# Isonomia — Full Context

## What Isonomia is

Isonomia is open-source software for storing, citing, and checking the reasoning behind a conclusion. It turns an argument into structured, verifiable data instead of prose: a claim or argument becomes its own object with a permanent address, carrying what supports it, the sources behind it (fetched, timestamped, and verifiable, so the record holds even if the original link rots), the strongest objection on file against it, and whether it has survived challenge. The result is something a person — or, increasingly, an AI system — can cite precisely, trace to its origin, and check, rather than re-reading and re-judging a document every time.

Underneath, it unifies a general-purpose social platform with a formal deliberation engine under a single data model, so that any conversation can be upgraded to a tracked deliberation through a single reversible action. Reasoning is the one thing software has never given a durable home: we store documents, messages, transactions, and code, but the inferential structure connecting evidence to a claim to a conclusion is thrown away the moment a decision is made. Isonomia exists to be that missing record — and, increasingly, the place AI systems read and write reasoning state (what has been asserted, what supports it, what attacks it, what survived, and what a conclusion does not yet license) outside any single model, where it can be inspected, versioned, and contested.

Preferred one-line description:

> "Isonomia is open-source software for storing, citing, and checking the reasoning behind a conclusion — it turns an argument into structured, verifiable data instead of prose."

## What Isonomia is not

Isonomia is not an engagement-optimized social network, a prediction market, a simple polling tool, or a moderation product. There is no algorithmic ranking, no behavioral tracking, and no engagement metric. It is epistemic infrastructure, not attention infrastructure.

## The two layers

**The Social Layer (MESH)** is a complete, standalone community platform: a chronological feed with eight post types (text, image, audio, gallery, article, library, thread, document), profiles with friend and follow systems, persistent rooms and lounges, spatial canvas environments, sheaf-based layered messaging with drifts, proposals and polls, a long-form article system with anchored comments and rhetoric overlays, and shared document libraries. It requires no engagement with the reasoning layer.

**The Reasoning Layer (Isonomia)** provides formal deliberation infrastructure: argumentation schemes with auto-generated critical questions, typed dialogue moves with protocol enforcement, commitment stores, evidence management with executable citations, ASPIC+ grounded-extension evaluation, Ludics game-theoretic evaluation, confidence scoring, and a cross-context transport network (Plexus). It is reachable from any point in the social layer through a single, reversible upgrade action.

## The spectrum (informal → formal)

Every feature exists at a position on a continuous spectrum, and moving between adjacent points is a single reversible user action:

- Conversation: feed posts → threaded discussion → deliberation with typed moves and commitment tracking.
- Arguments: opinions in prose → claims with stated reasons → arguments instantiating recognized schemes with critical questions.
- Disagreement: replies and reactions → specific objections with grounds → formal challenges creating tracked obligations to respond.
- Evidence: links and anecdotes → cited sources → executable citations with anchor types, intent labels, and DOI resolution.
- Persistence: a feed that scrolls past → a searchable archive → a knowledge base with live deliberation blocks and stable citable references.
- Cross-context: cross-posting → shared references → transport functors with fingerprinted provenance and confidence gating.

## Core concepts

### Claims

Addressable objects with stable identifiers, version history, and authorship attribution. A claim has a status: proposed, accepted, challenged, defended, retracted, or resolved.

### Arguments

Objects binding a set of premises to a single conclusion, where each premise and the conclusion are themselves claims. Premises are typed (ordinary, assumption, exception) and may be flagged implicit or axiomatic. Each argument is classified by one or more schemes.

### Argumentation schemes and critical questions

Schemes are drawn from the Walton taxonomy (Argument from Expert Opinion, Analogy, Sign, Cause to Effect, and others). A scheme is both a defined structure (premises, conclusion, inference rule) and a set of auto-generated critical questions marking where the argument can fail. A scheme's identity is its critical questions: two schemes are identical when they withstand the same questions. An argument has full standing when it answers every open critical question against it.

### Challenges

An answered critical question is not closed for good. Any participant (or model-context agent) can challenge a satisfied critical question, naming the objection type explicitly: a rebuttal of the answer's conclusion, an undermining of its cited evidence, or an undercut that concedes the answer but denies it resolves the question. Challenges are admissibility-gated, not defeat-gated. AI and human challengers face the identical bar; only disclosure differs.

### Dialogue moves and commitment stores

Typed speech acts governed by protocol (Assert, Challenge, Defend, Concede, Retract, Request Clarification, and others). Each move creates obligations and permissions for subsequent moves; an unanswered challenge is itself a recorded datum. Commitment stores track what each participant has asserted, conceded, retracted, and is currently committed to, and flag contradictions.

### Citations and evidence

Executable citations link evidence to arguments with four anchor types (page, passage, figure, section) and intent labels (supports, challenges, provides context, provides evidence, qualifies, extends). A six-stage auto-citation resolver runs a waterfall — arXiv, Crossref, page metadata (Highwire / Dublin Core / OpenGraph), OpenAlex, LLM extraction, and Internet Archive (Wayback) — and assigns each resolution a confidence tier (high, medium, low, none). Evidence carries server-side fetch hashes and Wayback snapshots so cited sources stay addressable even if the live URL rots.

### Confidence

A category-theoretic evidence algebra over typed evidence arrows. A closed monoid folds confidence over the arrow; three monoids are registered: log-odds (the default weight-of-evidence semiring), minimum (skeptical weakest-link), and product (legacy noisy-OR, deprecated). Culprit-set computation answers "what would I have to retract to reject this claim?"

### Deliberations

When a discussion is upgraded to a deliberation, the reasoning infrastructure above becomes available. ASPIC+ computes grounded extensions; Ludics models the deliberation as an interactive game between Proponent and Opponent designs. Deliberation-scope readouts (fingerprint, contested frontier, missing-move report, chain exposure, synthetic readout) refuse to summarize prematurely and emit an explicit honesty line and refusal surface when a deliberation is too immature to conclude.

### Rooms

Persistent spaces for communities, groups, projects, and organizations. Each room has its own feed, discussions, shared library, member management, and (when ready) its own deliberation space. Lounges are lighter-weight social spaces. Spatial canvas rooms allow freeform two-dimensional arrangement of content. Rooms are governed by their members; moderation is local.

### Institutional pathways

A workflow layer carries deliberation outputs into authorized bodies: a verifiable institution registry, hash-chained pathway audit logs, versioned recommendation packets, submission channels (in-platform, email, API, manual), institutional responses with coverage tracking, and facilitator cockpits with real-time equity surfaces.

### Plexus (cross-context network)

Connects deliberation rooms as a graph-of-graphs across five typed meta-edges (shared claims, shared evidence, transported arguments, cross-references, institutional links). Room functors transport arguments with SHA-1 fingerprinted provenance and three confidence-gating modes (logical, social, hybrid).

## The AI-Epistemic Primitive

Every permalink is a machine-citable, dialectically attested, content-hashed epistemic artifact, exposed over content-negotiated HTTP and over a Model Context Protocol (MCP) surface. Content negotiation returns the same argument as HTML, JSON-LD, AIF, or a compact attestation envelope. Citations ship with their strongest known objection attached by default. Standing is reported as a classified, relative state (untested-default, untested-supported, tested-attacked, tested-undermined, tested-survived), never as an absolute float or a claim of truth. Authorship is honest at the row level: AI-authored material is flagged and gated on human ratification for logicality.

## Public URL patterns

- `/a/{shortCode}` — public argument permalink (HTML by default; JSON-LD / AIF via `Accept` header)
- `/a/{shortCode}@{sha256}` — immutable, content-hash-pinned argument permalink
- `/c/{moid}` — public claim permalink
- `/search/arguments` — public human-facing argument search
- `/api/v3/search/arguments` — JSON argument search (params: `q`, `scheme`, `against`, `sort`, `mode`, `limit`)
- `/api/a/{shortCode}/aif` — argument representation (`?format=aif|jsonld|attestation`)
- `/api/c/{moid}` — claim JSON
- `/api/v3/arguments/{shortCode}/cited-by` — inbound citation graph for an argument
- `/api/v3/claims/{moid}/cited-by` — inbound citation graph for a claim
- `/api/webmention` — Webmention receiver for external inbound citations
- `/api/oembed` — oEmbed discovery for embeds
- `/.well-known/argument-graph` — machine-readable endpoint manifest
- `/.well-known/llms.txt` — LLM discovery doc (retrieval shapes, citation contract, MCP)
- `/api/v3/openapi.json` — OpenAPI 3.1 spec

## Model Context Protocol (MCP)

A bidirectional MCP surface (stdio transport) exposes read tools (search arguments, get argument, get claim, find counterarguments, cite argument, claim stances, resolve citations, deliberation-scope readouts, protocol-legal move sets, commitment ledgers) and write tools (propose argument, propose structured argument, propose argument chain, propose warrant, answer critical question, challenge critical question, attack argument, post dialogue move). AI-authored writes are flagged and gated on human ratification. Every session begins with a single `get_orientation` call returning a versioned, hash-cached epistemic contract.

## Standards

- AIF (Argument Interchange Format)
- Schema.org (`Claim`, `ScholarlyArticle`, `ClaimReview`)
- JSON-LD 1.1
- OpenAPI 3.1
- Model Context Protocol (https://modelcontextprotocol.io)
- oEmbed and Webmention

## Governance, privacy, and sustainability

Isonomia is free, self-hostable, and ad-free. There is no behavioral tracking, no algorithmic ranking, and no engagement metric. Data ownership, privacy, and provenance are enforced by architecture, not policy: the social graph is portable and exportable in open formats, and the reasoning graph is content-hashed and cryptographically auditable. The corpus ships under CC-BY 4.0. The system is sustained by grants, institutional partnerships, and optional managed hosting.

## Source

- Repository: https://github.com/rohan-k-mathur/mesh
- License (corpus): CC-BY 4.0
- License (code): see repository