Commonsent Signal: from raw issue to earned understanding

The public square is not short on information. It is short on integrity, structure, and memory. Signal is the layer that supplies all three: it turns a raw civic issue into measured understanding and into credibility that has to be earned, one topic at a time, and kept. It does not tell you what to think. It makes the act of thinking inspectable.

Every other module in Commonsent assumes people can reason together well enough to decide. Signal is the module that makes that assumption true. It is the civic-intelligence division, and its job is narrow and load-bearing: take a contested public issue, measure how trustworthy the information around it actually is, help a person genuinely understand it, and record who earned the right to be taken seriously on it. Then keep doing all of that as the issue evolves, because a fact checked last year is not a fact checked today.

The design begins from a refusal that runs through the whole project. Signal does not moderate content and does not rule on truth. The moment a civic system appoints itself the arbiter of what is true, it becomes a target, gets captured, and loses legitimacy. So Signal does something more durable instead. It measures the integrity of the information, makes the reasoning visible, and ties credibility to demonstrated understanding rather than to volume, identity, or conviction. It is a referee for the process, never a judge of the verdict.

01The whole pipeline at a glance

Signal runs in two passes around every issue. Phase A, content intelligence, runs once per issue and is machine-assisted but human-audited: it analyzes the information environment. Phase B, user reasoning, runs once per reader: it helps an individual understand the issue and, if they choose, deliberate on it. Underneath both sit continuous accountability services that every outcome feeds, and a set of loops that make understanding compound and force credibility to be maintained rather than banked.

02Phase A: reading the information, not the issue

Before anyone forms a view, Signal reads the environment the view would form in. Three analyzers run, each machine-assisted and human-audited, and together they answer a question almost no platform asks: how trustworthy is the information around this issue, and who is trying to shape it.

The Publishing Layer turns the issue into a living page rather than a frozen article. It carries a dual summary, one plain and one technical, and breaks the issue into claim cards, discrete assertions that can be examined one at a time instead of a wall of prose that hides its own structure. Source Integrity performs cross-source verification and builds a conflict-of-interest graph, mapping who funds, owns, and benefits from each source so that undisclosed ties become visible as structure rather than rumor. Coordination Detection looks for the fingerprints of manufactured consensus: it estimates an astroturf probability from patterns of coordinated inauthentic amplification, and flags the manipulation triggers, the engineered urgency and emotional spikes, that signal an attempt to bypass reasoning rather than inform it.

These three converge into the analyzed issue page and its Civic Nutrition Label, the single most portable idea in Signal. Just as a food label does not tell you whether to eat something but tells you what is in it, the civic label does not tell you what to believe. It tells you what you are consuming.

Each gauge has a definition and a calculation. Factual density is the share of an issue's claims that are linked to evidence objects with quality markers, against all asserted claims. Sponsor transparency is the share of detected interests that are disclosed and verifiable. Conflict-of-interest risk reads the concentration of undisclosed or cross-cutting ties in the interest graph. Astroturf probability and manipulation risk come from the coordination analyzer. None is a verdict on the issue. Each is a measurable property of the information, and all of them are auditable.

03Phase B: helping a person actually understand

Analysis is wasted if people skim it and react. Phase B is built to slow the reflex and reward the work. It runs four steps for each reader.

Personal Alignment takes a person's own priorities and produces a personalized brief, the same analyzed issue framed around what that person actually cares about, so the entry point is relevance rather than a generic feed. The brief is deliberate about which kind of relevance it is claiming, because four are easy to confuse: emotional relevance, where an issue simply makes you angry; identity relevance, where your group is expected to care; personal relevance, where the outcome will materially change your household; and sponsor relevance, where someone is paying to make you care at all. Most manufactured engagement lives in the first two. The alignment brief is built to surface the third and expose the fourth, so attention can flow to what actually shapes a person's life rather than to whatever is loudest.

The Comprehension Gate is the step that makes Signal unlike anything in the attention economy. Before a person can enter the deliberation room or earn standing on an issue, they pass a short check, on the order of a minute, that confirms they have grasped the core claims and the strongest case on more than one side. Fail it, and the system routes you back to re-read rather than letting you weigh in. This is not a quiz for its own sake. It is a structural answer to the fact that most public argument is conducted by people reacting to headlines they did not read.

Most public argument is conducted by people reacting to a headline they did not read. The Comprehension Gate is a structural answer to that, not a moral one.

Deliberation follows for those who want it: a pre-survey of position, a short structured room of roughly fifteen minutes with assigned roles and an AI moderator bounded to clerical and mapping duties, and a post-survey. The deliberation engine is itself a full protocol, detailed in its own piece.¹ Outcome closes the loop for the individual: it records the alignment shift between pre and post, scores the calibration of any predictions the person made, and produces an AI summary that keeps facts and values in separate columns, so a reader can see exactly where they updated because of evidence and where they simply hold a different value.

04The Topic Credit Ladder

Here is the mechanism that replaces followers, likes, and blue checks. On any given topic, a participant occupies a level on a credit ladder, from zero to eight. Credit is earned, never granted, and it is strictly topic-specific: deep standing on housing policy buys nothing on vaccine safety. You climb by demonstrating comprehension, by making calibrated predictions that later prove accurate, and by contributing reasoning that participants in other camps judge fair. You do not climb by being loud, early, or popular.

The decisive feature is decay. Credit is not a trophy you keep. It falls over time as the issue evolves and as the facts move, and it is rechecked through spaced recall, the same way real understanding has to be refreshed or it fades. Standing earned on last year's version of an issue does not entitle you to authority on this year's. This single rule does something no reputation system on the internet does: it ties credibility to current, demonstrated understanding rather than to accumulated history or audience size.

The score is continuous, but it is easiest to read as nine rungs, each unlocking a little more responsibility on that one topic. A person can sit near the top on housing and at the bottom on foreign policy, and any rung is held only while the underlying credit is maintained, so decay can move someone back down.

05The accountability services that never sleep

Every deliberation outcome feeds three standing services, and these are where Signal stops being an analysis tool and becomes civic memory.

The Missing Evidence Market turns the gaps surfaced during analysis and deliberation into open bounties. When a load-bearing claim has no good evidence behind it, that absence becomes a visible, fundable request rather than a quiet hole that each new participant rediscovers. The Public Memory System is the antidote to the internet's amnesia: it tracks claims, promises, and contradictions across time, so that a prediction made confidently last year, or a promise made and quietly abandoned, stays attached to whoever made it. The Alignment Drift Monitor is the most personal and the most carefully bounded. Private to each user, it shows them the difference between learning and being moved: whether their shifts in position over time track new evidence, or whether they have been steadily pulled by repetition and pressure. It is a mirror the system offers the individual, never a score it shows anyone else.

06Beyond the core: the wider feature surface

The pipeline described so far is the core, but it opens onto a wider surface of features that run on the same machinery rather than beside it. Three have already appeared as the accountability services: the Missing Evidence Market, the Public Memory System, and the Alignment Drift Monitor. The rest extend the analyzers, the label, and the gate, and they are worth naming because together they are what separate Signal from a bias mirror.

None of these is a separate product. Each is the same commitment applied at a different point: make the information inspectable, make reasoning earn its standing, and keep the public memory from being wiped between news cycles.

07Why the loops matter

The arrows that double back in the diagram are not decoration. They are the reason Signal improves rather than ossifies. Outcomes from Phase B enrich the analysis of future issues: recurring evidence gaps, common misreadings, and detected manipulation patterns all feed back into how the next issue is published and labeled. Comprehension and credit decay force maintenance, so the system never lets a stale understanding pass for a current one. The two passes compound: the more issues run through Signal, the sharper its analysis and the more demanding its gate.

This is also exactly how Signal serves the rest of Commonsent. The coordination and governance layers are only as good as the quality of public reasoning feeding them. Signal is what makes the demand side worth coordinating: a population that has actually understood an issue, whose credibility reflects demonstrated reasoning, and whose collective memory cannot be wiped between news cycles. In the language of the keystone essay, Signal is a direct contributor to systemic health and to the epistemic quality term inside the objective the whole system optimizes for.²

08What Signal is not

Because the failure modes here are severe, the boundaries have to be stated as plainly as the features. Signal is not a censor; it removes nothing and ranks no one's speech. It is not a truth tribunal; it measures the integrity of information and the quality of reasoning, and it leaves the verdict to people. It is not a social credit score; its credit is topic-specific, earned through demonstrated understanding, decaying by design, and never a portable judgment of a person's worth. And it is not an engagement product; the Comprehension Gate, the decay, and the calibration scoring all actively work against the addictive loops that every attention platform is built to maximize. If Signal ever became sticky, sortable into universal scores, or a place that told people what to think, it would have failed its own thesis.