Video summary

AI in Research Writing: Russellian Perspective of Truth and Responsibility

Main summary

Key takeaways

Educational

Main ideas, concepts, and lessons

  • Research writing now faces an ethical risk: The talk frames the modern era as one where AI can lead to distorted knowledge, not just increased productivity.
  • Russell’s principle applied to AI: Using philosopher Bertrand Russell as a moral lens, AI must be treated as a servant of truth, with ethical responsibility, not as a tool that manufactures illusions.
  • Citation contamination as knowledge corruption: A core analogy compares fraud in flood-control infrastructure (substandard materials/fake receipts) with research corruption caused by AI-related citation problems (e.g., citing hallucinated, fake, or retracted papers).
  • AI’s benefits come with limits: While AI can improve speed, efficiency, and accuracy, current tools have gaps (especially around real-time retraction checking) and can create misconduct risks.
  • Institutions must set policies and require disclosure:
    • Universities should have board-approved AI policies.
    • Students must declare which AI tools were used and for what purpose to ensure honesty and due diligence.
  • Practical ethical framework (Russell-inspired):
    1. Truth first (verify outputs and sources)
    2. Human reason above machines (AI supports, not authorizes)
    3. Responsible knowledge (avoid citing suspicious/unreliable work)
    4. Transparency (disclose AI use; avoid fabricated references)
    5. Ethical AI literacy (continuous education on critical/ethical use)
  • How to avoid AI-generated or unreliable research content: Use primary sources, treat AI as a draft aid, cross-check authoritative references, and avoid fabricated data/citations.
  • Natural Language Processing (NLP) is why AI affects writing: NLP enables summarization, grammar/style correction, translation, keyword extraction, sentiment analysis, etc. This improves writing—but also increases risks of overreliance and misrepresentation.
  • Use of similarity/AI-detection scores must be interpreted carefully:
    • Similarity index and AI-check scores (e.g., from Turnitin) are used as indicators, but the talk emphasizes false negatives and recommends knowing which parts were AI-generated.
  • Real-world consequences demonstrate seriousness of AI misuse: The talk cites multiple retraction/penalty cases (journal retractions, hidden prompts, false citations, terminated faculty), illustrating that undisclosed or fraudulent AI use undermines peer review and trust.
  • Tool walkthroughs are presented as examples of capability—not permission to bypass ethics:
    • Google Scholar AI Reader (browser/extension) for outlines, sections, citations, navigation.
    • Elicit/“S pace” for prompt-based literature review tasks (summaries, gaps, refactoring, journals).
    • DeepSeek for more robust literature assistance and analysis prompts.
    • StealthWriter/Humanizer and QuillBot are mentioned; their use is treated as ethically sensitive, especially when aimed at evading detection rather than genuinely improving writing.
  • Formal citation and disclosure rules: The talk provides guidance on how to cite AI tools (e.g., ChatGPT) in references and in-text, including timestamps for direct outputs.
  • AI Declaration Form as a key compliance mechanism:
    • Includes: researcher info, manuscript title, tools/models used, purpose, limitations/verification, responsibility statement, and signature/date.
    • Includes “acceptable vs not acceptable” paraphrasing behavior checklists.

Methodology / instructions (detailed bullets)

A) Ethical use of AI in research writing (core Russell-inspired practice)

  • Truth first
    • Verify AI outputs, facts, and sources before inclusion in any paper.
    • Require disclosure in an AI declaration form for sections where students used AI.
  • Human reason above machines
    • Use AI as support, not as an authority.
    • Ensure your reasoning and conclusions reflect your own scholarship.
  • Responsible knowledge
    • If sources are suspicious or potentially retracted, do not cite them.
    • Aim to prevent citation contamination.
  • Transparency
    • Disclose AI usage where required (and avoid fabricated references).
  • Ethical AI literacy
    • Maintain continuous learning about AI limitations, risks, and academic integrity expectations.

B) Verification checklist to prevent hallucinated or retracted citations

  • Use primary sources and read the original articles yourself.
  • Check:
    • DOI
    • whether the paper exists on the web
    • bibliographic details (accuracy)
  • Cross-check claims with authoritative references:
    • Don’t rely on a single source for one claim; compare multiple key references.

C) “AI as draft aid” workflow

  • Use AI to help generate drafts (writing support, structure, grammar), but:
    • Do not treat AI text as a final source
    • Paraphrase manually after understanding
    • Base discussion/results on your actual study findings

D) Citation and disclosure instructions (as presented)

  • Disclose AI use in the required sections (e.g., introduction/method/conclusion/recommendations/title crafting if needed).
  • Cite AI tools in references when used (examples given):
    • Include tool name, the institution/organization (if applicable), year, and URL.
  • In-text attribution
    • If using information generated by ChatGPT: add “(ChatGPT, [date time])” format.
    • If directly quoting/using a direct output, include a timestamp.
    • If multiple prompts: include short-form prompt attribution in text.

E) Acceptable vs not acceptable AI paraphrasing behavior

Acceptable checklist

  • Use AI only as support
  • Read and understand the original source
  • Verify the accuracy of AI paraphrases
  • Edit/refine personally
  • Cite the original source properly
  • Ensure the paraphrase reflects your own understanding

Not acceptable checklist

  • Copy paraphrase without revision
  • Don’t read/understand the original source
  • Rely on AI to generate paraphrasing for submission
  • Don’t check accuracy
  • Don’t cite
  • Results in risk of plagiarism/dishonesty flags

F) AI Declaration Form components (instructional elements)

  • Researcher information (name, institution)
  • Manuscript title
  • Declaration of AI use (tools/models used)
  • Purpose of use:
    • editing, data analysis, writing support, citation help, interpretation of graphs/results
  • Limitations and verification steps
  • Integrity statement/responsibility for author’s own conclusions
  • Signature and date
  • (Optionally) specify which pages/parts of the manuscript used AI

G) Interpretation guidance for AI detection/similarity scores (Turnitin context)

  • When interpreting scores, the talk encourages:
    • Aim for low AI score (example guidance: below 30%).
    • 31–60% = medium likelihood of AI use.
    • 61–100% = high likelihood (almost all content may appear AI-generated).
  • Consider false negatives:
    • Even if AI score is low, verify manually which parts were AI-assisted.

H) Tool prompt examples (as demonstrated)

Google Scholar AI Reader

  • No specific prompt workflow shown, but capability includes:
    • section outline and navigation
    • identifying where ideas were taken from
    • citation generation in multiple citation styles

“S space” / Elicit-like prompts

  • Examples of prompts:
    • “Summarize a paper in simple terms. What are the main findings and limitations?”
    • “Explain the methodology in layman’s terms.”
    • “Rephrase this paragraph for clarity and academic tone (don’t copy).”
    • “List possible research gaps.”
    • “Create five survey questions related to the topic.”
    • “Extract all references in APA 7th edition.”
    • “Suggest journals where your study can be published.”

DeepSeek

  • Example prompts:
    • “Summarize the latest studies on AI in higher education.”
    • The talk also includes options like thematic analysis, identifying research gaps, data interpretation, citation support, proposal drafting support, and policy briefing summarization.

StealthWriter / Humanizer

  • Example prompt behaviors shown as common:
    • “Rewrite the paragraph to sound more natural and human like”
    • “Simplify technical terms without losing academic quality”
    • “Make this abstract clear and more concise”
    • “Avoid AI detection and sound authentic”
  • Ethical caution implied: using this to evade detection is characterized as unethical.

QuillBot

  • Example prompts:
    • paraphrase literature review while keeping meaning accurate
    • summarize two pages into bullet points
    • rewrite abstract in formal academic tone
    • grammar clarity and APA 7 citations

Speakers / sources featured (identified)

Speaker(s)

  • Unnamed national president speaker (the lecturer):
    • The talk identifies the speaker as the National President of the Philippine Institute of 21st Century Educators (PI21), speaking on behalf of PI21 during the workshop.

Philosophers / authors

  • Bertrand Russell (also referred to as Bertrand Arthur William Russell)
  • Elsevier (source of “five responsible AI principles,” attributed to Elsevier’s Scopes/“scopo’s” database)

Institutions / organizations / publications cited or referenced

  • MIT Technology Review (study discussing retractions and “hallucinated science effect” / AI-generated retracted paper cycles)
  • Cambridge University (referenced for an “AI declaration / generative AI declaration” example)
  • Turnitin (used as an example for similarity and AI-detection scoring)
  • Google Scholar
  • Springer Nature (journal retraction example: retracted 129 articles due to heavy AI generation)
  • National University of Singapore (NUS) / Nanyang Technological University (NTU) (cases involving hidden AI prompt and penalties/accusations)

Tools / platforms mentioned

  • Google Scholar AI Reader
  • Google Scholar Paper Reader / Chrome extension
  • Elicit / “S space” (referred to in the transcript as “S space”)
  • DeepSeek
  • ChatGPT
  • Grammarly
  • Google Translate
  • StealthWriter / Humanizer dashboards
  • QuillBot
  • Turnitin
  • DALL·E / OpenAI (for citation examples)
  • MS Word / citation managers (generic mention; tools like “RefWorks/Refman” appear in transcript)

Original video