Last year, the Bounties Network team spent some time researching decentralized arbitration schemes, attempting to solve the difficulties that most marketplaces face around providing fair arbitration for users, which takes into account the changing contexts of the transactions they facilitate.
Out of this sprung Delphi, a mechanism for p2p dispute resolution that had users staking tokens and opening claims against each other. Most importantly, rather than those claims being ruled upon by us as a marketplace operator, they would instead be adjudicated by a group of 3rd party individuals, whose membership would be curated by a TCR. Using a TCR to curate a list of trustworthy arbiters for a particular set of domains made sense — the subjective nature of the evaluation necessitates a human touch. However, as time went on (and I talked to Dean Eigenmann more), it became increasingly clear that we shouldn’t be relying on TCRs as a finished component that didn’t require further scrutiny.
While we’ve since de-prioritized our work on Delphi for the immediate term, we feel compelled to share where we left off in our exploration, lest another brave mechanism designer wants to pick up the torch. The key, we believe, is self curation.
Self Curating Groups
One of the biggest issues we had with TCRs was based on the notion that the curation of the list would be undertaken by a group of token-holders, with no guarantees of honesty, trustworthiness, or even that the token holders knew anything about the contents of the list being curated. This made no sense to us — why would we blindly trust the evaluations of strangers without the ability to contextualize their evaluations based on their levels of understanding and bias? TCR1.0-maximalists will typically assert that the profit motive is strong enough to incentivize individuals into rationally evaluating their biases and understanding, to only vote when they felt they would improve the quality of the list. But are humans rational? How effectively do we evaluate our own biases and understanding towards a given topic?
In response to this, our thinking changed dramatically: what if instead of assuming that the curators and listees were independent groups (with some overlap), we treated them as one and the same?
When we look at how individuals organize into groups in the real world, we witness the natural processes of curation through invitation, exclusion, and expulsion happening organically by members of the group itself, rather than some external deciding body. While centralized planning and coordination produce efficiency benefits when that body can be guaranteed to be of a high quality, the efficiency is lost if just anyone can become a member of the governing body simply by purchasing tokens on the open market.
Humans organically self-organize all the time, usually when high quality individuals coagulate to separate themselves from their peers, seeking differentiation in an increasingly homogenous world. They evaluate their peers, and gravitate towards those within or above their quality bracket. When they encounter new individuals, they’ll invite those who meet the quality thresholds of the group. When members of the group misbehave, they’ll get ostracized when high quality members don’t want to associate with them anymore. Similarly, members of the group will seek to differentiate the group from those outside of it, emphasizing how members of the group differ in quality from the general population.
We witness and participate in this kind of curation on a daily basis, when we choose which friends to spend time with in our personal lives, or which organizations we should work with in our professional lives.
Listees are Domain Experts
Within an ecosystem, there will be certain metrics or values with which members are evaluated for their quality: among ballerinas it might be their flexibility or artistry; among restaurants it might be the speed of their service or quality of ingredients. Members of the ecosystem will know these metrics well — these are the same metrics they will use when optimizing their own performances. They will also be experienced in evaluating their peers using these metrics, as they constantly compare themselves to others in their race towards relative excellence.
As a result of this, it seems self-evident that the individuals which should be most trusted to curate a list are those within it, who already possess the skills necessary to curate within a particular domain. This is contrary to TCR1.0s, which desire for “the crowd” to manifest their wisdom, even when the crowd has no understanding or experience working within it.
Are people all equally fit to make various evaluations, or does our suitability depend on our knowledge and experience within a particular domain?
One of the elements which was recently included within the Moloch DAO specification was the notion that to become a token-holder and voter, one needed to apply to purchase shares, allowing existing token-holders to curate the set of individuals who would be voting to administer it’s funds. Individuals apply to purchase a given amount, and are accepted if the group feels comfortable giving that individual their desired voting weight over the funds.
While this wasn’t a detail we’d considered previously, it’s certainly interesting — should listees have equal weight in their curation voting?
Once again looking at natural systems, the answer seems to be no: we constantly give more credence to the opinions of our most highly regarded peers. Our stronger reliance on expert opinions is the reason why social media influencers succeeded in drawing a crowd to the ill-fated Fyre Festival, and is also the reason we prefer political pundits who have themselves worked within the sphere that’s being discussed.
By following in the footsteps of the Moloch DAO and having list applicants self-specify their desired voting weight (or changes to their weights if they’re already members), the group itself can determine whose voices should be the loudest. Ranking list members by their voting weight would likely also merit some interesting insights, for those who would wish to understand the breadth of quality within the list (derived as a result of the difference in voting weight between the highest and lowest ranked members), or simply the ordering of members.
However, if TCRs were to implement such a mechanism, they would differ substantially from organic vote-weighting due to the explicit nature of the weight specification: in natural systems, actors will weigh each others’ opinions based on their personal opinions of them, which may change when they become exposed to the public (or when they can see how heavily others weigh an individual’s opinions).
Finally, and in my opinion most importantly, self curating groups allow for healthy competition.
One of the main issues with TCRs was the high overhead in launching a competing TCR: not just the gas or technical costs of supporting a new system, but the high social costs required for people to begin coordinating around a newly minted token. These barriers to entry and competition meant that most TCR designers expected for there to be just one list for a particular domain, rather than a plurality of competing lists.
Are monopolies healthy?
Because of the way self-curated lists simplify the social constructs of TCRs, the overhead of starting a new list becomes much lower, improving the efficiency of the system when those who were wrongfully rejected from one list may start their own. Of course these lists will still suffer the consequences of being less relied upon, however this can easily be fixed by having TCR-consuming-systems respond to the creation of new lists and support a plurality of lists rather than just one.
When we once again look at natural systems, we see that in the face of inconsistent curation (where suitable members are unfairly excluded from a list), individuals will instead opt to create their own groups, sometimes including those members of the original list who agreed with their entry. From the playground to the fashion world, this kind of competition is the way that insecure incumbents are commonly disrupted.
When we think about the lists we rely upon as consumers, a number stand out that help us make decisions about where or how to spend our time and money. One such example is winners of The Oscars, the unquestionably most important curation that occurs in the film ecosystem. And by whom is this list curated? By a collection of past nominees (listees) and individuals who have been sponsored by existing members of The Academy, to be curated even further by The Academy’s Board of Governors.
Self-curating lists are all around us, and their existence rests on their simplicity. They don’t rely on the procurement of a 3rd party who should be incentivized to curate; instead they use the intrinsic incentives of list members to only associate with their highest quality peers to improve their own reputations.
In light of this we’re forced to ask: if curated registries exist in the natural world without their own tokens, why do they need a token to function now?
As the pace of Ethereum development continues to accelerate and teams continue to explore curation mechanisms for their various use cases, my hope is that more teams will continue to explore how we can construct crypto-systems which mirror the natural world, rather than eschewing the patterns that exist within our common social behaviours. If you are working on a TCR related use-case, please feel free to reach out and join our slack (and chat with us about these concepts at length), or evolve my nascent perspective into a more fully-formed Self Curating Lists concept.
As with most of my opinions, this blog post necessitates a disclaimer that my perspective on this topic is unrefined and unfinished — my hope is simply that by sharing it we can help accelerate the pace of TCR development towards an end which, to me, appears inevitable.