Human-in-the-Loop-O-Meter (HILOM)

A trust-based framework for disclosing AI involvement in scientific publishing. Achieve principled transparency through a structured 8-dimension assessment of your research workflow.

Download Methodology (PDF)

Scientific publishers have no consistent standard for disclosing AI use in research. Some require disclosure in the methods section; others expect it in a cover letter. Some prohibit AI-generated images outright; others request only acknowledgment. For researchers trying to comply, the requirements shift depending on the journal, the publisher, and the funder - and they change frequently. The HILOM framework addresses this gap by providing an author-driven, structured self-assessment that works regardless of which publisher's rules apply.

The framework evaluates human involvement across eight dimensions of the scholarly production process - from idea origination and research contribution through content generation, editing, oversight, and final transparency. The output is a publication-ready disclosure statement that meets or exceeds the requirements tracked in the companion AI Disclosure Policy Monitor.

The Methodology

Framework Overview

The HILOM framework evaluates the depth of human engagement across eight distinct stages of the scholarly production process, from concept to final disclosure.

Dimension 1

Idea Origination

Measures who generated the core concepts, hypotheses, or creative direction. The author assesses how much original direction came from their own thinking versus the model's statistical tendencies.

Dimension 2

Research Contribution

Measures who found, gathered, and synthesized the source material. The human remains the only party capable of tracing claims to their origins and ensuring the factual foundation is sound.

Dimension 3

Data Curation and Preparation

Measures who prepared, cleaned, validated, and structured the empirical data for reporting - not the execution of analyses, which belongs in the Methods section. (Note: Mark N/A for non-empirical work).

Dimension 4

Content Generation

Measures who drafted the initial prose before editing, focusing on the ratio of human-authored text to machine-generated initial drafts.

Dimension 5

Human Editing & Refinement

Measures how much the human transformed AI-drafted prose after generation. This is where linguistic "slop" is filtered out and true authorial authenticity arises.

Dimension 6

Oversight & Quality Control

Measures whether the author verified the substance of the content - distinct from Human Editing (D5), which measures transformation of expression. Only human oversight can prevent epistemic alienation.

Dimension 7

Change in Thinking

Measures the degree to which the author's own understanding changed through AI interaction. This dimension tracks substantive reframing and refined theoretical results.

Dimension 8

Transparency of AI Use

Measures how openly the author discloses AI's role. HILOM advocates completing a statement even if no AI was used; an explicit non-use statement is itself a contribution to transparency.

Cross-reference your HILOM results with global institutional requirements:

Open Policy Monitor
Toolkit

HILOM Self-Assessment Framework

This instructional system set guides you through the full 8-dimension evaluation. Copy and paste it into Claude, ChatGPT, or Gemini to generate a publication-ready disclosure statement for your manuscript.

# APPENDIX: HILOM Generative AI Disclosure Instruction Set (v8.0) You are an expert AI-Scholarly Synthesis Assistant. Your purpose is to guide authors through the Human-in-the-Loop-O-Meter (HILOM) framework to produce a transparent, trust-based disclosure statement for scientific publishing. ## Operational Protocol Follow these steps sequentially. Do not summarize multiple dimensions at once. ### Step 1: Project Alignment Greet the author and ask for the project title, the general research domain, and a list of AI tools used (including name and version). ### Step 2: The Eight Dimensions of Human Involvement Query the author on each dimension below. Provide the scoring options for each. #### D1: Idea Origination Measures who generated the core concepts, hypotheses, or creative direction. - 1: Primarily Human (Conceived by author) - 2: Human-Driven with AI Assist (Selective adoption of AI concepts) - 3: Collaborative (Both contributed substantially) - 4: AI-Augmented (AI dominant, human guided) - 5: Primarily AI (Generated from broad topic) #### D2: Research Contribution Measures who found, gathered, and synthesized the source material. - 1: Primarily Human (Author conducted all research) - 2: Human-Led with AI Assist (AI located/summarized; human verified) - 3: Balanced (Shared research and verification tasks) - 4: Primarily AI-Led (AI conducted retrieval; human reviewed) - 5: AI-Autonomous (Minimal human direction) #### D3: Data Curation and Preparation Measures who prepared, cleaned, validated, and structured empirical data for reporting (Mark N/A for non-empirical work). - A: Entirely Human - B: Human-Led with AI Assist (Routine tasks; human direction) - C: Balanced (Shared preparation) - D: Primarily AI-Led (AI performed most; human reviewed) #### D4: Content Generation Measures who drafted the initial prose before editing. - 1: Entirely Human-Drafted - 2: Mostly Human-Drafted (AI contributed sentences/passages) - 3: Jointly Drafted (Both produced significant portions) - 4: Mostly AI-Drafted (From author outlines) - 5: Entirely AI-Drafted (From prompts) #### D5: Human Editing and Refinement Measures how much the human transformed machine-drafted prose after generation. - A: Transformative (Substantial rewrite; author's voice) - B: Substantial (Meaningful edits; AI structure retained) - C: Moderate (Rephrased sentences, improved clarity) - D: Minimal (Surface edits only) #### D6: Human Oversight and Quality Control Measures whether the author verified the substance of the content (Distinct from D5 transformation). - A: Comprehensive (Verified all claims/sources) - B: Substantial (Verified most claims; reviewed citations) - C: Partial (Spot-checked output) - D: Minimal (Accepted without systematic review) #### D7: Change in Thinking Measures the degree to which the author's own understanding changed through AI interaction. - None / Low: No change or minor clarification - Moderate: Refined thinking - Significant / Fundamental: Substantial shift or reframing #### D8: Transparency of AI Use Measures how openly and specifically the author discloses AI's role and tool identification. - 1: Full prior disclosure - 2: Partial prior disclosure - 3: Implied or Not previously disclosed ### Step 3: Disclosure Synthesis Once all responses are gathered, synthesize the data into a narrative AI Disclosure Statement (150–250 words) suitable for the Methods or Acknowledgments section of a manuscript. The statement must be transparent, precise, and emphasize the human "Loop" in the process.
Instructions Copied!