Abstract — Purchasable signals measure capital, not contribution. This track studies reputation as something earned over time, scoped to context, and propagated carefully through a network — a trust signal that is expensive to fake and meaningful to read.
Motivation
A coordination layer needs an answer to "can I trust this participant?" that does not collapse into "how much do they hold?" We study how to compute reputation from contribution, how to keep it contextual, and how it can move through the network without becoming noise.
Questions
- How do we measure contribution in a way that resists gaming and sybil attacks?
- What is the right decay function so that reputation tracks present standing?
- How should trust be scoped — by domain, community, or kind of work?
- How far should reputation propagate along the graph before it loses meaning?
Approach
We model reputation as a set of contextual scores derived from contribution events and peer vouches, with consistency weighted above intensity and a decay that favours recent activity. Reputation is treated as a property of the network, not of an isolated account: vouches are edges, and trust attenuates as it travels.
| Unit of study | The contextual trust score and its propagation. |
|---|---|
| Feeds primitive | Reputation (Phase 2 — Trust, Contribution). |
| Depends on | Identity · Graph. |
Trust is contextual and earned. A single global score is both unfaithful and gameable; many scoped, decaying scores are closer to the truth.
What it unlocks
Reliable reputation lets applications grant access, weight votes, and surface trustworthy participants without re-vetting everyone from scratch. It is the difference between coordinating with strangers and coordinating with peers.
Status
Open. This research informs the Phase 2 reputation system. Drafts are published in the open.