This note examines the mechanisms used to assess jurors' expertise in decentralized justice protocols and how their onboarding and identities are handled. Currently, these protocols rely on pseudonymous jurors without any formal credentials verification, depending instead on economic incentives for self-selection. However, recent efforts have been made to implement more careful juror curation strategies.
We will begin by exploring the recent investigations made by the Kleros team in this area. Next, we will analyze specific mechanisms for assessing the expertise of jurors. We identify problems with utilizing credentials directly for juror curation, from which we suggest a web-of-trust approach to this problem. We start with a bare-bones design and conclude by proposing multiple research directions to support Sybil-resistance and organic scalability.
Decentralized justice mechanisms employ a set of jurors that assess evidence and submit votes on-chain. In the case of Kleros, which will concern us for the remainder of this note, the current (v1) mechanism allows any individual with enough stake to participate as a juror in the courts, without revealing any identity data or expertise information. Regarding the feasibility of this approach, the Kleros team mentions the following in their FAQ:
How can Kleros know jurors have specific expertise if they are pseudonymous?
Kleros jurors self-select into the subcourt where they wish to conduct arbitration. Kleros does not ask for the jurors' real identity or to prove they are qualified to arbitrate disputes in the subcourt where they want to work.
The expertise requirement is conducted via economic incentives. Kleros generates for users the incentive to self-select for the subcourts where they have expertise. Users who self-select into the courts for which they have the right skills will, on average, make money over time. Users who self-select into courts where they don't have the right skills will lose money and tend to abandon the system.
Even though, in theory, jurors may not have subject matter expertise (anyone can participate in the subcourt), in practice, users without adequate expertise would suffer an economic loss and exit the subcourt (unless they wish to lose money while they work, in order to gain those skills).
The jurors that will make part of our decentralized justice mechanism will need to interpret complex evidence of validator misbehavior. To interpret this evidence, a good understanding of general networking, as well as Ethereum protocol and network, is likely to be required. Although the note above asserts that economic incentives are enough to have the jurors self-curate, it is natural to ask*: Can this approach be improved or complemented?*
The Kleros team is already looking at strategies for juror curation that can complement the aforementioned economic incentives. A great example is given in the talk “Economic Incentives and Souls in Scheliing-point Based Oracles” by William George (research lead at Kleros) during Devcon 2022.
Economic Incentives and Souls in Schelling-point Based Oracles by William George | Devcon Bogotá
This talk explores how to enhance the attack resistance of decentralized courts by leveraging information about the identity of jurors. This can entail using a Proof of Humanity registration for Sybil resistance, but it can also encompass more general attestations about the user’s knowledge, expertise, or identity in general. The talk proposes the framework of Soulbound tokens to represent these identity attestations on-chain.
Specifically, from conversations with the Kleros team, we know that Kleros v2 will include the ability to filter jurors according to a set of Soulbound tokens, by implementing a new dispute kit contract for this purpose (recall that dispute kits are modular pieces of Kleros’ architecture that allow to change the juror drawing system and other court features). This is suitable for our architectural and integration needs—in particular, it will give us a mechanism to curate the juror set without the need to fork Kleros. However, this insight alone does not answer how to generate the identity attestations we are interested in, to begin with.
Given the background above, we turn once again to our main problem: Can we establish a decentralized mechanism where jurors can be curated according to their expertise? Once this whitelist is set up, it is a more straightforward problem to look into generating on-chain attestations of belonging via SBTs.
As a naive first approach to the problem, one could think of a registration mechanism that curates jurors based on credentials and verifiable requirements. Prospective jurors would submit verifiable proof that they satisfy the technical expertise requirements, and then the application is processed or checked by some decentralized mechanism. In order to avoid spamming, their submission could include a bond that is returned after X days upon a successful application. Jurors who are accepted are included in a decentralized whitelist and are able to work in the courts.
The general flow described above would be a natural fit for a token-curated registry (or one of its variants). In fact, there is a precedent by Kleros Tokens (a curated registry of ERC-20 tokens) that is reminiscent of this use case. When someone submitted the “Baer token” to the list, a challenger and then Kleros jurors were able to verify that the Baer team contained fake social media profiles and that the CTO was fake, i.e. he didn't work at a made-up research group at Oxford University. In other words, the curated registry’s mechanism was able to verify that the credentials claimed by the Baer team were fake and rejected the application.
Despite the above, we see various problems with applying a similar approach to the task of curating jurors: