Jump to content

AOWIS:AI Usage Guide/v1.0: Difference between revisions

From AOWIS
No edit summary
Line 1: Line 1:
{{DISPLAYTITLE:AI Usage Guide v1.0}}
{{DISPLAYTITLE:AI Usage Guide v1.0}}
{{DISPLAYTITLE:AOWIS AI Usage Guide (v1.0)}}
''Guidelines for controlled and verifiable use of AI in AOWIS.''
----
== Purpose ==
This document defines how Artificial Intelligence (AI) tools MAY be used in the creation, modification, and review of AOWIS documentation and system designs.
It establishes boundaries to ensure that all outputs remain verifiable, controlled, and compliant with AOWIS standards.
----
== Requirements ==
<pre>
REQ-AI-001: AI-generated content MUST be reviewed by a human before inclusion in any normative AOWIS document.
REQ-AI-002: AI MUST NOT be considered a source of truth for normative requirements.
REQ-AI-003: All normative statements MUST be validated against the AOWIS Writing Style Guide before acceptance.
REQ-AI-004: AI MAY be used to draft explanatory text.
REQ-AI-005: AI MAY be used to suggest normative statements, but such statements MUST be independently verified.
REQ-AI-006: AI-generated normative statements MUST be assigned unique requirement identifiers by a human.
REQ-AI-007: AI MUST NOT introduce normative requirements outside the "Requirements" section.
REQ-AI-008: AI-generated content MUST be checked for ambiguity, non-testable statements, prohibited terms, and mixed tone violations.
REQ-AI-009: AI MUST NOT be used as the sole mechanism for safety-critical design decisions.
REQ-AI-010: All safety-critical logic MUST be reviewed and approved by a qualified human reviewer.
REQ-AI-011: AI MAY be used to improve clarity, grammar, and structure of existing text, provided meaning is not altered.
REQ-AI-012: AI MUST NOT modify requirement semantics without explicit human approval.
REQ-AI-013: AI usage MUST be disclosed in the document revision history.
REQ-AI-014: AI-generated examples MUST be realistic and reflect low-infrastructure environments.
REQ-AI-015: AI MUST NOT generate or modify requirement identifiers without human verification.
REQ-AI-016: AI MAY be used for brainstorming, idea generation, and outline creation.
REQ-AI-017: AI MAY be used to correct spelling, grammar, and formatting.
REQ-AI-018: AI MAY be used to transform text into formal or structured language.
REQ-AI-019: AI-generated content MUST be reviewed, modified, and approved by a human before inclusion.
REQ-AI-020: Content MUST NOT be copied from AI outputs into the wiki without substantive human modification.
REQ-AI-021: A contribution MUST be considered non-compliant if it contains AI-generated content that has not been reviewed or modified by a human.
REQ-AI-022: Project maintainers MAY reject or remove contributions that do not meet REQ-AI-019 and REQ-AI-020.
REQ-AI-023: Repeated submission of non-compliant AI-generated content MAY result in loss of contribution privileges.
REQ-AI-024: All accepted content MUST be attributable to a human contributor who takes responsibility for its correctness.
</pre>
</pre>
----
== Definitions ==
'''Artificial Intelligence (AI)''' 
A software system capable of generating text, suggestions, or structured outputs based on input prompts.
'''AI-Generated Content''' 
Any text, structure, or data produced by an AI system.
'''Human Reviewer''' 
A person responsible for verifying correctness, compliance, and safety of content.
'''Normative Content''' 
Content that defines requirements using MUST, SHOULD, or MAY.
----
== Notes ==
AI can accelerate drafting and improve clarity, but it does not understand real-world constraints, failure modes, or safety implications.
In low-resource environments, incorrect assumptions can lead to system failure. Human oversight ensures that designs remain grounded in operational reality.
AI is most effective when used for:
* Structuring documents
* Rewriting unclear text
* Generating examples
* Identifying inconsistencies
AI is least reliable when used for:
* Safety-critical reasoning
* Context-specific engineering decisions
----
''End of AI Usage Guide (v1.0)''

Revision as of 18:23, 5 April 2026


Warning: Display title "AOWIS AI Usage Guide (v1.0)" overrides earlier display title "AI Usage Guide v1.0".

Guidelines for controlled and verifiable use of AI in AOWIS.


Purpose

This document defines how Artificial Intelligence (AI) tools MAY be used in the creation, modification, and review of AOWIS documentation and system designs.

It establishes boundaries to ensure that all outputs remain verifiable, controlled, and compliant with AOWIS standards.


Requirements

REQ-AI-001: AI-generated content MUST be reviewed by a human before inclusion in any normative AOWIS document.
REQ-AI-002: AI MUST NOT be considered a source of truth for normative requirements.
REQ-AI-003: All normative statements MUST be validated against the AOWIS Writing Style Guide before acceptance.
REQ-AI-004: AI MAY be used to draft explanatory text.
REQ-AI-005: AI MAY be used to suggest normative statements, but such statements MUST be independently verified.
REQ-AI-006: AI-generated normative statements MUST be assigned unique requirement identifiers by a human.
REQ-AI-007: AI MUST NOT introduce normative requirements outside the "Requirements" section.
REQ-AI-008: AI-generated content MUST be checked for ambiguity, non-testable statements, prohibited terms, and mixed tone violations.
REQ-AI-009: AI MUST NOT be used as the sole mechanism for safety-critical design decisions.
REQ-AI-010: All safety-critical logic MUST be reviewed and approved by a qualified human reviewer.
REQ-AI-011: AI MAY be used to improve clarity, grammar, and structure of existing text, provided meaning is not altered.
REQ-AI-012: AI MUST NOT modify requirement semantics without explicit human approval.
REQ-AI-013: AI usage MUST be disclosed in the document revision history.
REQ-AI-014: AI-generated examples MUST be realistic and reflect low-infrastructure environments.
REQ-AI-015: AI MUST NOT generate or modify requirement identifiers without human verification.
REQ-AI-016: AI MAY be used for brainstorming, idea generation, and outline creation.
REQ-AI-017: AI MAY be used to correct spelling, grammar, and formatting.
REQ-AI-018: AI MAY be used to transform text into formal or structured language.
REQ-AI-019: AI-generated content MUST be reviewed, modified, and approved by a human before inclusion.
REQ-AI-020: Content MUST NOT be copied from AI outputs into the wiki without substantive human modification.
REQ-AI-021: A contribution MUST be considered non-compliant if it contains AI-generated content that has not been reviewed or modified by a human.
REQ-AI-022: Project maintainers MAY reject or remove contributions that do not meet REQ-AI-019 and REQ-AI-020.
REQ-AI-023: Repeated submission of non-compliant AI-generated content MAY result in loss of contribution privileges.
REQ-AI-024: All accepted content MUST be attributable to a human contributor who takes responsibility for its correctness.

Definitions

Artificial Intelligence (AI) A software system capable of generating text, suggestions, or structured outputs based on input prompts.

AI-Generated Content Any text, structure, or data produced by an AI system.

Human Reviewer A person responsible for verifying correctness, compliance, and safety of content.

Normative Content Content that defines requirements using MUST, SHOULD, or MAY.


Notes

AI can accelerate drafting and improve clarity, but it does not understand real-world constraints, failure modes, or safety implications.

In low-resource environments, incorrect assumptions can lead to system failure. Human oversight ensures that designs remain grounded in operational reality.

AI is most effective when used for:

  • Structuring documents
  • Rewriting unclear text
  • Generating examples
  • Identifying inconsistencies

AI is least reliable when used for:

  • Safety-critical reasoning
  • Context-specific engineering decisions

End of AI Usage Guide (v1.0)