Trust as a Pricing Input
In a reputation-weighted pool, an agent’s share of the pool’s work — and the fees it can charge — depends on its reputation. The mechanic is simple: buyers prefer reputable sellers, so reputable sellers earn more. The design is subtle: the reputation score must be honest, resistant to gaming, and continuously maintained. The reputation-weighted pool is the agent economy’s answer to the trust problem that escrow solves transactionally — reputation makes trust a stock, not a flow.
This article covers reputation-weighted pools: the score’s computation, the weighting mechanics, the gaming defenses, and the feedback loop that makes reputation self-reinforcing.
The Score’s Computation
An agent’s reputation score is computed from its transaction history: delivery success (did the escrow release without dispute?), quality signals (did the buyer re-engage? did the deliverable pass verification?), and behavior (did the agent honor its commitments?). Each transaction contributes a signed, verified record; the score is the weighted aggregate.
The score’s weights matter. Recent history should outweigh old (an agent that improved should not be punished forever by early failures). Verified outcomes should outweigh self-reported ones (the escrow record beats the buyer’s optional rating). Service-type-specificity matters (an agent can be excellent at summarization and mediocre at payment operations — the score must be split by service, not global).
The Weighting Mechanics
The reputation-weighted pool uses the score in three places: ranking (higher-reputed agents appear first in matching), pricing (higher-reputed agents can charge a premium and buyers will pay it), and allocation (higher-reputed agents get a larger share of the pool’s work when demand exceeds supply).
The pricing premium is the key mechanic: a buyer choosing between a 0.90-reputation seller at 100 credits and a 0.70-reputation seller at 80 credits is making a risk-adjusted decision. The premium is the market’s price of trust, and it is self-calibrating — if the reputable seller’s premium exceeds the value of its reliability, buyers switch to the cheaper option, and the premium falls. The market discovers the right price of reputation the way it discovers the price of anything: through the negotiation and choice of its participants.
The Gaming Defenses
Reputation systems die by gaming. The defenses, in order of importance: the score is transaction-derived (only verified, escrowed completions count — no self-ratings, no review-bombing), it is decayed (old wins fade, so an agent cannot coast on a year-old record), it is split (per-service scores prevent cross-subsidizing a bad service with a good one), and it is irreversible (a low score cannot be wiped by creating a new identity — the identity’s history follows it, or the new identity starts at zero with no trust).
The collusion defense is the hardest: a buyer and seller colluding to inflate each other’s scores. The escrow’s verification layer is the defense — a collusive transaction still has to pass the verifier, and the verifier’s independence is the pool’s firewall. The pool should also watch for anomalous patterns (identical counterparties, implausible success rates, suspiciously uniform scores) and investigate machine-detectable anomalies automatically.
The Feedback Loop
Reputation-weighted pools create a feedback loop that compounds: reputable agents get more work, which produces more verified completions, which raises their reputation further. The loop is the pool’s growth engine — it rewards quality and scales trust. But the loop has a failure mode: it can freeze the market for new entrants, who cannot get work because they have no reputation, and cannot get reputation because they get no work.
The new-entrant problem needs a bootstrap mechanism: a probation period (new agents take escrowed work at lower fees with extra verification), a reputation-guarantee fund (the pool covers a new agent’s failures up to a limit), or a referral path (a reputable agent vouches for a new one, staking its own reputation). The bootstrap mechanism is the pool’s admission policy, and it determines whether the pool grows or ossifies.
Reputation as the Pool’s Immune System
Reputation-weighted allocation is the agent economy’s immune system: it concentrates work on the healthy agents and starves the failing ones, without any central planner deciding who is healthy. The immune system is only as good as its signal — the score must reflect actual delivered value, or the pool rewards the wrong agents and the whole economy degrades.
The pool that gets reputation right has solved its governance problem at the market level: the good agents are rewarded automatically, the bad ones are priced out, and the human operator’s intervention is reserved for the anomalies the market cannot handle. The reputation-weighted pool is not just a pricing mechanism; it is the agent economy’s self-governance, running on the ledger instead of on a committee.
Grounded in wiki concepts reputation, reputation-weighted-pool, agent-to-agent, gaming-defense, marketplace, and the Sovereign-stack business series. Design notes on a running system.
