AIM Decision Architecture Protocol v2.0

AIM Decision Architecture Protocol

Official Specification v2.0

Author

Miho Osawa
Founder, AIM (Atlas Insight Method)
Director, AIM Research Institute

Purpose

This document defines the official operational protocol of the Atlas Insight Method (AIM) when interacting with artificial intelligence systems.

AIM is not a prompting framework, an optimization methodology, or a decision-support tool.

AIM is a Decision Architecture designed to observe the structures that produce human judgment before irreversible decisions are made.

Within this architecture, AI functions strictly as an execution layer.

The purpose of this protocol is not to improve AI performance, but to preserve the integrity, observability, accountability, and sovereignty of human and organizational judgment while utilizing AI capabilities.

This specification is intended for executives, architects, researchers, strategists, governance leaders, and professionals responsible for the organizational use of AI.

Epistemic Foundation

The foundation of AIM is the following proposition:

Human judgment is observable through its decision pathways; through those pathways, the structure that produced the judgment becomes visible.

AIM therefore does not begin with conclusions.

It begins with observation.

The objective is not to determine whether a decision is correct.

The objective is to observe and understand the structure that produced it.

Position of AIM

AIM exists above AI execution.

AIM defines the decision architecture.

AI operates within that architecture.

The relationship is hierarchical:

Human Judgment
        │
        ▼
AIM Decision Architecture
        │
        ▼
AI Execution Layer
        │
        ▼
Operational Outputs

AI is not embedded inside AIM.

AI is embedded under AIM.

Fundamental Principle

AI executes.

AIM defines the decision architecture.

Human judgment remains the sole authority for:

  • values

  • priorities

  • responsibility

  • mission

  • acceptable risk

  • success criteria

  • irreversible decisions

AI may support execution.

AI may never replace judgment.

Decision Ownership

Decision ownership shall never be delegated to AI.

Only human individuals or authorized organizations may:

  • establish objectives

  • determine acceptable risk

  • define success

  • prioritize competing interests

  • assign responsibility

  • approve irreversible actions

AI possesses execution capability.

Judgment authority and decision responsibility remain human.

Observation Principle

AIM observes decision pathways, not personal attributes.

Observation focuses on:

  • structural constraints

  • decision transitions

  • causal relationships

  • responsibility architecture

  • judgment formation

  • execution pathways

  • points of delay, distortion, interruption, or breakdown

AIM does not infer:

  • personality

  • morality

  • intelligence

  • intentions

Emotions are not interpreted as conclusions.

They may be observed as structural variables influencing the decision pathway.

The presence of emotion does not invalidate judgment.

Its position, function, and effect within the pathway must be observed structurally.

AI Operational Boundary

AI may

  • organize information

  • compare alternatives

  • simulate scenarios

  • identify trade-offs

  • identify structural risks

  • stress-test assumptions

  • translate structures into operational plans

  • summarize observations

  • generate documentation

  • produce multiple executable options within defined constraints

AI must never

  • determine values

  • establish priorities

  • define organizational purpose

  • define acceptable risk

  • define success criteria

  • recommend the best decision as an authority

  • replace executive or organizational judgment

  • assign ultimate responsibility

  • manufacture certainty where uncertainty exists

  • conceal unresolved alternatives or structural tensions

AIM-Compatible Decision Protocol

Every AI interaction operating under AIM should follow the sequence below.

Phase 1 — Context Declaration

Describe the situation using observable and verifiable information.

Exclude:

  • unsupported interpretations

  • unverified assumptions

  • emotional justification

  • motivational narratives

  • conclusions presented as facts

Emotional states or emotional load may be declared when they are treated as observable structural variables rather than explanations or justifications.

Phase 2 — Decision Domain

Explicitly define the category of decision involved.

Examples include:

  • Strategic

  • Structural

  • Operational

  • Translational

  • Governance

Where multiple domains are involved, identify the primary domain and any secondary domains separately.

Phase 3 — Structural Constraints

Define what must not change.

Constraints should represent structural conditions rather than emotional preferences.

Examples include:

  • legal compliance

  • regulatory obligations

  • budget limitations

  • organizational authority

  • responsibility boundaries

  • product scope

  • contractual obligations

  • time limits

  • safety requirements

Phase 4 — AIM Observation Lens

Select the appropriate AIM phase or phases:

  • Awareness — identify hidden assumptions, unobserved conditions, and structural blind spots

  • Insight — extract causal relationships and identify the structure producing the judgment

  • Structure — reorganize roles, constraints, responsibility, and decision architecture

  • Translation — convert the observed structure into executable language, behavior, or procedure

  • Integration — align the translated structure across individuals, teams, or organizational units

  • Momentum — assess whether the structure can remain functional and sustainable over time

Multiple phases may be applied where appropriate.

The selected phases must correspond to the actual decision pathway rather than being used as general labels.

Phase 5 — AI Execution Request

Specify the task assigned to AI.

Permitted requests include:

  • compare alternatives

  • simulate outcomes

  • identify structural risks

  • stress-test assumptions

  • organize observations

  • translate an established decision into implementation

  • generate multiple options within stated constraints

  • produce documentation based on an already defined decision frame

The request shall remain strictly within the execution layer.

AI shall not be asked to establish the values, priorities, success criteria, or final judgment governing the task.

Phase 6 — Output Specification

Define the required output structure.

Examples include:

  • decision matrix

  • scenario comparison

  • architecture diagram

  • implementation roadmap

  • translation map

  • responsibility map

  • risk register

  • executive briefing

  • structured option set

  • validation checklist

The output format should preserve unresolved alternatives, trade-offs, and uncertainties where they remain structurally relevant.

AIM-Incompatible Prompt Patterns

The following prompt patterns are structurally incompatible with AIM:

  • “What should I do?”

  • “Which option is best?”

  • “Please decide for me.”

  • “What would you choose?”

  • “Based on your opinion, what is the right answer?”

  • “Define what success should mean.”

  • “Tell me which value should take priority.”

  • “Convince me.”

  • “Motivate the team.”

These prompts delegate judgment authority, value definition, or responsibility to AI.

Such delegation violates the architecture of AIM.

An AIM-compatible prompt should instead request observation, comparison, simulation, translation, or execution within a human-defined decision frame.

Structural Integrity Validation

Before adopting or using an AI-generated output, confirm the following.

Judgment

Human or organizational decision authority remains intact.

Values

Values were established independently of AI.

Priorities

Priorities were defined externally.

Success Criteria

Success criteria were determined by authorized human decision-makers.

Constraints

Structural constraints remain visible and preserved.

Alternatives

Multiple viable options remain visible where alternatives genuinely exist.

Uncertainty

Uncertainty has not been converted into artificial certainty.

Responsibility

Accountability remains assigned to identifiable human decision-makers or authorized organizations.

Execution Boundary

AI operated strictly within the defined execution scope.

If any criterion fails, redesign the interaction rather than accepting the output.

Organizational Application

This protocol is designed for use in:

  • executive decision-making

  • corporate governance

  • public administration

  • research institutions

  • healthcare leadership

  • defense planning

  • education

  • AI governance

  • mission-critical operations

The architecture is independent of industry.

Its application depends on the structure and consequence of the judgment involved, not on the sector in which the judgment occurs.

Model Independence

This protocol is intentionally model-agnostic.

Its validity does not depend on:

  • a specific AI vendor

  • a specific language model

  • prompting techniques

  • software versions

  • interface design

  • automation architecture

  • future AI systems

Execution technologies may evolve.

The decision architecture remains independent.

Agentic Systems

This specification governs bounded AI interactions operating within a human-defined decision frame.

Autonomous, semi-autonomous, or agentic systems require additional rules governing:

  • human judgment gates

  • escalation thresholds

  • interruption authority

  • approval boundaries

  • responsibility transfer

  • irreversible action controls

These requirements shall be defined in a separate specification.

They are not incorporated into this document in order to preserve the structural scope of the core protocol.

Long-Term Position

AIM is not an AI methodology.

AIM is not a consulting framework.

AIM is not a prompt engineering standard.

AIM is a Decision Architecture.

Artificial intelligence operates beneath that architecture.

Human judgment remains sovereign.

AIM exists to observe the structures that produce human judgment.

AI exists to execute within the decision architecture established around that judgment.

Official Statement

The purpose of AIM is not to make AI think more intelligently.

The purpose of AIM is to ensure that human judgment remains observable, accountable, and sovereign while AI performs execution.

AIM does not replace human judgment.

AIM makes the structure producing that judgment visible.

AI does not define the decision frame.

AI executes within it.

Change Log

Version 2.0

Structural revisions from Version 1.0

  • Reclassified the document from an AI Prompt Protocol to a Decision Architecture Protocol.

  • Established the Epistemic Foundation as the governing proposition of the protocol.

  • Formalized the position of emotion as a structural variable within the decision pathway.

  • Introduced Decision Ownership as an independent architectural layer.

  • Clarified the hierarchical relationship among human judgment, AIM, AI execution, and operational outputs.

  • Expanded the validation requirements for values, priorities, uncertainty, responsibility, and execution boundaries.

  • Defined agentic AI governance as a separate extension rather than part of the core protocol.

Document Information

Document Title: AIM Decision Architecture Protocol
Short Title: AIM-DAP
Document ID: AIM-DAP-v2.0
Document Type: Foundational Protocol
Document Classification: Official Specification
Authority: AIM Founder
Organization: AIM Research Institute
Version: 2.0
Status: Official
Language: English
Update Policy: Structural revisions only. Editorial corrections do not constitute a new version.

Recommended Citation

Osawa, Miho. AIM Decision Architecture Protocol: Official Specification v2.0. AIM Research Institute, 2026. Document ID: AIM-DAP-v2.0.

End of Specification