Token Curated Registries probably aren’t going to change the world
Token Curated Registries (TCRs) are a curious concept. At their heart they are a means of creating a list, a rather prosaic sounding goal. Scratch the surface, however, and they are an entangled web of incentivization and game theory that could cause more problems than they solve.
I’m not going to dig too deep into what TCRs are a there are a number of good such explainers. In short, TCRs tie together lists, tokens and voting in such a way as to allow for curated lists to be created in which items are proposed and voted on for inclusion. The idea is that by creating a tailored list, it will generate value for those included. This will then see others from similar fields wish to be included on it, necessitating they buy the list tokens so they can submit themselves for consideration. Those curating the list likewise receive tokens for their efforts. Consumers, meanwhile, benefit from better lists that aren’t subject to the whims of one entity and where voting is transparent and accessible to all.
The focus of this article, rather, is to list all the things I see as problems for TCRs. Many of these should be immediately obvious just from the summation above. Although I discuss three examples of live or proposed TCRs, I am focused more on the concept as a whole.
Lists? You’re seriously talking to me about lists?
The nature of TCRs makes them hard to get too excited by. A list? Great. Who cares? This is partly as a result of people giving really poor examples of use cases. A list of coffee shops in San Francisco? Really, who cares? These dull examples make the output of TCRs hard to conceptualize.
However, there are some interesting use cases of TCRs. One such example comes from FOAM Space, which is a good representation of how seemingly banal lists can be used to generate something far greater.
FOAM aims to provide a decentralized map through which users can add Points of Interest. These can then be voted on and, if approved, are then added to the FOAM map. The TCR in this example is represented by these Points of Interests on a map.
Bearing in mind that I think FOAM is one of the better examples of TCRs currently proposed, let’s look at some of the potential positives and negatives TCRs bring:
I wanted to start at the top with a clear example because there are many different proposals for TCRs but few concrete examples given. There are also a great deal of different TCR implementations being worked on and there is no one ‘TCR’ model that can be referred to as standard. However, they have the same overarching structure and my many issues with TCRs can be boiled down to four core categories:
- Weak core proposition
- Misalignment of incentives
- Limited and poorly applied use cases
- Token model
Weak core proposition
TCRs are designed to allow disparate parties to come to an agreement. However, the information being agreed upon can be both objective and subjective. Subjective information brings many obvious issues, but even objective information, as outlined in the FOAM example, is often hard to verify. “Don’t Trust. Verify” isn’t some abstract concept. That anyone can verify the blockchain themselves sits at the heart of Bitcoin. This is lost when it comes to TCRs, because there will always be an element of trust. People can still verify the data recorded and transactions made, but because these transactions are recording objects external to the blockchain they then have to trust that this is correct (the Oracle problem).
Bitcoin enabled censorship resistant transactions, an obviously useful function for anyone remotely familiar with either history or the current global financial system. Decentralized lists are hard to create the same compelling narrative around, at least not one requiring tokens and a blockchain. Wikipedia covers a broader range of content than any TCR could aim to solve, and yet does so admirably. Are there inaccuracies? Yes. Is it subject to people vandalizing it? Sure. Are either of those problems solved by TCRs? No. I am not convinced that TCRs are the solution for the curation from which they take their name.
Centralized entities are incentivized to make their lists as good and as accurate as possible due to the competition of the free market. Google and Apple will have their own biases, but their continued success lies in part (the power of incumbency can’t be understated) on their ability to remain competitive and deliver an experience that users want. Any bias they have likely rests at a higher level rather than filtering down to individual decisions e.g. if a coffee shop should feature more prominently than another.
TCRs may not have an overarching bias dictated from above, but there is a greater chance of issues at a lower and individual level. However, platform bias is also likely depending on how demographically widely distributed the curators are. A predominantly American/Western European, wealthy, white, male and middle aged userbase — such is common on many technological platforms — will undoubtedly bring through bias through homogeneity.
There will still be plenty of objectionable things that are accepted into curated registries, just as they affect a system like Wikipedia. FOAM notes the issue with OpenStreetMap, which suffers from vandalism and outdated information. While staking tokens may reduce some of the lower levels of mischief, it won’t stop determined users. Preventing them relies heavily on active governance of token holders, something I do not see as a given despite incentive. Moreover, attaching value to such a system can cause issues of its own.
Misalignment of incentives
There are a range of potential issues with how TCRs implement incentives.
A prediction market like Augur needs to incentivize responders so that the correct answer is provided to the questions set. This is what allows the bets to be resolved, and the payouts made to the correct party. Responders earn rewards for correctly settling bets (although there have been ambiguous questions which have clouded this process). This can lead to some issues. For example, if a losing bettor controls enough of tokens being staked, they can 51% attack the system so that they receive the payout despite being wrong.
This is unlikely to happen with Augur. Poorly written questions aside, the correct answer is usually a clear and objective answer. “Who will win the Super Bowl” is not difficult to answer for most responders, requiring a simple google.
Ryan Selkis aka TwoBitIdiot opined that AdChain’s (which aims to produce a TCR listing ‘honest’ websites for advertisers to use so that marketers don’t overspend on websites inflating viewing figures) TCR implementation was “simple…black and white”.
Unfortunately, even these seemingly clear cut decisions rapidly devolve once you bring a voting public into the fold. Curated lists often depend on more subjective qualities. Early votes on Adchain saw both Facebook and New York Times challenged and rejected. The challenge for the New York Times was brought about due to their journalism in respect to the Iraq War, an element that the TCR was not intended to consider or reflect.
This is not an extreme case but rather is the entire game. To prevent such egregious errors, actors have to be incentivized to act in the best interests of a) users and b) the system. Instead, actors are incentivized to act in the best interests of themselves.
Consider these examples:
- Token holders receive tokens for voting correctly on a decision. But voting correctly does not necessarily lead to the right decision. Let us assume that the New York Times does not game its viewer metrics. Rejecting them from the AdChain TCR would be the wrong decision. But, if I felt the community was going to reject them, on a financial basis I would have to reject them too — because otherwise I would get no tokens.
- I am a business that wishes to be on a list. My rivals are on the list and it is causing my business to lose significant business as a result. Therefore, I either need to get on the list or — if rejected — ensure the list isn’t used as much. As such, I can spam the network with poor submissions to try and make sure that my real submission gets through. Or I buy enough tokens to be able to sway voting my way. Or I issue numerous challenges to block the network. It could cost me a significant amount of money, but if this amount is less than the amount that I am already losing on an ongoing basis as a result of not being on the list then it is still worth doing.
Bad actors or those with a vested interest in appearing on the list will be more driven than honest actors who aren’t on the list themselves.
This is reminiscent of my complaints about the RecDAO system that was trialed on Reddit. It aimed to decentralize moderation of the ethtrader subreddit such that users could vote on whether posts should be promoted or deleted. The obvious problem with such a solution is that those who care about the post have a far greater interest in voting on it. The average person doesn’t care or may not know. However, TCRs rely on achieving a mass of users, because otherwise you might as well just take a count of the 20–30 people making decisions. This needs the average person to be engaged.
Achieving this engagement is unlikely. Voting turnout tends to be low for important events, let alone for ongoing governance. Blockchain governance participation has tended to be very low to date, meaning that those engaged carry an outsized impact. Voting takes time and effort, as well as being constantly engaged.
This low level of engagement means that it is frequently not a 51% attack, but a 5–10% attack. This is exacerbated when you consider that just to challenge an application, the user must usually put up a financial contribution of their own. This means that I have to feel strongly enough to challenge and also confident enough that other voters will vote my way. This is not a small hurdle.
Furthermore, governance usually takes the form of voting on issues on an irregular and dispersed basis. However, most successful TCRs will likely require constant voting. This means that voters will have to be involved on a weekly if not daily basis. I am not convinced this is likely and for this reason I prefer the approach of something like Messari, where the default result of an application is a rejection to others which see the default as approval. This means that a quorum has to be achieved to add new entrants, preventing additions from acceptance following a vote with miniscule turnout.
Regardless, getting people to vote is a large concern. For a TCR to succeed they will likely need to gamify or otherwise hold the interest of users. I don’t think financial rewards by themselves will prove sufficient.
Poorly applied use cases
Although currently limited in numbers, TCRs are currently being proposed for use for a whole manner of means. Most of them are simply bad fit for TCRs, with Civil a particularly egregious use.
Journalism is highly subjective and emotionally charged. Using TCRs as a means to create a list of ‘trustworthy sources’ is a marriage made in hell. On a practical level, publications are uneven. Some journalists at Company A will be great, others less so. The Guardian produces a decent newspaper, but their online clickbait journalists are awful. Practicalities aside, it is difficult to reconcile left and right wing views when it comes to newspapers.
I struggle to see how TCRs are going to help us here. Linking subjective moral judgments to payments doesn’t seem like it will end well.
TCRs will have their fits, but will likely be more narrow than originally envisaged.
TCRs are built upon the premise of the token underpinning the list being valuable for incentivization purposes. This leads to questions such as:
- How are tokens distributed initially? Are they distributed widely enough
- If TCRs are successful, does this mean the need for thousands or tens of thousands of tokens?
- If so, how can hundreds or thousands of lists maintain the value required to secure them?
- What happens if the token starts off and is undervalued while the market catches up? This could lead to risks for the project on an ongoing basis if bought up by a limited number of investors and presents attack vectors while the value of the token is realized
- How are TCRs going to solve the issues faced by other projects such as friction, regulations and lack of trading volumes?
Of particular concern is the initial distribution. Teams taking large proportions of tokens are bad enough for projects in which they have no utility — no team should be taking significant amounts of tokens when they are used to vote on these subjective matters.
Adchain, for example, saw half of the token supply sold to the public with 20% each reserved for MetaX and ConsenSys. This 40% of tokens were locked for between 12 and 18 months post public sale. I have no idea if either entity are using their tokens to vote but equally if they still hold these positions then they can win any vote they desire. Trusting centralized entities to abstain from voting usually only holds until people vote ‘wrong’.
I don’t necessarily want to like scare people off [by outvoting others with less coins to stake as group norms are built]. But New York Times, they actually lost that [challenge], which like is offensive to me as an adToken holder. So, now I’m personally going in much harder with my own tokens. But it’s interesting to see the curation process play out, to be a witness to it. Mike Goldin, Adchain Lead Engineer (CryptoInsider)
Besides anything else, the idea of a token for a list having value is quite jarring. It just doesn’t feel ‘right’. The assumption that lists should have value in a direct form is unproven. Even if ultimately they do, given wider crypto adoption patterns it is likely a decade or more before the infrastructure is in place for them to be widely used. Other worker tokens seem easier to implement with the potential to be as effective as TCRs. For example, paying users in ETH to add Points of Interests makes sense, with advertisers or companies able to pay to for better spots on the map could make sense (although it sounds horrible). Paying FOAM to maintain a TCR doesn’t make sense to me. But this is what many of the TCRs models amount to.
It always feels like punching down aiming at laughable projects, so let’s look at another well-grounded and useful project that has been at the forefront of TCRs thus far. Messari, led by the previously quoted Ryan Selkis, aims to help the industry self-regulate by introducing disclosure and transparency standards. This can only be a good thing.
Where I diverge from the project is in their belief that a TCR is instrumental to success. This TCR would be used to create a whitelist of projects which have conformed to a set of standards, as defined by the TCR Curators. The idea is that:
- Applicants (crypto projects) pay a $25k fee to apply to the Messari TCR (MTCR), submitting their disclosures and other information as required
- Curators (token holders) accept or reject their application, as well as setting transparency standards and authenticating disclosures. However, they can also delegate this work to Validators
- Validators are selected and paid by the Curators (in the form of MTCR tokens) to validate and curate the whitelist
- Consumers use the end product, which is the whitelist and associated database of content
Selkis outlined how this model would work on the Epicenter podcast, in which he provided the example of university trustees (TCR discussion starts around the 30 minute mark). The trustees, or Curators, don’t review every application. Instead, they set the standards that must be met and applied by those at a lower level, the Validators. Again, I think it is broadly a good idea to have a committee setting and approving standards.
Selkis also noted that you have to control the initial distribution and raised the notion that 15–20 funds could be invited to act as curators. Messari themselves would also take a chunk of the supply. Other parties would include exchanges, advisory groups, and other accredited investors.
This raises a few issues. One, the advantages of having a distributed group is lost when the parties are so interwoven. 15–20 funds could and likely would quickly dovetail. Secondly, by having a “relatively illiquid” token to start with you lose some of the incentives for people to do the validation and curation. As an aside, and forgive me my cynicism, but I have seen enough so-called leading luminaries of crypto act in such a manner that I would have little faith in their role as arbiter of any list.
Rewards only flow to the curators and validators In such a case, the idea of a TCR almost becomes something of a tithe. If a group of funds is running the MTCR, it would be in their best interests to force as many projects as possible to apply to the MTCR, thus generating a nice source of stable profits. 100 projects a year at $25k a go is $2.5m, get that to 1,000 projects and it’s $25m.
As Selkis himself notes, Messari could build a lot of what they’re trying to build without the use of a TCR. The TCR is used to incentivize data completeness and avoid the incomplete/lag in updating data that afflicts the likes of Crunchbase. I’m not convinced the TCR offers sufficient rewards to achieve this though. Why could this not be charged for in a normal manner? The TCR is used to incentivize companies, not individuals, as far as I can tell — so if this information is so crucial would they not pay for it anyway?
What the TCR does is transfer the onus on payment for this information from funds doing diligence to the teams themselves. I’ll happily take the free information, but I’m not sure it’s the best model. If incentivization was used to get tens of thousands of contributors I would understand, but as analysts aren’t eligible for rewards and they only go to the curators and validators (professional firms), I’m not sure how this model of incentivization is really adding much beyond creating a token.
There are other issues beyond token distribution when considering launching. Lists aren’t going to launch with any inherent value because they won’t be of use to people. There are means around this, with established lists better placed to transition to a TCR model and thus do a form of Initial Token Offering (ITO) valuing their list at however many millions of dollars.
For example, I am sure that if the Michelin Guide decided to do an ITO, there would be restaurants trying to buy their way onto the list. I am also fairly confident that such a move would irreparably damage the iconic brand and, given Michelin hold a tight leash over listings, this is obviously not a move they would ever consider.
Secondly, the framing of the TCR is crucial. This can be changed over time, but by and large a curator at launch will have more ability to shape what the TCR demands than a late contributor (it is always harder to change something once initially set). This isn’t an issue as such, but TCRs should be mindful of the problems that initial ambiguity can bring (as Augur experienced).
The size of a list will also be important. I am unsure how something like Messari could cope if there were suddenly thousands of projects to review. I assume they would just scale up in a normal manner, but it’s arguably not as easy as other lines of business where there are defined contracts and business needs. Instead, there is simply an uptick in applications which curators can then delegate to the validators. Remember that validators must also constantly be checking all those applicants who have already been accepted to make sure they are still keeping to their prior submissions.
Being too small is as big a problem as too big. If there are no new applications then many TCRs will collapse, because there are none of the transaction fees that applications bring. As such there has to be a fine line trodden between admitting too many projects, and being so exclusive that applications dry up.
Finally, there are issues such as collusion and off-chain lobbying which are readily imaginable. Again, there is not much that can be done about this (which is why you want the tokens distributed as much as possible) but it’s fairly unavoidable. The best you can do is limit it.
Long road ahead
Similar to STOs, I think the absence of as much scrutiny that ICOs have faced means there is less criticism of them. ICOs themselves were long heralded as revolutionary forms of global equity raising that bypassed the legacy financial system. It is only now that they have become everyone’s favorite punching bag.
I actually think TCRs may work better when they are twinned with more basic exclusivity at an individual level and are based upon social capital rather than financial. This sees payment become the cost of keeping the TCR exclusive, but is a very limited use case. It may also work for hobbyists curating lists. Again, this is a form of social capital, rather than using them as the basis of creating value. I think TCRs need some form of identity/reputation accountability, to go along with their financial incentives.
The recent TCRParty popularity list on Twitter, which uses a TCR to form a list of trusted crypto personalities, should provide additional clues as to the problems TCRs face. My initial reaction was annoyance, but reading Alpine’s (the ConsenSys project behind it) comments were illuminating:
My hunch was that TCRs will end up being plutocracies, where the rich get richer and control the lists — Steve Gattuso, CoinDesk
CoinDesk noted that:
Both [Gregory] Rocco and Gattuso are skeptical about TCRs, concerned that they inevitably become plutocratic popularity contests where people amass disproportionate control through token holdings.
People often confuse negativity with dislike or some personal agenda. This is obviously a more negative take than a balanced one but it should be clear I’ve chosen what I think represent stronger examples. However, I also think TCRs face a significant amount of challenges and have more limited applicability than is often trumpeted.