Methods, Protocols & Prompt Systems
Method profile 02 Source-pinned Interactive state model

Dialogue lifecycle

EXPLORE → CLARIFY → DECIDE → EXECUTE → VERIFY → CLOSE

A six-state control model for human-AI work. It keeps investigation, clarification, commitment, action, verification and disposition distinguishable so that a promising idea is not silently treated as an authorized instruction, and a completed action is not mistaken for a verified or properly closed result.

01

Artefact classification

The lifecycle, AICS contract and 5PP are related control surfaces, but they solve different problems.

Dialogue control

Six-state lifecycle

Controls when work may move from open inquiry to commitment, action, audit and durable closure.

Contract representation

AICS

Represents lifecycle phase, decisions, authorization, blockers, evidence and closure as a versioned instruction contract.

Execution discipline

5PP Variation A

Controls how the instruction envelope is clarified, locked, structured, executed and mechanically audited.

02

Interactive dialogue state machine

Drag the states, select a node and inspect its purpose, exit conditions, authority level and characteristic failure mode.

View
03

Synthetic transition-guard laboratory

Attempt a state transition and see which source-derived guards permit or block it. This browser model is educational and is not the AICS validator.

Synthetic contract conditions
Not evaluated

Transition result

Select a transition and evaluate its synthetic guards.

    TransitionEXPLORE → CLARIFY Computed statuspending
    04

    Dialogue lifecycle versus 5PP

    Two parallel control systems. Click either track to inspect the difference. The alignment bands are explanatory, not a normative one-to-one mapping.

    CLOSE is not a sixth 5PP phase. It is a separate disposition and durable-writeback state that consumes verified evidence and preserves residuals, deviations and re-entry conditions.
    05

    Observable lineage and representative adoption

    First-party dialogue provenance, formal AICS increments and selected operational projections. The graph does not claim to expose every internal use.

    06

    Evidence and formalization receipts

    Measured repository receipts are separated from analyst-coded evidence coverage.

    Measured repository receipt

    AICS test-suite growth

    AICS PR receipts report 5 passing tests for 0.1 and 15 passing tests for 0.2. These tests were not rerun for this publication.

    Registry coding, not efficacy

    Current evidence surface

    The profile measures available artefact evidence, not whether the lifecycle improves outcomes. No independent effect study was identified in the inspected source set.

    6

    Lifecycle states

    From open inquiry through durable disposition.

    5

    AICS profiles

    Core, conversation, repository, execution and full.

    15

    Tests at AICS 0.2

    Source-reported in the self-hosting increment.

    3

    Negative fixtures

    Duplicate ID, blocked execution and failed closure.

    1

    Explicit reopen operation

    REOPEN requires a decision and preserved history.

    0

    Independent effect studies

    None identified in the inspected source set.

    07

    Evaluation interpretation

    Mechanism and implementation evidence are explicit. Outcome efficacy remains an open empirical question.

    Problem definition

    Human-AI conversations often mix investigation, recommendation, authorization and action. That collapse makes it difficult to know whether a candidate was merely discussed or actually accepted for execution.

    Mechanism hypothesis

    Explicit state boundaries should make premature commitment, unauthorized execution, incomplete verification and silent closure easier to detect. This is a design hypothesis, not a proven universal effect.

    Operationalization

    AICS represents the current phase, blockers, accepted decisions, version-bound authorization, verification evidence and closure state in a schema-validated contract.

    Evaluation routes

    Matched-task studies could measure premature-action rate, unresolved blockers at execution, rework, verification coverage, operator effort, closure defects and recovery quality after reopening.

    Explicit non-claims

    • The dialogue lifecycle is not the same artefact as 5PP.
    • A visible state label does not prove the participants are cognitively operating in that state.
    • An accepted decision can still be wrong, poorly informed or unethical.
    • Authorization consistency does not substitute for domain, safety or legal review.
    • CLOSE does not mean every future question is resolved.
    • Reopening must preserve prior history rather than silently rewriting the closed state.
    • The earliest located conversation date is first-party provenance, not independent historical validation.
    08

    Representative project connections

    Selected implementations and operational projections, not a complete estate inventory.

    AICSCanonical formalization, schema, validator, self-hosting contract and round-trip projection. control-center-opsOperational governance, project-change protocol, runbooks and AICS contract use. SPINERepresentative project and agent workflow context in which dialogue modes and 5PP are used together. 5PP profileThe separate five-phase protocol that hardens and audits work inside the lifecycle. ChangePlaneRepresentative protocol-trace and decision-record operationalization. Adaptivearts.ai/blogAdjacent public context on structured AI work, SPINE and 5PP. No dedicated lifecycle-origin article is claimed.
    09

    Sources and reproducibility notes

    Claims are bounded to the inspected sources and provenance classes below.

    Canonical formalizationfbratten/aics @ cd88e41bcadcdc68155b384ecf723f3b91a80b18
    Normative lifecycleprotocol/lifecycle.md and SPECIFICATION.md
    Mechanical representationschema/aics.schema.json and tools/validate_contract.py
    First formal incrementAICS 0.1, 31 Jul 2026, 5 source-reported tests
    Self-hosting incrementAICS 0.2, 31 Jul 2026, 15 source-reported tests
    Operational projectioncontrol-center-ops project-change package @ b2f44ab5fc340317a30ac1fbe9d5aacd8b1df53f
    Private first-party provenanceearliest located conversation formulation: 11 Jul 2026; raw transcript not reproduced
    Publication surfacefbratten.github.io/methods/dialogue-lifecycle/

    This page performs no AICS validation, repository writes, LLM calls, private-memory access or external state transition. All interactive examples remain in browser memory.