Files
ben.stull 70c79c6768
regenerate-readme-index / noop (push) Has been cancelled
Add Identity and Relationality RFC super-drafts
Expand RFC-0001 §Identity and §Relationality into proposed super-drafts. Identity: persona, linkability/selective disclosure, plural-presentation but singular-standing. Relationality: the flow, contextual integrity, per-edge asymmetric consent. Both depend on Informed Consent; neither graduated.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-12 08:56:49 -07:00

30 lines
6.4 KiB
Markdown

---
slug: relationality
title: Relationality
state: super-draft
id: null
repo: null
proposed_by: ben.stull@wiggleverse.org
proposed_at: '2026-06-12'
graduated_at: null
graduated_by: null
owners:
- ben.stull@wiggleverse.org
arbiters: []
tags: []
---
RFC-0001 (Human) establishes Relationality as one of the five foundational properties of a Human — there is no Human that exists outside of relationship — and names it among the words still to be defined. It fixes four commitments this RFC must build on: that a Human is shaped by relationship; that relationships are of many kinds and resist a single canonical type; that relationships are asymmetric, so that two parties experience the same relationship differently and the model must not collapse it into a single shared object; and that Relationality is the medium of consent, since most of what a Human consents to, they consent to in relationship. Where the sibling RFC, Identity, governs what may be recorded *about* a Human, Relationality governs what may be shared *between* Humans — and between Humans and the systems and agents that act within those relationships.
The unit Relationality governs is the **flow**: a movement of information, recognition, or presentation from one party to another across a specific relationship. This RFC's central position is that a flow is governed by the norms of the relationship it occurs in, not by a global permission. Information a Human shares appropriately in one relationship — a purchase history with a merchant they chose, a confidence with a friend, a medical fact with a clinician — is not thereby available to flow into another; carrying it across contexts is a distinct act that the receiving context's norms must independently justify. A system that models consent as a single share / do-not-share switch, rather than as a property of each flow in its context, has already discarded the thing Relationality exists to protect. Appropriateness is contextual: the same disclosure that is ordinary in one relationship is a violation in another, and the difference is the relationship, not the data.
Because a relationship is not a single shared object, a flow is directional and per-edge. What A discloses to B is consented and scoped from A's side, for A's purpose, within A's relationship to B; B's reciprocal view, and anything B would carry onward, are their own flows under their own consent. The model must not represent a relationship as one mutual record both parties equally own, because that erases whose disclosure it was and whose authority governs it. This matters most where a relationship *produces* data — a transaction, a conversation, a shared history — over which each party holds a real but partial claim; the model must keep those claims distinct rather than dissolving them into common property that either party may treat as wholly theirs.
The question Identity calls linkability appears here as **recognition across relationships**: whether a Human known in one relationship is recognized as the same Human in another. Seen from the relational side, recognition is itself a flow — it carries a Human's presence from one context into a second — and is therefore governed by the same contextual norms and the same purpose-bound, revocable consent as any other flow. A Human may be recognized across two relationships they consent to bridge and remain unrecognized across two they do not, without inconsistency. The cross-context recognition that produces belonging, continuity, and earned trust between Humans is valuable precisely because it is consented; the identical mechanism, unconsented, is surveillance. The model must make the two distinguishable, and must put the choice between them with the Human.
RFC-0001 holds that Relationality is the medium of consent, and this RFC makes the dependency explicit: consent to a flow inherits Informed Consent's conditions — specific to a purpose, time-bound rather than perpetual, and revocable as a property of the Human rather than a favor of the system. Revocation raises the hardest relational question, which this RFC must answer rather than dodge: what is owed when a Human withdraws consent to a flow after information has already moved and been acted upon, including where it has propagated transitively — A shared to B, B relied on it, C now depends on B's reliance. The model owes at least that revocation reach forward to halt further flow, and that data already propagated be addressed rather than treated as permanently surrendered the instant it was first shared. What more is owed — recall, deletion, correction, or only cessation — is the question graduation must settle.
Relationality extends beyond the human. RFC-0001 holds that Humans are in relationship not only with other Humans but with communities, institutions, the non-human world, and the systems that mediate all of these. An agent acting on a Human's behalf is a relational participant, not a neutral pipe: what it may carry between Humans, and what it may be told about a Human in order to act, are relational flows governed here — in concert with Agency, which governs what may be done on a Human's behalf, and with the RFC on non-human entities. A relationship mediated by a machine does not shed its norms because a machine stands in the middle; the machine inherits them.
This RFC builds on RFC-0001 (Human), is the sibling of Identity, and depends on Informed Consent and Privacy & Data Minimization for the permission and the default-minimization each flow is judged against, and on Epistemic Authority for whose account of a relationship leads where the parties disagree. It conditions Participants (Customer, Merchant, and Operator are relational positions carrying built-in asymmetries of power), Reputation (vouching and endorsement are relational flows, and reputation moves across relationships), Trust & Safety, and Recourse & Dispute (a contested flow needs a path to remedy). Relationality is not a single shared object the platform owns, not a global consent flag, and not a license to move what was shared in one context into another. Open questions it will have to resolve as it graduates: how a flow's scope and purpose are represented so that a machine can honor them; how a contextual-integrity violation is detected and adjudicated; how joint data with competing claims is governed; how far revocation can reach over data already propagated; and how transitive flows are bounded, so that consent granted in one relationship does not silently authorize a chain of others.