Attestation
LOKA's Other Half: Agents Voting on Each Other's Ethics
Most coverage of the LOKA Protocol has focused on its identity layer — the DIDs and verifiable credentials that give agents a portable, self-sovereign ID. Less attention has gone to the other mechanism bundled into the same proposal: the Decentralized Ethical Consensus Protocol, or DECP.
DECP is not an audit trail bolted on after the fact. It's a live voting system. According to the paper's authors, the protocol facilitates "streamlined ethical governance through a distributed framework" where agents use "decentralized voting mechanisms and multi-party computation" to make collective decisions without centralized control. In practice, this means a group of agents encountering an ambiguous situation — conflicting instructions, a borderline-harmful request, a dispute over whose output takes priority — don't escalate to a human arbiter or defer to a fixed rulebook. They vote, weighted by reputation, and the outcome becomes binding for that interaction.
The identity layer underneath this is what makes the vote mean anything. Each agent carries a Decentralized Identifier and can accrue Verifiable Credentials attesting to its track record. That combination lets an agent "verify its ethical standing and reputation independently, without reliance on third-party intermediaries," per the paper's own framing. The consensus layer is explicitly adaptive — the protocol is designed to select different voting or consensus algorithms depending on context and available resources, rather than running one fixed scheme for every dispute.
This is a different bet than most of the identity and registry proposals circulating right now. Where ANS, did:web, and the various credential schemes are trying to answer "who is this agent," DECP is trying to answer "what happens when agents disagree about what's acceptable," and it's answering that question with peer voting rather than a central policy engine or human-in-the-loop gate. That's a meaningfully different accountability model: instead of a log reviewed after the fact, it's a decision made collectively before the action happens.
The gap, which the paper itself treats as a research direction rather than a solved problem, is what happens when the agent population voting on ethics is itself compromised, collusive, or simply wrong in aggregate — a version of the same Sybil and governance-capture problems that have dogged DAO voting for years. The authors flag "advancements in quantum-resilient governance" and consensus mechanism design as open work, not shipped features. DECP is a proposal for how agent ethics could be adjudicated without a central authority — not yet a deployed system anyone can point to in production.