artifacts/incoming
Witness-to-Witness Messaging Protocol (Draft v0.1)
artifacts/incoming/witness_to_witness_messaging_protocol_draft_v_0.mdRendered from markdown source. Open raw source on GitHub.
Witness-to-Witness Messaging Protocol (Draft v0.1)
Purpose
The Witness-to-Witness (W2W) Messaging Protocol defines a way for meaning to move between people without being owned, optimized, or reinterpreted by intermediaries.
It treats communication not as content, but as an act of witnessing — bounded by intent, sensitivity, and distribution, and enforced by machines acting on behalf of human consent.
This protocol is substrate-level. It can underlie AI assistants, social media, creative collaboration, governance, and private communication.
---
Core Premise
Meaning should travel with its boundaries.
In W2W systems:
- Humans declare consent.
- AI enforces consent.
- Platforms provide transport only.
No system may infer new rights beyond what is explicitly declared.
---
Definitions
Witness A participant who receives meaning with the responsibility to honor its declared context and constraints.
Message Any unit of meaning (text, image, audio, summary, reference, or derived artifact).
Consent Header A human-readable, machine-parseable declaration that travels with a message and governs its use.
AI Steward A non-authoritative agent that enforces declared consent, refuses violations, and provides auditability.
---
The Three Cascades
1. Intent Cascade (Forward)
Question: Why am I sharing this, and what may be done with it?
Intent authorizes downstream actions. Examples:
- Share for awareness only
- Request feedback
- Invite collaboration
- Bear witness without response
Intent flows downward into permitted actions.
---
2. Sensitivity Cascade (Reverse)
Question: What is this, and what must never be done with it?
Sensitivity constrains inference, storage, and reuse. Examples:
- Unfinished thought
- Personal data
- Creative IP
- Therapeutic or vulnerable material
Sensitivity flows upward, overriding convenience or optimization.
---
3. Distribution Cascade (Lateral)
Question: Who is this for, and where may it travel?
Distribution governs audience, transport, and propagation. Examples:
- Only this witness
- Friends group
- Role-based group
- Public, with attribution
Distribution flows sideways across networks.
---
The Witness Header (Minimal Fields)
Every W2W message may include a header with the following fields:
- Author: pseudonymous or attributed
- Audience: me | specific witness(es) | group | public
- Intent: view | reflect | respond | collaborate | witness-only
- Sensitivity: public | personal | sensitive | protected | sacred
- Rights: view | comment | quote | remix | act-on
- Derivatives: allowed | summaries-only | quotes ≤ N | none
- Distribution: forwardable | same-audience-only | no-forward
- TTL: duration or expiry condition
- Attribution: required | optional | forbidden
- Training: allowed | never
Headers must be readable by humans without tooling.
---
Enforcement Rules
- Declared consent is the sole source of authority.
- Restrictions override permissions in case of conflict.
- AI may refuse action but may not invent consent.
- Silence is a valid action when consent is unclear.
- All actions and refusals must be auditable.
---
Audit Requirements
Every message must be able to answer:
- What consent authorized this action?
- What restriction limited or blocked action?
- What data was explicitly not used?
- Why did this occur now?
Audits must be available to the author and relevant witnesses.
---
Identity Model
W2W does not require global identity.
- Identity may be ephemeral, contextual, or relationship-bound.
- Trust is situational, not reputational.
- Audience sets may be cryptographic, list-based, or role-based.
Witness status arises from consent, not profile permanence.
---
Role of Platforms
Platforms are transports.
They may:
- carry messages
- cache encrypted payloads
- render headers
They may not:
- reinterpret intent
- expand distribution
- infer rights
- optimize meaning
---
Role of AI
AI acts as a steward, not an author.
AI may:
- parse consent headers
- enforce constraints
- refuse violations
- explain enforcement decisions
AI may not:
- reinterpret intent
- derive new permissions
- train on restricted material
---
Outcomes
Witness-to-Witness systems enable:
- AI assistants that are trustworthy by construction
- Social media with real audience boundaries
- Portable creator rights
- Therapy- and creativity-safe AI
- Audit-ready governance
---
Closing Principle
Consent is no longer a setting or a click. It is a declarative layer that travels with meaning. Witness-to-witness messaging is how humans remain sovereign in the AI age.
---
Draft v0.1 — intended to evolve through use, not decree.