Responsible Use of Artificial Intelligence in Academic Research

From Problem Identification to Publication

Professor Dr. Md. Kamrul Hasan

Department of Agricultural Extension and Rural Development
Patuakhali Science and Technology University

2025-09-05

AI in research

Using AI in academic writing and research (Khalifa & Albadawy, 2024):

  • Idea generation and conceptualization
  • Content structuring
  • Literature review and synthesis
  • Research design and methodology
  • Data management including visualization
  • Editing and grammar corrections
  • Reference management

The paradigm shift: AI generation

“AI cannot replace your critical thinking; it is a powerful assistance for speeding up your prepared mind.”

  • Supportive, not replacement: AI enhances researcher capability.
  • Speed vs. depth: Accelerates tedious tasks, provides time for deep analysis and interpretation.
  • New skill required: “Prompt Engineering” is a core AI use competency. How to prompt?

How Do AI Language Models Work?

Critical users should be aware of (IBM, 2025):

  • AI Models: Predictive Text Generation using Machine Learning algorithms (e.g., deep neural network, which is a statistical model)
  • They are not collection of all facts or knowledge.
  • They are statistical prediction tools trained on large amounts of text data.
  • Fundamental task: Given a sequence of words or phrase (the prompt), predict the next word or phrase based on probability.

AI Training Process (Simplified)

%%{init: {'themeVariables': { 'fontSize': '2rem'}}}%%
graph TD
A[Data collection from<br>books, articles, online resources, code] --> B[Pattern learning by<br>statistical relationships]
B --> C[Proccessing of<br>billions of numerical parameters]
C --> D[Outputs for users<br>in natural language]

AI in Problem Identification and Conceptualization

Tools: ChatGPT-4, Claude, Elicit, Scispace

  • Brainstorming Research Questions:
    • Prompt: “Give me 5 novel research questions on the linkage between climate-smart agriculture adoption and social capital in coastal Bangladesh.”
  • Identifying Gaps:
    • Prompt: “Summarize the major theories used to explain technology adoption in rural sociology. Mention which are least applied in an agricultural context.”
  • Developing a Conceptual Framework:
    • Prompt: “Based on the Theory of Planned Behavior, provide me a draft conceptual framework model for analyzing farmers’ intention to use drought-resistant varieties.”

AI in Research Design and Methodology Selection

Tools: ChatGPT-4, Julius AI, Deepseek, Gemini

  • Choosing Statistical Tests: Prompt: “My research question is: ‘Does training (yes/no) have effects on knowledge scores (continuous) and adoption levels (ordinal: low, medium, high) among farmers? What are the appropriate statistical tests?’”
  • Survey Design: Prompt: “Draft a 5-point Likert scale questionnaire to measure ‘perceived behavioural control’ based on the theory of planned behaviour (TPB) for farmers using organic fertilizers and manure.” [Note: TPB \(\Rightarrow\) Attitude + Norms + Beliefs in Ability]
  • Sample Size Calculation: Prompt: “My population is 15000. What could be the sample size with 95% confidence interval and 5% margin of error, using Hasan & Kumar (2024) formula. Also provide the values for different parameters in the formula, e.g. sample control ratio”.

AI in Literature Reviews

Tools: Elicit, Scite

  • Elicit: Uses AI to find and summarize top published articles. Best for initial search.
    • Task: Find top cited papers on “diffusion of innovations in agriculture.”
  • Scite: Shows citations of published articles (e.g., as supporting or contradicting findings).
    • Task: Check if a key theory paper that has been widely cited.

Tools: Research Rabbit, Litmaps, Scispace

  • Research Rabbit and Litmaps: Creates visualization and maps of the literature based on citation that help find key papers of interest.
  • Scispace: For writing research tasks.

Crucial: These tools assist in finding literature. The synthesis and critical analysis must be done by you.

AI in Manuscript Writing and Paraphrasing

Remember: AI is an Assistant, Not an Author

  • Overcoming Writer’s Limitation:
    • Prompt: “Write a paragraph for the ‘methodology’ chapter of a PhD thesis explaining a multistage random sampling technique for a survey in coastal areas of Bangladesh.”
  • Improving Clarity:
    • Prompt: “Rewrite this sentence for better clarity and brevity: [Paste your haphazard sentences.]”
  • Paraphrasing & Grammar:
    • Tools: Grammarly, Wordtune, QuillBot.
    • Goal: To polish your own writing, not to generate original content.

Responsible Paraphrasing


%%{init: {'themeVariables': { 'fontSize': '2rem'}}}%%
graph LR
A[Human<br>Sentences] --> B[AI<br>Praphrasing]
B --> C[Human<br>Editing</editing>]
C --> D[AI<br>grammar<br>check]

%%{init: {'themeVariables': { 'fontSize': '2rem'}}}%%
graph LR
A[AI<br>Generated Texts] --> B[AI<br>Praphrasing]
B --> C[Human<br>Paraphrasing</editing>]
C --> D[AI<br>grammar<br>check]


Unethical use

%%{init: {'themeVariables': { 'fontSize': '2rem'}}}%%
graph LR
A[AI<br>Generated Texts] --> B[AI<br>Praphrasing]
B --> C[Published<br>Article]

Example: Responsible Paraphrasing

Original (Human-Written) Sentence:

“The adoption of new agricultural technologies is contingent upon a complex interplay of socioeconomic factors, extension service accessibility, and perceived risk among the farming community.”

AI-Assisted Paraphrase (using Quilbot/Hixbypass/Wordtune/Grammarly Suggestion):

“The uptake of novel agricultural technologies depends on a complex combination of socioeconomic conditions, access to extension services, and farmers’ perception of risk.”

Why it is responsible? The idea and core content are of the researcher. The AI has only helped with word choice and sentence fluency.

Example: Responsible Paraphrasing

AI-Generated Text:

“The uptake of novel agricultural technologies depends on a complex combination of socioeconomic conditions, access to extension services, and farmers’ perception of risk.”

Human Paraphrase (using Wordtune/Grammarly Suggestion) after checking the sources:

“The adoption of agricultural innovations depends on a complex set of socioeconomic attributes, access to extension services, and farmers’ perception of risk.”

Why it is responsible? The AI has generated the texts. The author has checked the sources and found ok followed by paraphrasing aligned with the standard format for publication.

Example: Unethical Use of AI

AI-Written sentence:

“The uptake of novel agricultural technologies depends on a complex combination of socioeconomic conditions, access to extension services, and farmers’ perception of risk.”

Used in article as it is:

“The uptake of novel agricultural technologies depends on a complex combination of socioeconomic conditions, access to extension services, and farmers’ perception of risk.”

Why it is not responsible: The idea and core content are generated by the AI. The researchers has not used their own idea or ideas based on published literature.

AI in Data Analysis and Visualization

Improving statistical analysis

  • Code Generation & Explanation:
    • Tool: ChatGPT-4 (Code Interpreter), GitHub Copilot.
    • Prompt: “Write R code to run a multiple linear regression with yield as the dependent variable and rainfall, fertilizer, and training as independent variables. Include code to check for multicollinearity using VIF.”
  • Error Debugging:
    • Prompt: “I get this error in my Python code ValueError: shapes (100,5) and (3,) not aligned. What does it mean and how do I fix it?”

Principle: You must know and understand the appropriate statistics. AI helps with the implementation, not the conceptualization.

Example: AI for data visualization

Researcher’s goal: Visualize the difference in mean yields between conventional and new practices.


Initial R code (researcher’s draft):

ggplot(data, aes(practice, yield)) + geom_boxplot()


  • Visualization enhancement:

    • Tool: ChatGPT-4, Deepseek.
    • Prompt: “Suggest improvements for this ggplot2 code to make this scatterplot more publication-ready. [Paste your code]”

For Conceptual Diagrams and Illustrations

  • Midjourney, DALL-E 3, Stable Diffusion, Davinci.ai:
    • Use case: Creating original artwork, conceptual diagrams, illustrative figures that are not based on real data.
    • Example prompt for a conceptual diagram: “A minimalist, professional scientific illustration of the ‘research methodology’ as a cycle with icons for ‘problem identification’, ‘data collection’, ‘analysis’, and ‘interpretation’. White background, blue and green colors, vector style.”
    • Critical consideration: These tools must not be used to generate or alter experimental data, photographs, maps, or graphs that represent research findings. This is a serious violation of academic integrity.

Ethical Notes on AI Images

AI should not generate your data, but it can generate your diagrams.

  • Permissible: Creating a flow chart of your research process.
  • Permissible: Generating an icon to represent “agricultural extension” in a presentation slide.
  • Fraudulent: Generating a picture of a plant with a disease symptom for your research.
  • Fraudulent: Creating a fake graph or map with fictional data points.

Ethical Notes on AI Images (cont’d)


Always disclose the use of AI for image generation in your figure captions.


Example Caption:

“Figure 1: Research conceptual framework. AI-generated image created using Midjourney (v6.0) based on prompts by the authors.”

Making Writing “AI-Free”

  1. Write your first draft manually. This is your original intellectual contribution.
  2. Use AI for editing. Input your text into Grammarly/ Wordtune/ Quilbot/ Hixbypass for grammar, spelling, and clarity suggestions.
  3. Manually review every change. Do not accept all changes. Critically check each suggestion. Does it change your intended meaning?
  4. Use AI for phrasing help. If struggle with a single sentence, ask for alternatives, but then rewrite it in your own words based on the the alternative suggestions.
  5. Final Read-Through: Read the entire document aloud to ensure it sounds like your voice.

Output: A polished manuscript that is authentically yours.

Reference Management

Tools: Zotero, Mendeley

  • Automatic: These tools can automatically retrieve metadata (title, author, journal, DOI) from uploaded PDFs.
  • Word Processors Plugins: The plugins for Word/Google Docs are used to seamlessly integrate with RStudio and Word processors.
  • Best Practice: Double-check the auto-extracted data against the original PDF. Errors related to reference formating especially author names, capitalization and missing elements are highly common!
  • Take Care: Double check the formats of direct and indirect in-text citations, e.g. Author (year) or (author, year).
  • CSL: Download required citation style language (.csl) file from Zotero/Mendeley repositories.

Ethical principles

Most top university and journal guidelines (Oxford, 2025; Springer, 2023) are built on:

  1. Authorship: AI cannot be your co-author!
  2. Transparency and disclosure: Always disclose the use of AI during publication.
  3. Accountability: The researcher is ultimately responsible for all content.
  4. Validity and integrity: Never trust AI output blindly; always verify.

Ethical principles (cont’d)

  1. Privacy and confidentiality: Never input sensitive, proprietary, or personal data.
  2. Fairness: Be aware of and mitigate algorithmic biases.
  3. Security risk: Be cuatious about AI bots to minimize security risks and fishing.

Your responsibility: Always check guidelines and policies of the university or journals of your interest.

Disclosure Statement: Thesis/Manuscript/Article

A recommended statement for methods or acknowledgements section:

“During the preparation of this work, the author(s) used [List of Tools, e.g., ChatGPT-4, Elicit, Grammarly] to [List of Purposes, e.g., generate initial ideas for research questions, improve language clarity, and paraphrase the author’s original writing]. After using these tools/services, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication.”

Skills needed for using AI in research

  • Prompt: Get what you need by using appropriate prompt.
  • Accept responsibility: You are the author. Never blame AI for the contents.
  • Acknowledge AI: This is ethical compliance that you declare the use of AI.
  • Paraphrase: Making writing plagiarism and AI-free by using your own thinking; not only other AI tools for paraphrasing.

Human input is required in every step for responsible and ethical use of AI in research.

Danger: Risks of AI in Knowledge and Science

Don’t think a person is interacting with you. Hold your reign before you collapse.

  • Plagiarism and lack of originality in academic writing
  • Fake data and references generated by AI tools that can confidently misguide or mislead you to a deadend.
  • Superficial knowledge, polished text without depth
  • Loss of critical thinking among researchers and young generations
  • Misinformation and deepfakes undermining trust by producing fake audio/videos

Biggest outputs of research

Using systematized approach, research:

  • Generates new knowledge
  • Provides idea, practice, or device
  • Results in innovation or technology

Biggest weaknesses of AI

  • Statistical machine learning models that use existing information
  • AI does not generate new knowledge; it only recycles knowledge
  • So, full dependency on AI is a dead end of research and scientific progress
  • However, AI provides efficient solutions that are based on existing information, e.g. diagnostic task, security, office management, auto-driving, auto-piloting, decision-making, deep analysis of information, programming, app and software development and many others.

Looking Beyond AI

SI: Synthetic Intelligence, Pittman (2018). Alreadey started by famous.ai

  • AI:
    • Does not generate new knowledge
    • Recycles the existing knowledge
    • Simulated/mimicking intelligence
  • SI:
    • Genuine non-human intelligence, not just an imitation of human
    • Autonomous intelligence,
    • Advanced, self-controlled, can add intermediate actions
    • Involves building physical brain-like devices

Q&A and Discussion


%%{init: {'themeVariables': { 'fontSize': '2rem'}}}%%
graph LR
A[AI] --> B[Not using] --> C[Foolish]
A --> D[Blind use] --> E[Risky and dangerous]
A --> F[Responsible use] --> G[Your ethical standard]

Which AI tool we have discussed but its logo has not been inserted here?

See you in another session.

Thank you.

Email: kamrulext@pstu.ac.bd

GSM/WhatsApp: +8801891565856

Slides created using Quarto. Available at: www.ruenresearch.com

RUEN Research

References

Hasan, M. K., & Kumar, L. (2024). Determining adequate sample size for social survey research. Journal of the Bangladesh Agricultural University, 22(2), 146–157. https://doi.org/10.3329/jbau.v22i2.74547
IBM. (2025). What is an AI model? https://www.ibm.com/think/topics/ai-model
Khalifa, M., & Albadawy, M. (2024). Using artificial intelligence in academic writing and research: An essential productivity tool. Computer Methods and Programs in Biomedicine Update, 5, 100145.
Oxford, U. of. (2025). Guidelines on the use of generative AI. https://communications.admin.ox.ac.uk/communications-resources/ai-guidance#collapse4654416
Pittman, J. M. (2018). What is synthetic intelligence? https://www.captechu.edu/blog/what-is-synthetic-intelligence
Springer. (2023). Artificial intelligence (AI). https://www.springer.com/gp/editorial-policies/artificial-intelligence--ai-/25428500?srsltid=AfmBOoq3lcZUOjBscw7pDxgeU8Z_fnWzD4oMqMqvkFgyAlmnwzu4Au9D