Graded Token-Curated Decisions with Up-/Downvoting — Designing Cryptoeconomic Ranking and Reputation Systems
Joint work with Sidd Bhasin. Videos below.
A Graded Token-Curated Decision Protocol is a cryptoeconomic mechanism allowing an amorphous group of token holders to up- or downrank an object. Graded Token-Curated Registries implement fair ranking and reputation systems by alleviating the need for a centralized third party and by disincentivizing participants to provide fake or biased decisions.
1. Introduction
Token-Curated Registries, for short TCRs, are the new kid on the block in the blockchain community, primarily because they uniquely demonstrate two properties at the heart of every blockchain, namely application design paired with cryptoeconomic incentive mechanisms. They tackle some very important problems and have all it takes to disrupt curation lists. Some of the world’s richest tech companies, such as Google, Amazon and Facebook, are so successful because they are expert curators of relevant data and goods. However, existing curation protocols have an unstable mechanism: one must trust the centralized curator (as in the case of google data curation) or malicious reviewers can bias the outcome of the ranking (as in the case of amazon’s reputation system). In this article, we put forth the theory of graded token-curated decisions which build upon and considerably extend the concept behind TCRs with graded up and down voting of an entry. They remedy the above shortcomings of centralized curators and significantly increase the trust and value of curation lists. Graded token-curated decisions (gTCDs) give raise to a plethora of new applications and services. We demonstrate the versatility and practicability of gTCDs by showcasing a decentralized, trustless ranking/reputation system.
2. Token-Curated Decisions and Applications to Registries — A Recap
Every public blockchain uses tokens as a way of incentivizing participation in the network. The mechanism design is so crucial because it represents the shared belief among users about the value of the token. Tokens have enabled companies to create entirely new business models, particularly decentralized business models. For example, Steem uses its tokens to incentivize its users to contribute news and content who eventually get paid in Steem tokens for their contribution to the platform. Another interesting example is Augur, where people are incentivized through tokens to contribute to the outcome of events as ‘oracles’. Even Ethereum and Bitcoin themselves provide a monetary incentive to people for providing their computing powers to process transactions.
Playing with the value of the token to incentivize good behavior and disincentivize bad behavior based on predefined rules is the premise on which TCR’s can be further explored. As Mike Goldin puts it best:
Token-curated registries are decentrally-curated lists with intrinsic economic incentives for token holders to curate the list’s contents judiciously.
TCR’s have a plethora of interesting applications, including but not limited to
- whitelisting good sites to protect against scam and phishing, as laid down for example by MetaCert
- provide advertisers with a community-vetted stamp of approval on the websites best suited for serving ads, as deployed for example by adChain
- decentralized, community-governed maps of points of interests based on crypto-spatial coordinates, as introduced by FOAM
The mechanism TCRs implement can be seen as a community-vetted list. In a nutshell the token-curated decision protocol runs between a listee and the community as illustrated in Fig.1.
Protocol Goal: Generate and maintain a list of good objects (e.g., web sites).
Protocol Assumption: Listee wants to be on the list, e.g., because she receives access to some exclusive service or discount.
Protocol Setup: Smart contract implementing the token-curated decision mechanism. Any (third) party can implement the contract. The nature of the blockchain makes the contract trusted, as the code is publicly verifiable and the proper implementation of the decision mechanism is scrutinized by the trustless blockchain network. Hence the big advantage of the blockchain is to implement trust and alleviate the need for a trusted third party (being hard to find in practice by the way).
Protocol: The protocol runs between the listee Alice and a set of community members, abbreviated as the voters in the following way:
- List(): Alice stakes some tokens with the TCR to register an object. In the previous example the object is a website that shall be advertized.
- Challenge(): The voters counter-stake the same amount of tokens proportional to the total amount of voters. Each voter casts a vote and decides, if the the object gets listed. (In our example, the voter is a participant of the ad network and decides, if she likes to see ads provided from that site.) The quorum of voters decides whether the objects gets on the TCR. As reward for taking part in the process, the stake of voters deviating from the quorum choices (including that of Alice in case of rejection) is shared among the winning voters.
- Unlist(): At any point in time, Alice may request to unlist her object. In return she obtains her stake back.
Analysing the token-curated decision protocol, let’s look at the incentive strategies of all players. Listees are incentivized to stake tokens because they desire to be on the list (by assumption). However it is worth noting that their expected outcome from being on the TCR must amortize with the staked tokens (and a rigorous analysis needs to take that into account). Voters are incentivized to participate in the challenge because they obtain a reward. Most importantly, malicious voters are disincentivized to cast a biased or unfair vote, as this leads to forfeit of their stake.
3. Graded Token-Curated Decisions and Applications to Registries
All in all, token-curated decisions are a powerful new technology to implement attractive and exclusive lists. These lists have a significant value for many business models and applications. In my opinion a vast amount of applications beyond what we understand at the moment is possible. TCRs are the first applications, but their wide-spread adaption may be impeded in various emergent settings. The limitations with TCRs in their current form are twofolds.
- Static List: TCRs judge a listing on an in-or-out manner. It’s like passing the bouncer at the Berghain club (known for the hardest door policy in Berlin). Once you made it in, you enjoy literally a timeless night club experience. But life is not binary. It is rather strictly dynamic and subjected to ongoing changes. Suppose we curate a list for CO2-emission friendly cars. Before the scandal, Volkswagen cars met the requirements of the list. But after the revelations, some cars not meeting the CO2-emission threshold should be dropped from the list. While the Volkswagen scandal serves only an illustrative purpose, the point to make here is that a TCR can outdate and loose its value as a trust and validation anchor. The workaround is to create a novel TCR and ask all listees to apply for a new listing, which is generally an inefficient and time-consuming process.
- Binary Metric: TCRs assume the uniqueness of the object. But objects co-exist in an environment. We do experience similar objects. Suppose, for example, you want to list restaurants in a token-curated registry. A TCR for a Thai restaurants allows you to classify the restaurants only. But it leaves out the capability to add more valuable information about the restaurant, namely the quality of the food in comparison to related restaurants.
To harness the full potential of token-curated decisions, we introduce the concept of graded token-curated decisions. Graded token-curated registries (gTCRs) remedy the above weaknesses and extend TCRs with the utility of dynamic object up-/downvoting. The additional feature gives rise to the cryptoeconomic dual application of ranking and reputation systems where objects can compete with related objects.
As before, the token-curated decision protocol for a ranked list runs between a listee and a group of voters, depicted in Fig. 2.
Protocol Goal: Generate and maintain a list of ranked objects (e.g., web content).
Protocol Assumption: Same as before. Listee wants to be on the list, e.g., she receives access to some exclusive service or discount provided her rank meets a threshold.
Protocol Setup: Same as before. Smart contract implementing the token-curated decision mechanism. Any (third) party can implement the contract.
Protocol: The protocol runs between the listee Alice and a set of community members, abbreviated as the voters in the following way:
- List(x): Alice stakes x tokens with the TCR to register an object. Here x is the minimum stake required by the gTCR. In response to the request, Alice is challenged by the community. If she passes the challenge, her object is ranked according to the total amount of locked stakes.
- Upvote(x): Alice or any other party adds x tokens to the TCR in order to upvote an object. For example, Bob likes Alice’s object and believes it deserves a higher ranking. As the list is public, Bob can easily identify the current rank of the object and push for a higher ranking.
- Downvote(x): Alice or any other party adds x tokens to the TCR in order to downvote an object. For example, Bob dislikes the object because after time the object does not meet the expectations. In order to reflect the changes, Bob casts a challenge to downgrade the ranking of Alice. If the total amount of staked tokens is zero, then the object is taken from the gTCR.
- Challenge(x): The voters counter-stake the same amount of tokens, i.e. x tokens, in proportion to the total amount of voters. Each voter casts a vote and decides, if the the object gets listed through a quorum vote. As reward for taking part in the process, the stake of voters deviating from the quorum (including that of Alice or the staker on behalf of her in case of rejection) is shared among the winning voters.
- Unlist(): At any point in time, Alice may request to unlist her object. In return she and her supporters obtain the stake back.
Analyzing the graded token-curated decision mechanism for ranked registries, the difference to the previous analysis are the upstake and downstake functions. Upstaking is clearly in the interest of the listee (and her supporters) as it may lead to a higher ranking. Malicious up or downstaking in order to bias the ranking is disincentivizes as the up/downvoter risks to loose her stake.
Acknowlegement
The theory of graded token-curated decisions and applications to ranked registries is clearly a community effort and a result of the interaction with key players in this young research field — be it through face-to-face discussions or be it through remote channels like email, telegram, medium posts, whitepapers or github releases. Kudos to Consensys people Wayne Chang and Harrison Hines who brought up the topic and Mike Goldin, Ryan John King, Paul Walsh who demonstrated with their projects first meaningful applications. Big shout out also to Achill Rudolph and Ocean’s people Dimitri De Jonghe, Trent McConaghy and Bruce Pon whose protocol based on curation marketplaces influenced the ideas behind gTCRs. (In fact, we should flesh out the differences next time we meet for some beers.)
Addendum
Below videos of two talks a gave on graded token-curated registries. The first happened at the Token Engineering Meetup during Blockchain week NYC 2018. The second is recorded at the Curation Market Meetup #3 in Berlin a week thereafter. Credits go to team Ocean and Way.
