant can issue. We’ll still have to choose whom to trust very carefully: capping issuance won’t help us if lots of people abusively issue without performing favors for others. But if we are mostly capable of selecting trustworthy pods, capping influence will limit the damage that can be done by a small fraction of bad apples.

This is the sense in which Circles looks like a UBI: The number of me-coins you can issue is limited at any given time, but grows over time, in sync with the growth of everybody else’s limit. This means that at any point in time, there are only so many coins you can issue, or more to the point, that each other participant can issue to you, limiting the risk that a bad actor can purchase favors in overwhelming quantities without sufficient intention or capacity to reciprocate.

Self-organizing via a social graph

In practice, coordinating this kind of insurance scheme would be challenging. Someone needs to identify a group of trustworthy individuals, then get everybody in the group to agree that everybody is trustworthy and they should form a mutual insurance circle. That’s not easy.

But what if we simply Facebook style each decided who we wished to “friend”, where friending means accepting one another me-coins on a one-to-one basis? Then we could organize these circles incrementally, without a centrally coordinated big-bang. Also Facebook-style, it would have the effect of expanding the range of those with whom we could transact, as we could work transitively through our own direct friends. If Alice is friends with Bob, and Bob is friends with Carol, Alice can issue an Alicecoin in exchange for a Bobcoin, and use the Bobcoin to get a favor from Carol, who accepts Bibcoins. So, to a degree, Alice can now purchase from Carol, even though they are not friends. However, there is a limitation. Alice can purchase only up to the minimum of her and Bob’s unused issuance cap. As chains of transitive connection grow long, the exposure of two parties to one another is restricted by the weakest link, by the pair in the chain of trust that has the lowest current issuance caps.

In broad strokes, this is all pretty nifty. You can start up a mutual insurance arrangement just by pairwise enlisting some friends whom you trust, both to provide valuable favors when redeeming or accepting coins, and to choose other trusted partners well. (You are not just sleeping with your partner, as they say, you are sleeping with everyone your partner has slept with.) You get a very natural exposure (in both the good way of being able to transact for favors, and the bad way of risking delivering favors to those who unjustifiably won’t reciprocate) to the people that you’ve chosen, and — with some attenuation! — to the people they have chosen.

Still it remains challenging to ensure that the people in the social graph you are building have similar expectations. Do we agree kind of favors, or value, a “me-coin” should purchase? Do all the parties in our social graph make goods or services with that value available for purchase, except when some hardship (that members of their social graph would consider justifying) prevents them from doing so?

Explicit trust limitation

Even with the (somewhat unpredictable) attenuation over transitive connections, the necessity to really trust the people we’re willing to exchange me-coins one-to-one with might mean we’d need to be very cautious about whom we “connect” to. In order to encourage bigger social graphs and more activity, Circles incorporates a notion of partial trust. You still exchange coins with your friends on a one-to-one basis, but you can limit the fraction of your own coins (i.e. the unused cap on your own issuance) you permit your friends to trade for.

You set a trust percentage. If it is zero, basically, you do not accept a partner’s coins at all. If your trust percentage is 50%, you’d accept that partner’s coins up to 50% of your own current issuance limit. The reason that it’s important to limit how much you accept (besides not wanting to feel morally obliged to deliver favors of a very high value) is because you are obliged to exchange your own coins in the transitive transfers described above, and maybe you want to limit how much issuance a friend can compel of you. This limitation of trust further reduces the possibility that very long transitive transfers will be possible for large numbers of coins. The maximum you can send in each step becomes a fraction of the next party’s issuance cap, rather than the full cap. If limited trust is widely used, the “weakest links” in long chain transfers would likely become much weaker, helping ensure the (desirable!) attenuation of trust beyond parties to which a person has explicitly connected.


In sum, Circles defines a kind of me-coin ecosystem in which individuals can issue their own coin for favors, meaning “I owe you one”, and further, define social networks of people whose favors they accept. If most people in such a social network are trustworthy, and have similar understandings of what the value of their “circle’s” coins should be (expressed in the value of the favors they purchase or are redeemed for), such a social network might become a site of commerce, and more interestingly, mutual insurance.

There are a lot of “ifs” in all of that, and the degree to which social networks of people with consistent and reciprocal expectations will actually emerge remains, I think, quite uncertain. Also, the manner and degree to which trust attenuates over distance in the social graph remains, I think, pretty hard to predict or understand, creating a risk that trustworthy shared expectations among your friends and even their friends might be undone by “freeloaders” several further degrees of separation away.

But the ecosystem is an experiment. It remains, I think, an open question whether people will offer valuable favors for issued coins, and make good on mutual expectations to reciprocate with similarly valuable favors except in extremis. If so, it could be an interesting form of mutual aid. We’ll see.