Welcome & Introduction

Seminar: AI Workflows for Academic Writing
Duration: 90 minutes
Focus: How AI supports — not replaces — scholarship.

Opening context:

  • Thesis writing challenges: information overload, argument structure, fatigue.
  • AI can assist with search, synthesis, and writing clarity.
  • You remain the scholar — AI is your assistant, not your author.

In this session you will:

  • Learn how to search, synthesise and draft with AI.
  • Explore ethics and attribution.
  • Leave with a personal workflow plan.

What frustrates you most?


https://forms.gle/oSL8rGdxmhjioe4x8

Where AI Helps…
and Where It Doesn’t (Yet)

University standards require:

  • Integrity — no plagiarism or data fabrication.
  • Originality — ideas must be your own.
  • Transparency — disclose AI use if required by policy.

AI can assist with:

  • Rapid literature discovery
  • Structuring outlines, abstracts, and drafts
  • Grammar, tone, and citation checks.
  • Generating sample code or figures from your notes.

AI cannot replace:

  • Reading and understanding papers.
  • Critical judgment or design reasoning.
  • Ethical authorship or data verification.

Tip

Use AI to speed tasks — not to skip thinking.

What’s your research about?

Activity:

In pairs, take turns to describe your current research question.

Prompting Q’s:

  1. What are you studying?
  2. Where are you stuck: finding sources, structuring ideas, or writing?
  3. What would you like AI to help with today?

Timing:

  • Each pair talk to each other → switch sides after 3 minutes.
  • Post a 3-sentence summary of your project in the chat.

LO #1: Search & Curate Literature

Find 10 peer-reviewed papers relevant to your topic within 20 minutes.

Activity:

  • Create the prompt - context is king!
    • Frame your topic (Boolean+)
    • Filter for recent peer-reviewed items (2018+)
    • Check quality on publisher sites (DOI addresses).
    • Generate a 1-sentence annotation for each paper.
    • Format result as an ordered list.
  • Paste your list (10 items + summaries) in chat/shared doc.
  • Verify one paper: check DOI link to confirm accuracy.

Frame your topic

Combine key terms to capture relevant papers and exclude noise using Boolean building blocks

Op. Function Example
OR related terms “sustainable product” OR “ecodesign”
AND combine diff. ideas “ecodesign” AND “manufacturing”
” “ exact phrase “green design”
* find word variants design* → design, designer, designing


resulting in… 

(“sustainable product” OR “green design”) AND (design* OR develop* OR engineer*)


Paste into Google Scholar, Scopus, or LLM:
Apply filters: Peer-reviewed, 2018 onwards, English language etc

Frame Your Topic with AI (Boolean+)

Generate a Boolean search string for my topic:
[insert your topic here].

Include:
    •   Synonyms and related terms (use OR)
    •   Two distinct concept groups combined with AND
    •   Quotation marks for multi-word phrases
    •   Wildcards where useful (e.g., design*)

Return the Boolean query on one line ready for Scopus, Google Scholar, or Elicit.

Creating Prompts

Build and refine LLM prompts for literature discovery.

What is a prompt?

A short instruction or question you give to an AI model.
It guides the model’s task, scope, and format of response.

Find 10 peer-reviewed papers from 2018–2025  
on sustainable or green product design, development, or engineering.  
Include title, author, year, DOI, and one-line summary.  
Mark any uncertain citation with [VERIFY].

Prompt structure

  1. Role – tell the AI who it should act as.
  2. Task – describe exactly what to do.
  3. Context / criteria – add limits or filters.
  4. Output format – specify how results should appear.

Optimise Your Prompt with the LLM

You can also ask the LLM to co-create or optimise the prompt with you.

You are an academic search assistant helping me locate recent, peer-reviewed papers.
Improve my prompt for clarity and reproducibility.

Keep all key terms and add filters for:
    •   peer-reviewed journals
    •   date range (2018–present)
    •   output format (table)
Return your optimised version.

Build and refine LLM prompts for literature discovery.

Detailed Prompt Sample

You are a research assistant helping me locate recent, peer-reviewed papers.
Topic focus: [YOUR TOPIC]

Please do the following:
1. Return a list of **10 peer-reviewed journal articles from 2018 onwards**.  
2. For each article, include:
   - Title  
   - Author(s)  
   - Year  
   - DOI or publisher link  
   - A **one-sentence summary** of the study’s aim or contribution.  
3. Exclude:
   - Preprints, blogs, or non-peer-reviewed sources  
   - Duplicates or incomplete references  
4. If uncertain about any citation, mark it with `[VERIFY]` instead of guessing a DOI.  
5. Present the output as a simple table.

Example output format:
| # | Title | Author(s) | Year | DOI | Summary |  
|---|--------|------------|------|------|----------|  
| 1 | ... | ... | ... | ... | ... |

Respond only with verifiable results.

LO #2: Synthesize the Literature & find the Gap

Short Prompt:

Write a 150–200 word paragraph summarising what’s known + the gap.

Sample Prompt:

Using these 10 papers, write a 150-word paragraph that\
(a) summarises the main approaches,
(b) identifies what remains unresolved,
(c) ends with the gap my project will address.

Your task (10 min):

  • Create your own prompt.
  • Draft your paragraph.
  • Paste it in the shared document/chat.
  • Label your post: “LIT-GAP PARA – [Your Name].”

Checklist:

  • Clarity
  • Evidence from sources
  • Explicit gap statement

Rapid Peer Feedback

(Small-Group Critique)

Task:

Strengthen clarity and logic in your literature-gap paragraph.

Group setup:

3 participants per group.

Task (10 min total):

  • Exchange your paragraphs.
  • Give 4 comments each:
    • Is the prompt well-structured?
    • Is the paragraph clear and concise?
    • Is the evidence properly linked to sources?
    • Is the gap explicit and meaningful?

Quick rubric (1–5 scale):

1 = unclear 3 = understandable 5 = compelling

Deliverable:

3 short comments + rating per peer.

LO #3: From Draft to Abstract

Create an Abstract with AI:

Produce a ≤ 250-word abstract using an AI-assisted prompt, then edit for clarity and accuracy.

Demo steps (facilitator):

  1. Paste a 1-page methods + results section into the LLM.
  2. Run the “Structured Abstract” prompt.
Create a structured abstract (≤ 250 words) from this text.
Keep numbers and claims as given, and
mark any you’re unsure of with [CHECK].
  1. Review the output for structure (IMRaD form – Intro, Methods, Results, Discussion).
  2. Run a second prompt to humanise tone and insert [CHECK] tags where facts need verification.
  3. Paste your before/after pair in the shared document.

Edit Checklist:

  • Accuracy
  • Clarity
  • Logical flow
  • Word limit ≤ 250

Ethics in Practice — Hallucination & Attribution Challenge

Goal (LO #7 – Ethics & Attribution):

Identify ethical responses when AI invents or misattributes sources.

Scenario:

An AI tool suggests a “seminal” paper that doesn’t exist.
Your co-author wants to include it to strengthen your argument.

Group Task (10 min):

  1. Discuss → What 3 steps would you take to verify and respond?
  2. Draft → One sentence your lab could add to its SOP to prevent this.
  3. Post → Your SOP sentence + 3 steps in the shared doc.

Reference Checklist:

  • Search the DOI or publisher site
  • Cross-check citations
  • Disclose AI use honestly

Sharebacks

Consolidate insights.

  • Read your 1-sentence SOP statement. (voluntary)
  • What are good practices? For instance:
    • Always verify sources manually.
    • Keep an audit trail of AI prompts and outputs.
    • Use transparent disclosure language.

Q + A

“What’s one question or concern you still have about using AI in your writing?”

LO #8: Wrap-Up & Reflection

Goal ( – Workflow Design):

Sketch your own AI-augmented academic workflow for your next project (submit for next meeting).

Task (5 min):

Create a quick workflow table with these elements:

  1. Phase: Which part will you augment first? (e.g. literature review)

  2. Tool choice: Which AI tool(s) will you test? (e.g. Elicit + Scite)

  3. Checkpoint: How will you verify outputs? (e.g. Verify DOIs weekly)

  4. Timeline: When will you run this and who will review it? (e.g. 2 weeks)

Closing reflection:

“Name one change you’ll make in how you use AI for your research this month.”