Dear Blockchain, You Were Not the First Decentralized Network.
What we can learn about off-chain governance from the Medieval Law Merchant.
The Role of Institutions in the Revival of Trade: The Law Merchant, Private Judges, and the Champagne Fairs, was written by Paul Milgrom, Douglas North, and Barry Weingast. Published in 1990, the paper outlines the Lex Mercatoria, or the Law Merchant, which were the laws governing commercial transactions between merchants during the Middle Ages. Their research shows the cost-benefit relationship of reputation systems for incentivizing honest behavior. Using the prisoners dilemma, along with other game-theoretic systems, Milgrom and his colleagues explain how the Law Merchant succeeded for such a long time as a completely decentralized network.
The article proposes that traders must be informed of their trading partner’s past behavior to serve as an adequate bond of trust. Well informed merchants can boycott those who have cheated in the past. So long as the cost of attaining and communicating that information does not exceed the cost of engagement, the Law Merchant can thrive.
While this paper has been discussed by others, briefly, for its applicability in blockchain, here I will attempt to outline how its game-theoretic systems prove germane to off-chain governance. Off-chain governance is an important, and often overlooked, sector of the decentralized network. How do we make judgments on those who break the rules? Those who cheat or lie? Or steal from honest players on the network? All while maintaining the decentralization that is so critical to the culture of blockchain technologies.
What is the Law Merchant?
Nation-states, as a centralized entity, have been around for a really long time — but not forever. During 11th century Europe, most of the continent was a feudal society of locally governed townships that were responsible for their group, and their group alone. As time went on, and geographic trade specialization increased, Medieval Merchants found themselves without a system of governance that could incentivize honest behavior among one another, specifically when trading outside their own township. The Law Merchant evolved from a need to mitigate new types of cheating afforded to a new, inter-communal system of trade. While the feudal government could enforce agreements between two members within their jurisdiction, they had no control over visiting merchants from other towns. As a result, the system developed to deter merchants who would cheat in one location, never to be seen again. The Law Merchant was a means of reducing uncertainty associated with trading outside ones typical jurisdiction.
The Law Merchant’s reputation system is supplemented by a private judge. The private judge served as a central authority of who had cheated the system, and who had not. He promoted private resolution and transmitted just enough information for the community to maintain reputation equilibria. The services a judge provided were bundled together to benefit both the community, and the individual seeking adjudication. A trader seeking a ruling against a former partner would benefit himself by receiving a judgment, and benefit the community by identifying a dishonest actor in the network.
“Laws are a dead letter without courts to expound and define their true meaning and operation.”
The Law Merchant sought to solve a multi-fold incentive problem in the trading community. The main focus are the following:
- Induce members of the community to behave honestly;
- Those who have behaved dishonestly must be boycotted by the community;
- Members must keep informed on who has been dishonest;
- Members must provide evidence against those who have cheated; and,
- Rulings of the judge must be recognized as legitimate in the community.
The subsequent sections examining the Law Merchant system and it’s game-theoretic counterparts are not my original thoughts, but rather summaries and reiterations from Milgrom et al.
The Prisoners Dilemma (PD) is a famous cooperation game that depicts incentive dynamics between two actors’ interactions with each other. In the case of the Law Merchant, the two potential choices are: Honest and Cheat. Let α > 1 and α — β < 2.
Honest behavior maximizes both traders’ utilities (1). But, if a trader cheats when his partner is honest his individual payoff is higher (α > 1). Assuming the traders only transact with each other once, then his maximized individual utility will be to cheat if his partner is honest. But, if they both cheat, both traders are much worse off than if they played honest (0).
Now, if the relationship was assumed to be ongoing, a trader could condition his decisions based on the information he has on his partner’s past trading behavior. If trading levels are frequent, then a Tit-For-Tat (TFT) strategy can be employed in which a trader will play whatever his partner played in the previous iteration. For example, a trader that is cheated today will cheat his partner in the next period.
This logic holds for merchants that do not engage in bilateral trade frequently, but trade frequently with the community. If information on cheaters is widely shared in the community, then even if a cheater changes partners he will be cheated in his next partnership if that merchant is honest. Milgrom terms this a transferable reputation of honesty. Why would an honest merchant who was not cheated, want to cheat someone else? Well, cheating the cheater is more profitable, per the PD, and because the punishment is delivered to the dishonest merchant by cheating him.
By extension, this logic then holds that under typical circumstances of cooperation, no merchant can gain from deviating from the rules at any time so long as he assumes other merchants will also be honest in the future. The boycott mechanism of the system remains effective when all merchants in the community are informed of other’s reputations.
Reputations of Honesty
Without centralized institutions, each player has little to no access to information about traders in the community. The limited information on a current partner’s past behavior, and lack of institutional enforcement, means there are no incentives to be honest.
An institution capable of informing each trader of the other’s past behavior may be able to restore these incentive problems, but they are costly and inefficient. The Law Merchant’s solution to this is adequate information — meaning that neither trader need know all past behavior, but rather the information on their partner in the previous period. If a trader is aware of his partner’s behavior in the preceding period, he will be capable of executing the punishment against his partner, even if he was not the one who was cheated. In doing so, the Law Merchant ensures that a player who formerly cheated cannot simply start over with a new partner to avoid punishment from cheating the previous period.
The costliness, inefficiency, and information inaccessibility of reputation systems are mitigated by the ‘Judge’. This specialized role — also known as the ‘Law Merchant’ (LM) — serves as a centralized figure for adjudicating disputes and holding information.
Parties have the ability to query the judge, prior to finalizing their trade agreement, for records on their partner’s previous behavior. Without querying the LM, this model assumes a trader has no information on his partner’s past. If a party is cheated, they may appeal to the LM. He has the ability to award damages if the defendant was found to have cheated, however payment of the damages is voluntary as there is no police enforcement. A player may not bring a case against his partner to the LM if he did not query the LM prior to finalizing their trade agreement.
The system is not viable if the cost of investigating a complaint is too high. If costs exceed benefits, then no trader will expect that others will make claims or pay judgments. When costs are low, and threats are credible, is when the system actively acts as a deterrent. The benefits of cheating must be small compared to the value of continuous honest trade.
In this way, the Law Merchant solved the highest problem: information cost. The ‘Judge’ is a centralized information center and each trader need only refer to one place. Traders do not need to seek out the information on their current partners history independently, a method that would be far too costly for the system to be effective.
Applicability for the Blockchain
Like the Medieval Merchant, the blockchain network is a transnational actor; it operates in every jurisdiction, all around the world, sometimes despite a state’s attempts to restrain it. Other transnational groups, even those as barbaric as terrorist groups, all have rules on how they govern themselves. This happens for one of two reasons: 1) the state does not recognize the transnational group, and therefore its laws do not transcend to the group, or 2) the transnational group does not recognize the state as having jurisdiction over its network, and therefore do not recognize the state’s laws.
Blockchain falls in somewhat of a grey area between those two. Its technology is so new that states are not sure how to proceed in handling it. But at the same time, the decentralized network of the blockchain was borne from an inherent distrust in the centralized state as an honest enforcer. The Law Merchant was borne from a lack of police enforcement by a centralized nation-state. As such, both share the premise that the governing structures in place are not fitting for the purpose called for by the systems. But, the Law Merchant faced this struggle first, and its discoveries are highly applicable to Blockchain governance — a concept still in its infancy.
For the blockchain, when a transaction occurs completely on-chain, the contract is self-enforcing and no adjudication is required because it is nearly impossible to cheat. But, not all transactions are on-chain. In fact, as the blockchain scales, more and more transactions will be dependent on some off-chain element that is anticipated to be completed at a future time.
But, what about Lightning network security?
The Lightning 101 series on Medium defines security for Lightning Transactions as:
“a signed bitcoin transaction with a special smart contract that has not been included into the bitcoin blockchain yet. This smart contract allows us to chain unconfirmed transactions together in a secure fashion. The participants in a Lightning smart contract are allowed to repeatedly update the chain of unconfirmed transactions, with only the newest one being valid.”
While off-chain elements have the potential to be governed by on-chain elements, such as those employed in Lighting Transactions, it is not the most effective nor time-optimizing manner to govern off-chain behavior. Let me add depth with an example:
Say Lily wants to buy socks from sock producer, John. They enter into a contract together in which Lily agrees to pay one token for one pair of socks. The possible outcomes from this game are vast, but a few are:
- Lily receives her socks and pays John.
- Lily receives her socks, but tells John she never received them.
- Lily does not receive her socks, and tells John she never received them.
- John never sends the socks, but tells Lily he sent them.
- John sends socks of much poorer quality than those promised to Lily.
If this transaction was secured on-chain, it would most likely be in the form of a special smart contract that has not been included on the blockchain yet. So, in that contract, it would have to control for every possible outcome of the transaction — a difficult task that is easy to manipulate. Beyond this, how would the contract determine if Lily is or isn’t lying when she says her socks never arrived? To control for this, a third party, the delivery boy, would need to confirm if the socks were received by Lily.
If this transaction was secured off-chain, by the methods outlined in the previous section, preventative action would have taken place by querying the LM prior to agreeing to the contract. Lily would know if John was a cheater, and vice versa. If either players had cheated in the past, the other would refuse to continue the partnership, or cheat the cheater. In addition, the risk for cheating is higher in off-chain governance than it is on-chain; off-chain, the cheater is risking all future payoffs for α, whereas on-chain the cheater is simply risking a collateral cost, c, of burning tokens tied in the special smart contract for a single period.
In the absence of information, the PD shows us that in these off-chain instances a player is going to cheat, every time. Because the information on a players previous trading history is not easily accessible, transferable reputations of honesty are not strong enough to deter the player from cheating consistently. A reputation systems has the potential to work, so long as those engaging in the community are incentivized to remain informed on who has cheated. In fact, in governing off-chain elements, there is no reason that the Law Merchant reputation system shouldn’t be effective, according to the simplified PD dynamics.
In my work with CasperLabs, I have proposed a reputation system as a means to create self-enforcing dynamics between players without formal models of enforcement provided by nation-states. This idea stems from ostracizing those who cheat by community boycott after the judge rules that specific player guilty. In doing so, the player may benefit α from cheating once, but his play will result in grim trigger responses with discount factors so high that every period after, his payoff will be zero. This is an ideal situation, assuming the community legitimizes the judge’s rulings and are willing to give up their future payoffs from trading with the cheater to benefit the overall community.
This idea reiterates Milgrom’s argument that players must remain just enough informed to be able to reciprocate TFT strategies against other players. Without any information on a trading partner’s past behavior, there is no incentive to play Honest. Introducing information into the game on each trader’s past experience can restore incentives to participate honestly. In addition, non-cooperative games can be mitigated by on-chain smart contracting tools, as aptly described by Virgil Griffith here. But as he concedes, warping games can only be triggered by on-chain events; off-chain events can side step measures set in place by smart contracts.
These incentive dynamics between the judge and the players promote Honest behavior between parties, and punish those who do not value informing themselves enough to protect their transactions. The effective execution of an off-chain ‘judge’ has yet to be seen; while blockchains such as EOS have tried their hand at creating a court-system off-chain, it proved wildly ineffective shortly after its initial implementation.
And who would this person be? Validators? The elected-board? Validators have the largest supply of information on trading behaviors — but mostly for on-chain transactions. The elected board members have conflicting incentives, both for themselves and their constituents. I argue that the judge should be a completely independent body separate from other governance sectors on the blockchain.
The Judge is only a central authority if people grant him legitimacy and report instances of cheating. Only by logging these instances of misfortune can the Judge expand his knowledge on past behaviors to create sufficient transferable reputations on enough people to be accurate most of the time. This means that a certain number of off-chain transactions will have to occur, initially, for the Judge to develop sufficient reputations of honesty for Users.
Beyond this, my largest concern for judicial systems on the blockchain is corruption. On-chain governance systems are already suffering widespread issues with vote-buying and uneven stake-weighted systems that favor the wealthy. With this in mind, I propose a judiciary which is tenured for life with one caveat: judges must release all financial statements once, annually to ensure all rulings are being handled with impartiality and fairness. In this system, extortion is unlikely as judges receive payments for conducting investigations at a Cost, C. If he extorts a player, that participant will stop bringing cases to the judge, and the judge will stop earning C per investigation until his total earnings go to zero.
Above I have shared with you, what I believe to be, the most critical aspects of Milgrom et al.’s piece. The age of the Internet has completely altered the cost and accessibility of information. Reputation systems are already developing through technologies such as Uber, Fiverr, UpWork, and Airbnb, among others. The beauty in this is that blockchains inception occurred at the best point in history for governance. The accessibility of information has never been higher, and by extension, reputations as a form of governance have never been as pervasive as they are today.
Both the Medieval Merchants and the Blockchain have a shared ancestry. As Benson (1989) stated:
“The Law Merchant was conceived of as coordinating the self-interested actions of merchants, but perhaps an equally valuable insight is gained by viewing it as coordinating the actions of people with limited knowledge and trust.”
This also means that governance has come full circle. Yet, I am not so convinced that the past cannot serve as a foundation for present technology to build upon. The Law Merchant proved successful for decades, without the technologies of today. Imagine what kind of system could be built atop this already successful model for self-enforcing agreements of the past, in the present.