Idea Generation — Introduction

Student at a computer brainstorming with GenAI If you get stuck as you work through this in-class exercise, ask the instructor. Have fun—this one is hands-on.

Goal: leave with 2–3 strong ideas and 1 question you can actually test.

Reminder on academic integrity

You must have permission from your instructor to use GenAI in any assessed work. Some courses forbid it; others allow limited or full use. Follow your course outline and citation rules. Using GenAI without permission is academic misconduct under UVic’s Academic Integrity Policy.

What you’ll learn

  • A simple, repeatable ideation pipeline: diverge → cluster → converge.
  • How to turn topics into measurable research questions.
  • How to summarize safely (tool limits, chunking, verification).

Before you start

Open one tool: Copilot, Gemini, ChatGPT, Perplexity, or Claude (any is fine).

Safety: Don’t paste confidential/personal data. Redact names/emails/IDs (e.g., [Researcher_A], [Email_1]).

A. Diverge (5–7 min): create many options

Goal: generate a broad list of plausible directions without judging them yet.

Prompt (copy/paste, replace angle brackets):

  • **Role**: Research mentor.
  • **Action**: Propose 20 distinct topic ideas about <your broad area>.
  • **Format**: Table with columns = Idea, One-sentence rationale, Tags(3).
  • **Constraints**: No duplicates; vary methods (survey, experiment, case study), contexts (K–12/higher ed/industry), and scales (classroom/institution/community).

Now try your own: If 20 is too many, ask for 10. If it’s too generic, add constraints (e.g., “Canada only,” “undergraduate makerspaces,” “low-cost data”).

B. Cluster (5–7 min): find patterns

Goal: reduce noise by grouping related ideas.

Prompt:

  • **Action**: Cluster the 20 ideas into 5 labeled themes.
  • **Format**: Table with columns = Theme, Ideas included, Why this theme matters (≤20 words).
  • **Constraints**: No theme overlaps; each idea appears once.

Optional follow-up:

  • **Action**: For each theme, generate 2 “stretch” variations that increase novelty without losing feasibility.

C. Converge (8–10 min): pick winners with criteria

Goal: select 2–3 ideas worth pursuing using measurable criteria.

Prompt:

  • **Action**: Score each theme 0–5 on Novelty, Feasibility (skills/time), Data availability, Scope fit, Potential impact.
  • **Format**: Table with columns = Theme, Novelty, Feasibility, Data, Scope, Impact, Notes.
  • **Constraints**: Brief notes with concrete risks or assumptions.`

Decide: Keep the top 2–3 themes. Write 1–2 risks you will check next (e.g., “access to participants,” “IRB/ethics needed,” “data exists?”).

D. Turn topics into measurable research questions (RQ drill)

This is where ideas stop being vibes and become a real project.

RQ templates (choose one)

A) Comparison

  • “In <population>, does <approach A> vs <approach B> change <measurable outcome> over <time>?”

B) Effect

  • “What is the effect of <intervention/exposure> on <outcome> for <population> in <setting>?”

C) Relationship

  • “How is <variable X> associated with <outcome Y> among <population>?”

RQ generator prompt (Copy/paste)

  • Role: Research methods tutor.
  • Context: My chosen theme is: <paste theme label + 2–3 included ideas>.
  • Action: Propose 5 research questions that are specific and measurable.
  • Format: Table with columns = RQ, Population/Setting, Variables (IV/DV), Method (1 line), Metric (how measured), Feasibility note.
  • Constraints: Avoid vague verbs (explore/understand). Each RQ must include a measurable outcome and a plausible data source.

Tighten ONE RQ with acceptance criteria (Copy/paste)

  • Action: Rewrite RQ #__ to meet all acceptance criteria below.
  • Acceptance criteria:
    • Names a population and setting
    • Defines what is being compared or measured
    • Specifies at least one measurable outcome
    • Time-bounded (e.g., "over 2 weeks")
    • Ethically feasible (no sensitive data)
  • Format: Return (1) final RQ, (2) IV/DV, (3) method, (4) metric, (5) biggest risk.

E. Summarize a document safely (tool limits & chunking)

Free tiers and some enterprise tools limit input size. If a tool can’t “read” the whole article, it should say so. Ask it to be explicit—and chunk long text.

1) Pick an article (example news story):
LEGO helps Langford man recapture life after induced coma in 2018

2) Ask the tool to confirm capacity first:

Can you summarize the article at this URL? If not, say "TOO LONG" and ask me to paste text in chunks of your preferred size. URL: https://www.saanichnews.com/local-news/lego-helps-langford-man-recapture-life-after-induced-coma-in-2018-7333837

3) If too long, paste in chunks:
Here is CHUNK 1/3 (do not summarize yet). Acknowledge receipt only.

(Repeat for CHUNK 2/3 and 3/3.)

4) Then summarize across chunks: Action: Produce a 5-bullet summary using only the content from CHUNKS 1–3. Constraints: No new facts; quote 2 short phrases (≤10 words) with bullet numbers.

5) Compare summaries: Are the “quick” and “expanded” versions different in quality/coverage?

Guided examples you can try right now

1) Topic exploration (makerspaces)

I am an undergraduate student starting an honours project about university makerspaces. Generate 15 distinct topics. Vary method (survey/experiment/case), context (course/program/library), and outcome (learning/employability/access). Return a table with Idea, Rationale, Tags.

Now try a topic you care about.

2) Research questions from a bullet Suggest 5 research questions with measurable outcomes for: "Evaluate the effectiveness of makerspace programs." Return a table with Research Questions, Variables, Method, Outcome, Feasibility note.

Then run the Research Questions drill above to tighten one question.

3) Article summary (with limits)

  • Ask for capacity, then either summarize the URL or paste chunks.
  • Follow with: Expand to ≥5 bullets; keep quotes short; no new facts.

Note: Limits change over time. If a tool doesn’t acknowledge limits, assume it may not have processed everything—ask it to show its work.

Reflection (2–3 min)

  • Which two criteria mattered most for your down-selection?
  • Which prompt change improved quality the most—tone, scope, or format?
  • What risk will you check next (data, ethics, access)?

Badge evidence: Save a screenshot of your “top 3” table (with scores) or your “5 research questions” table.

Self-check (2 min)

  • Did you diverge (≥10 ideas), cluster (clear themes), and converge (scored & selected)?
  • Are your Research Questions measurable and scoped to your time/resources?
  • Did you verify at least one claim or data assumption?

Go further

For more prompting techniques (tone, examples, format control), a short overview you can skim next:
Prompt engineering techniques

NEXT STEP: Critical Review