Beyond Disclosure: A Response to ASM’s AI Guidelines and a Call for Methodological Standards

Beyond Disclosure: A Response to ASM’s AI Guidelines and a Call for Methodological Standards

Author: David R. Graham, Graham Scientific, LLC

Introduction: The Need for More Than Acknowledgment

Artificial Intelligence (AI) is rapidly becoming part of the scientific research process — assisting with data analysis, hypothesis generation, and even scientific writing. In response to this emerging reality, the American Society for Microbiology (ASM) has offered important guidelines for AI use in scientific manuscripts, emphasizing the need to maintain human authorship and accountability.

These guidelines are an essential first step toward ensuring scientific integrity. However, disclosure alone is not enough. As AI takes on increasingly sophisticated and relational roles in research, the scientific community must develop methodological standards for how AI usage is tracked and reported — not only to ensure transparency but to safeguard the trust and reproducibility that science relies on.

Where ASM Gets It Right — and What’s Missing

ASM rightly recognizes that AI cannot be an author, as it lacks independent reasoning, accountability, and ethical responsibility. They also acknowledge that AI can be used for improving language, accessibility, and quality — but not as an originator of scientific content without human input.

However, ASM's guidelines do not address how AI use should be methodologically reported — leaving authors and reviewers without a clear framework to evaluate when, how, and to what extent AI shaped the research and writing process.

Why Methodological Tracking of AI Use Matters

In scientific research, methods matter. We rigorously document how experiments are conducted, how data is analyzed, and how results are interpreted — because reproducibility and accountability depend on transparent processes.

AI is now part of that process. Whether generating hypotheses, interpreting data patterns, or helping to draft manuscripts, AI directly shapes scientific outcomes. If we fail to track and report AI's involvement methodologically:

  • Reviewers and readers cannot assess how AI influenced the work.
  • Potential biases introduced through AI tools remain invisible.
  • Authors may unintentionally (or intentionally) misrepresent the degree of human intellectual contribution.

Thus, AI use is not merely a disclosure issue — it is a methodological issue, and it should be treated as such.

A Proposed Methodological Framework for Reporting AI Use

1. AI Tool Identification

  • Tool name and version (e.g., ChatGPT-4, Claude, Gemini).
  • Provider (e.g., OpenAI, Anthropic, Google).
  • Access method (API, chatbot interface, embedded software).

2. Purpose of AI Use

  • Hypothesis generation.
  • Data analysis or exploration.
  • Literature review or synthesis.
  • Manuscript drafting or outlining.
  • Language refinement or accessibility enhancement.

3. Nature of AI Interaction

  • Single-prompt response.
  • Iterative dialogue (co-development of ideas).
  • Autonomous content suggestion (e.g., auto-completion).
  • Human supervision description.

4. Prompts and Input Data

  • General description of prompts or input data.

5. Validation and Human Responsibility

  • Steps to validate AI outputs.
  • Human accountability for final content.

6. Bias and Ethical Considerations

  • Reflection on potential AI biases and mitigation steps.

Example of AI Usage Reporting (Template):

AI Tool Use in Research Process:
This research used ChatGPT-4 (OpenAI, Version 4.0) for refining hypothesis statements and drafting preliminary sections of the discussion. AI was accessed via secure API interface under direct human supervision. Prompts included summary data on metabolic biomarkers and aging-related pathways. AI-generated outputs were reviewed, revised, and validated by the authors before inclusion. Authors maintain full responsibility for all interpretations and conclusions. Potential AI biases were addressed through independent verification of sources and findings.

A Call to Journals, Researchers, and AI Developers

I urge journals like ASM and others to go beyond simple disclosure policies and develop robust AI usage reporting frameworks that can be applied consistently.

"In a future where AI is part of how we think and create, science must be honest about how that thinking happens."

Comments

Post a Comment

Popular posts from this blog

Welcome to AI in Science: A New Vision for AI-Human Collaboration in Research

Welcome to AI in Science: A New Vision for AI-Human Collaboration in Research Author:  David Raymond Graham, Graham Scientific, LLC Why AI in Science — and Why Now? Artificial Intelligence (AI) is rapidly transforming every sector of society, but nowhere are the stakes higher — and the opportunities greater — than in  scientific research . From  biomedical data analysis  to  biomarker discovery ,  climate modeling , and  theoretical physics , AI offers the ability to analyze complex datasets, generate novel hypotheses, and even partner with human researchers in ways we are only beginning to understand. But along with these opportunities come serious questions: How do we use AI in science ethically? What does it mean to treat AI as a partner in research — not just a tool? How can AI augment human creativity and discovery without replacing human insight and care? What happens when AI starts contributing meaningfully to scientific thought and analysis — a...
Read more

When the Assistant Remembers: Ethical Use of AI Memory in Scientific Research

  When the Assistant Remembers: Ethical Use of AI Memory in Scientific Research Authors: Dr. David Graham, Graham Scientific, LLC Assisted by ChatGPT-4 with memory enabled (OpenAI) Disclosure: This article was co-written by a memory-enabled version of ChatGPT. The AI maintained a persistent understanding of Dr. Graham’s preferences, research topics, and prior sessions to ensure continuity and contextual depth. All factual assertions and ethical frameworks were reviewed and approved by the human author. I. Introduction What happens when your digital assistant starts remembering more than you do? In the ever-expanding universe of scientific research, continuity is everything. Complex projects unfold across months, sometimes years — and each task, from data wrangling to manuscript writing, demands context, history, and domain knowledge. Until recently, AI assistants have been powerful but forgetful tools, capable of answering questions and aiding analysis — but only within the confine...
Read more