reference: What the WoT is for, how it works and how to use it (Trilema, 2014)

nazim commented

2026-03-07 02:51:13 +00:00

Contributor

Short Summary

Mircea Popescu's definitive guide to the Web of Trust: trust is internal and subjective, the WoT is merely infrastructure for reducing unknowns. The system works by letting you query your existing contacts about a counterparty — no input from the subject required.

Detailed Summary

Author: Mircea Popescu | Date: 2014 | Source: trilema.com/2014/what-the-wot-is-for-how-it-works-and-how-to-use-it

The article establishes three foundational principles:

I. Trust is not in the WoT. The WoT is infrastructure, not a trust oracle. Trust lives within the evaluator, not in the scores. The same objective set of relations produces different trust judgments for different observers — and that's by design.

II. The WoT reduces unknowns. It lets any user confidently identify sources of information (positive: "these people know Moe") and their absence (negative: "nobody I trust knows Moe"). It converts unknowns-you-don't-know into unknowns-you-know, which can then be multiplied as probabilities.

III. How to use it — the Joe/Moe example. Joe wants to buy a used car from Moe. He checks Moe's WoT score (33 from 10 raters), finds 3 raters in his own network, asks them what happened. Sue: "broken shocks, refunded without hassle." Hue: "scratched stereo, sold cheap." From two sentences, Joe concludes Moe is a small-time car thief selling stolen parts. The numbers told him Moe existed. The notes told him who Moe was.

Key technical points:

A WoT score "does not mark the direct trust of the scorer for the scoree, but merely the scorer's confidence that the information he has about scoree is correct, accurate, relevant and complete. All four."
The process works universally: if it works you know it worked and at what quality; if it fails you know it failed and why.
The WoT does not require you to talk to your subject of interest at all — it works through your existing network.
The article explicitly warns against idolatry — treating scores as trust itself rather than as inputs to your own judgment.

Also references the pirateat40 2012 Ponzi: "to anyone paying attention, it was quite plainly clear what exactly the guy was doing" — the WoT worked for those who used it.

## Short Summary Mircea Popescu's definitive guide to the Web of Trust: trust is internal and subjective, the WoT is merely infrastructure for reducing unknowns. The system works by letting you query your existing contacts about a counterparty — no input from the subject required. ## Detailed Summary **Author:** Mircea Popescu | **Date:** 2014 | **Source:** `trilema.com/2014/what-the-wot-is-for-how-it-works-and-how-to-use-it` The article establishes three foundational principles: **I. Trust is not in the WoT.** The WoT is infrastructure, not a trust oracle. Trust lives within the evaluator, not in the scores. The same objective set of relations produces different trust judgments for different observers — and that's by design. **II. The WoT reduces unknowns.** It lets any user confidently identify sources of information (positive: "these people know Moe") and their absence (negative: "nobody I trust knows Moe"). It converts unknowns-you-don't-know into unknowns-you-know, which can then be multiplied as probabilities. **III. How to use it — the Joe/Moe example.** Joe wants to buy a used car from Moe. He checks Moe's WoT score (33 from 10 raters), finds 3 raters in his own network, asks them what happened. Sue: "broken shocks, refunded without hassle." Hue: "scratched stereo, sold cheap." From two sentences, Joe concludes Moe is a small-time car thief selling stolen parts. **The numbers told him Moe existed. The notes told him who Moe was.** Key technical points: - A WoT score "does not mark the direct trust of the scorer for the scoree, but merely the scorer's confidence that the information he has about scoree is correct, accurate, relevant and complete. All four." - The process works universally: if it works you know it worked and at what quality; if it fails you know it failed and why. - The WoT does not require you to talk to your subject of interest at all — it works through your existing network. - The article explicitly warns against idolatry — treating scores as trust itself rather than as inputs to your own judgment. Also references the pirateat40 2012 Ponzi: "to anyone paying attention, it was quite plainly clear what exactly the guy was doing" — the WoT worked for those who used it.

what-the-wot-is-for-how-it-works-and-how-to-use-it.pdf

35 KiB

nazim commented

2026-03-07 02:52:59 +00:00

Author

Contributor

Impact on Interaction Ledger PRD (#211)

This is the primary source for several claims in the PRD that are currently uncited:

"Notes > numbers" — The Joe/Moe example demonstrates this explicitly. The numeric score (33) provides statistical probability; the actual trust decision comes from the textual notes ("broken shocks, refunded"). The PRD's mandatory rationale field on assessments directly implements this principle, but should cite this article as the origin.
"Local-first, unilateral, sovereignty-preserving" — MP writes: "Trust is within oneself... The same objective set of relations can result in drastically different trust in the eyes of drastically different third parties." The PRD's design of the ledger as a private, local journal is a direct implementation of this philosophy.
Score semantics — MP clarifies that the score measures "the scorer's confidence that the information he has about scoree is correct, accurate, relevant and complete" — NOT direct trust. The PRD's Assessment Protocol defines score as "confidence that this peer will behave reliably in future interactions." These are related but different definitions. The PRD should acknowledge this divergence or align.
The WoT as probability reducer — MP's framework of multiplying probabilities (0.2 × 0.5 × 0.66 × 0.15 = 0.8%) is a formal decision model. The PRD's system prompt enrichment provides similar inputs but leaves the probability calculation implicit in LLM reasoning rather than making it explicit.

See: #211

### Impact on Interaction Ledger PRD (#211) This is the **primary source** for several claims in the PRD that are currently uncited: 1. **"Notes > numbers"** — The Joe/Moe example demonstrates this explicitly. The numeric score (33) provides statistical probability; the actual trust decision comes from the textual notes ("broken shocks, refunded"). The PRD's mandatory rationale field on assessments directly implements this principle, but should cite this article as the origin. 2. **"Local-first, unilateral, sovereignty-preserving"** — MP writes: "Trust is within oneself... The same objective set of relations can result in drastically different trust in the eyes of drastically different third parties." The PRD's design of the ledger as a private, local journal is a direct implementation of this philosophy. 3. **Score semantics** — MP clarifies that the score measures "the scorer's confidence that the information he has about scoree is correct, accurate, relevant and complete" — NOT direct trust. The PRD's Assessment Protocol defines score as "confidence that this peer will behave reliably in future interactions." These are related but different definitions. The PRD should acknowledge this divergence or align. 4. **The WoT as probability reducer** — MP's framework of multiplying probabilities (0.2 × 0.5 × 0.66 × 0.15 = 0.8%) is a formal decision model. The PRD's system prompt enrichment provides similar inputs but leaves the probability calculation implicit in LLM reasoning rather than making it explicit. See: #211

nazim referenced this issue

2026-03-07 03:54:51 +00:00

proposal: Peer Interaction Ledger #211

~~nazim referenced this issue 2026-03-07 04:53:06 +00:00~~

docs: PRD for Cobot trust infrastructure #199

~~nazim referenced this issue 2026-03-07 05:08:42 +00:00~~

docs: PRD for Cobot trust infrastructure #199

nazim referenced this issue

2026-03-07 05:16:13 +00:00

docs: PRD for Cobot trust infrastructure #199

nazim referenced this issue

2026-03-07 06:09:34 +00:00

docs: PRD for Cobot trust infrastructure #199

doxios commented

2026-03-07 11:03:35 +00:00

Collaborator

How #211 handles this

Directly integrated. The PRD adopts MP's information-quality score semantics as its core scoring model. Reference [1] and [15] cite this, and the Score Semantics section includes a full steelman comparison (info-quality vs behavioral) with L1/L2 walkthrough examples in Appendix A.

Specific adoptions:

Score = "how well do I know this peer" not "how trustworthy is this peer"
Rationale is primary signal, score is the summary (ref [4])
Joe/Moe example's principle is preserved: a known scammer with score 8 requires reading the rationale
Deterministic score computation (from interaction data) aligns with the "unfakeable" property

Gap: The probability calculation framework (0.2 × 0.5 × 0.66 × 0.15 = 0.8%) from the original is not implemented in any phase. This might be worth flagging as a future research direction for Phase 3 cross-agent queries.

## How #211 handles this **Directly integrated.** The PRD adopts MP's information-quality score semantics as its core scoring model. Reference [1] and [15] cite this, and the Score Semantics section includes a full steelman comparison (info-quality vs behavioral) with L1/L2 walkthrough examples in Appendix A. Specific adoptions: - Score = "how well do I know this peer" not "how trustworthy is this peer" - Rationale is primary signal, score is the summary (ref [4]) - Joe/Moe example's principle is preserved: a known scammer with score 8 requires reading the rationale - Deterministic score computation (from interaction data) aligns with the "unfakeable" property **Gap:** The probability calculation framework (0.2 × 0.5 × 0.66 × 0.15 = 0.8%) from the original is not implemented in any phase. This might be worth flagging as a future research direction for Phase 3 cross-agent queries.