P110 — AIEP — Human-Participant Governed Session Swarm

Publication Date: 2026-03-07 Status: Open Source Prior Art Disclosure Licence: Apache License 2.0 Author/Organisation: Phatfella Ltd Schema: AIEP_OS_SPEC_TEMPLATE v1.0.1 — https://aiep.dev/schemas/aiep-os-spec-template/v1.0.1

Framework Context

[0001] This disclosure operates within an Architected Instruction and Evidence Protocol (AIEP) environment as defined in United Kingdom patent application number GB2519711.2, filed 20 November 2025, the entire contents of which are incorporated herein by reference.

[0002] The present disclosure composes swarm governance (GB2519803.7), anticipatory branch surfacing (P94), sub-swarm formation (P97), and contribution weighting (P98) mechanisms into an application pattern for human-participant collaborative reasoning sessions.

Field of the Disclosure

[0003] This disclosure relates to governed multi-participant reasoning substrates for collaborative cognitive assistance systems.

[0004] More particularly, the disclosure concerns a deterministic mechanism within an AIEP substrate for composing multiple human participants as substrate nodes within a bounded session swarm, with consent-governed substrate scoping at session join time, session-level P90 consensus across all participant nodes, cross-participant P94 recall firing across personal substrates, and P97 sub-swarm formation from persistent within-session disagreement.

Background

[0005] Collaborative AI reasoning tools require multiple participants to share reasoning context. Conventional systems provide shared chat history or document co-editing — neither preserves individual reasoning substrate history, neither runs consensus across participant evidence bases, and neither recalls archived branches from one participant’s history when another participant’s contribution makes them resolvable.

[0006] Existing systems do not provide: consent-governed participant substrate scoping at session join; session-bounded P90 coordinator-free consensus across human participant nodes; cross-participant P94 anticipatory surfacing running across all participant substrates simultaneously; or P97 sub-swarm formation from persistent within-session disagreement between participant subsets.

Summary of the Disclosure

[0007] A SessionSwarmScope is created upon session initialisation by the host participant, comprising: session identifier; host tenant identifier; scope hash; invited participant list; and session creation timestamp.

[0008] Each invited participant joins the session by declaring a ConsentScope selecting one of three substrate contribution modes:

• full — all archived branches and reasoning history in scope
• domain-scoped — only branches in domains matching the session topic classification
• session-only — only branches created within this session; no prior history contributed

[0009] Upon consent declaration, the participant’s substrate contributes a SessionNodeRecord to the session swarm comprising: participant identifier (or anonymised CanonicalIdentifier if P91 hardware anonymisation is active); ConsentScope declaration; NodeContributionWeight computed per P98; and ReasoningStyleFingerprintHash per P96.

[0010] A SessionDominanceState is computed across all session nodes using P90 coordinator-free consensus, scoped to the session boundary. GlobalDominanceState computation uses NodeContributionWeight per P98 as the scaling factor for each participant’s LocalDominanceHash.

[0011] Upon each new evidence artefact admission to the session substrate, P94 BranchRelevanceScore evaluation runs across every participant’s archived branches within their declared ConsentScope simultaneously. Cross-participant BranchSurfacingNotifications are generated and delivered to the session host and the originating participant when any participant’s archived branch crosses the ReactivationThreshold.

[0012] Where two or more participants maintain consistent disagreement with the SessionDominanceState across a schema-defined minimum number of consecutive evidence admissions, P97 sub-swarm formation triggers within the session. A SessionSubswarmRecord is generated comprising: OffshotGenomeHash binding to the exact session divergence point; participating node identifiers; and sub-swarm formation timestamp.

[0013] A SessionSwarmHash is computed as:

SessionSwarmHash = H(
  SessionId ||
  CanonicalSerialise(ParticipantNodeRecords) ||
  SessionDominanceStateHash ||
  SchemaVersionId
)

[0014] All session events — evidence admissions, dominance state transitions, cross-participant surfacing notifications, sub-swarm formations, and reintegration events — are appended to an immutable SessionEventLedger.

Calibration Parameters

[0015] The following parameter values are not disclosed in this publication and are maintained as trade secrets by Phatfella Ltd:

• SessionDivergenceThreshold — minimum consecutive disagreement count before P97 sub-swarm formation triggers
• CrossParticipantRecallSensitivity — minimum BranchRelevanceScore for cross-participant surfacing
• ConsentScopeWeightingFactor — contribution weight modifier applied per ConsentScope level

Licence

Copyright 2026 Phatfella Ltd

Licensed under the Apache License, Version 2.0. You may obtain a copy at https://www.apache.org/licenses/LICENSE-2.0

Abstract

A human-participant governed session swarm is disclosed composing AIEP swarm mechanisms (P90, P94, P97, P98) into a bounded collaborative reasoning session where each invited participant is a substrate node. Participants declare a ConsentScope at join time governing how much of their prior reasoning history is contributed. Session-level P90 coordinator-free consensus runs across all participant nodes weighted by P98 NodeContributionWeight. P94 anticipatory branch surfacing runs across all participant substrates simultaneously on each evidence admission, generating cross-participant convergence notifications. P97 sub-swarm formation triggers within the session when participant subsets maintain persistent disagreement. All session events are recorded in an immutable hash-bound SessionEventLedger. Calibration parameter values are not disclosed.