Quality Control Solutions For Token Curated Registries
Member-curated registries, such as Token Curated Registries (TCRs), typically operate like this: I start a club. To be considered for membership in my club, you need to pay an application fee and enter the Thunderdome. For a period of time, any club member can challenge the validity of your application by anteing up an amount equal to your fee, and putting the matter up for vote by existing network members. In a challenge, the loser loses their coins to the winner and to voters in the majority bloc; two participants enter, and only one leaves with their coins! If there are no successful challenges for the duration, then you are accepted into the registry.
The goal of this scheme is to maintain a high quality registry by aligning incentives among the Voters, Challengers, and Applicants. Challengers stand to gain when they successfully oust a bad Applicant. Voters are rewarded by participating in the winning bloc. Applicants are rewarded with network entry for passing applications.
There are at least two critical assumptions upon which these structures are riding:
- Challengers and Voters are able to acceptably assess applicant quality.
- Network participants are sufficiently motivated by prospect of long-term network value growth to behave well, participate, and resist short-term gains at the expense of the network.
The rest of this post dives into issues around the first assumption and improvements that could assuage it. The second issue plagues decentralized ecosystems at all levels, but it is briefly addressed near the end.
How do you measure and ensure quality?
Voters and Challengers will have a tough time assessing Applicant fitness in the absence of reliable data and audit know-how, yet it’s a key activity to create desired system behaviors. If voters vote blindly, then they perform no useful curation activity for the network, and thus the network will be on its merry way to death. No one wants to transact if there is no quality.
This may result from severe information scarcity or asymmetry. For example, in a registry of car owners, it’s unreasonable to expect that a car owner would be able to acceptably assess the quality for cars on the network, perhaps even their own! For complex assets, this market failure can be even harder to address than in the market for lemons, which makes the assumption that asset owners know a thing or two about the quality of their own assets. With little available information about the listings on the network, the problem goes from very hard to virtually impossible.
Solution 1: Membership Standards
We can require that members implement verifiable standards as part of the application process. Each item in the standard can be checked by network participants to provide anchors of trust. The standards may be machine-verifiable, human-checked, or a mix of both depending on the marginal costs of each check against the added usefulness for assessment.
This already happens today. AirBnB has several anchors of trust including social proof, phone number verification, and inspection of documents. Exchanges have longed used KYC to fight bad behavior. Ride sharing firms run background checks. Several blockchain projects today have already devised new ways to regulate the marketplace such as Proof of Space-Time to validate digital storage properties and Proof of Spank to verify age. Keybase, uPort, Civic are creating universal identity systems, the latter two utilizing blockchains by default. Registries should borrow or build such tools to ensure quality.
Solution 1 Example: Car Owners
A registry for car owners may require that applicants also submit (likely off-chain) VIN, digital scans of the title, a non-expired vehicle registration, and an image of the vehicle with recent emissions stickers. Information may be partially omitted or provided on a need-to-know basis to increase user privacy.
With this data available, network participants are better equipped to decide whether to challenge or not and also which way to vote. Steps like checking the VIN against an Internet database and confirmation of non-expiry of the vehicle registration might be handled automatically by project-provided software clients ran by or on behalf of network participants. Trickier checks such as photo verification might be completed by humans using client software also provided by the project. In the long term, technologies such as machine learning will likely alleviate verification costs.
Solution 1 Example: Ad-Supported Websites
Advertisers believe that web publishers regularly defraud them for millions of dollars. Projects (notably adToken) have attempted to use TCRs to provide a whitelist of websites for fair exchange. However, the verification that websites are honest and serve genuine traffic is no easy task. How can this be improved?
Membership standards are one answer. If we require participants of the marketplace to provide KYC and prove that they are obligated to bear the counterparty risk of sanctionable punishment in the case of illicit bot fraud, we can go a long way in building advertiser confidence that the websites are indeed genuine and will have decent returns for the claimed categories. The network governance might even decide that an independent third-party validator conducts periodic audits of all network participant claims against the correct nation state databases.
Solution 2: Network-Appointed Validators
Who does the checks? Network participants are incentivized to do so thanks to rewards from challenge victories, but as discussed, they are actually in a poor position for evaluation of applications. Self-regulation would be tricky without a sophisticated arbitrator to have the last call.
In Rules for a Flat World, Gillian Hadfield imagines a system of for-profit private regulators who are held accountable to super-regulators. In her theory, the competition among private regulators will drive the regulation products to higher quality and/or lower cost. Regulators would be able to specialize to assess very complex and intricate applications. Hadfield’s paper recommends that regulators demonstrate the following properties:
- The regulator is a private (non-state) actor (profit or non-profit).
- Regulation is mandatory: regulated entities must choose and submit to the regime established by a regulator (and pay any fees charged by the regulator.)
- In order to participate in the market for regulators, the private regulator must be approved by the state.
- The market for regulatory services is competitive.
- The regulator has the capacity to recruit the public enforcement apparatus of the state to sanction non-compliance with the requirements of its regulatory system (beyond contract enforcement.)
In our model, the “state” and super-regulators could be satisfied as a democratically-elected council per registry-based marketplace. They would have working sessions and create proposals for network rule adjustments and measurable high-level objectives that approved validators must be capable of assessing. Approved validators may exist on a registry maintained by the council and through network vote. All applicants could be required to solicit an approved validator as part of their application process. Whether the validator’s services are paid for by network transaction fees, the applicant, or a mix is to be determined.
Solution 2 Example: Car Owners
Auto mechanics can apply to become validators that evaluate images of car internals for defects, ensure that vehicle mileage matches usage (e.g., CarFax) records, and perhaps even conduct a live in-person or virtual inspection. On-chain attestations such as ASE certifications, proof of business existence and operation, proof of insurance, and physical address verification may be used as part of the application process for the governing council to accredit validators.
The auto mechanics acting as validators are compensated for their efforts. They earn a new stream of revenues in addition to their main business lines, and we may eventually see some auto mechanics fully evolve into specialized auditing agencies, which enjoy better economies of scale after devising robust inspection processes. The amount and intensity of audits will be subject to market forces, as high-value applications will command more in-depth inspection, and vice versa.
To keep the auto mechanics aligned with the high-level network objectives, their audits can be tracked across all the applicants they evaluate, which is already an excellent use case of blockchain technologies. If one validator makes calls that are consistently contradicted by other validators in future audits, then this may be grounds for the council to conduct an investigation of its performance.
Since all the audits will be available to the council (and possibly to the public) on the blockchain, investigation of a validator will become very cost-effective compared to absurdly expensive e-discoveries suffered today that can cost over $18,000 per GiB. Other signals that may help regulate the validators include percentage of approvals compared to the network average, time to approval, and ratings from applicants. This model is contingent on a competitive marketplace of validators and acceptable oversight performance from the governing council.
Again, there is nothing new here. Already, doctors and nurses are paid to perform part-time medical reviews for health insurance companies. Anyone can start a mobile or virtual notary business on the side. These are patterns we’ve seen before in other industries that can be applied to satisfy the quality demands of member-curated registries.
Solution 2 Example: Overseas Business
In Poorly Made in China, business consultant Paul Midler discusses his insights from working for an American company with products manufactured in China. Due to a lack of accountability and legal enforceability of contracts, ailments such as quality fade from manufacturers and non-payment from buyers are commonplace. The cheated party then eats the costs and finds new partners with the same risks. Playing the prisoner’s dilemma for only a single round results in reminiscent outcomes.
Could we formulate a game of iterated prisoner’s dilemma to achieve tit-for-tat? A curated registry of U.S. importers and Chinese manufacturers has the potential to do so, but has high upstart costs, large required network effects, and some elements of the tragedy of the commons. It is unlikely that a single company will see enough benefit to pay massive fixed costs to bootstrap such a registry. Even if a high-quality registry existed under a central entity, users on the opposite side of the world would need to trust that the entity is unbiased towards its own sector and nation state. Fortunately, these are all problems that we think blockchain technologies will be good at solving.
There could exist two ad-hoc registries: one for buyers and another for manufacturers. Both registries would include cross-referencing transaction logs tied to members. Each registry could maintain its own governance of high-level network objectives, as different groups have different assessment criteria.
The buyer group could require its members to maintain transaction ratings from manufacturers, demonstrate cash-on-hand minimums, and transact volumes in an acceptable range. Network validators can determine for themselves the best way to inspect adherence to the outcome-based goals such as reducing fraud rates.
The manufacturing group could require its members to maintain transaction ratings from buyers, demonstrate proof of factory and land ownership, produce educational certificates for employees, and release on-site inspection histories by auditors. Again, validators may create their own regulatory products that will meet network goals such as high quality transactions or low rates of fraud and defects.
A New (Validating) Hope
As the network grows, the buyer and manufacturer registries themselves may split and specialize. High-volume buyers may best benefit from different membership standards than those of low-volume buyers. Manufacturers of electronics may have very different auditing procedures than those of clothing. Enough difference could justify a mitosis, wherein a fraction of the group members opt to found a new further-specialized registries with better-suited rules, not unlike splitting in The DAO.
With the large assumption that the contract code is free of serious defects, the prospective benefits are that the new registries should have the simplest network rules required to satisfy their average participants, along with newly-specialized validators that cater to that specific registry.
When there are sufficient children registries, we may even consider a registry of registries to institute basis standards and interoperability among groups. This master registry would also add further accountability, encouraging sub-registry members to boot unscrupulous members before they commit actions that may result in booting of the entire group itself from the master list.
With curated registries and membership standards, we have a tentative solution to the overbearingly large regulatory structures of today. In software engineering, large monolith codebases are split into microservices to reduce complexity through separation of concerns. This may be a worthy approach to investigate for quality-oriented marketplaces on the blockchain.
***
Nothing in this article should be taken as legal or investment advice.
