Instructor Notes

About This Course

Responsible Use of Generative AI is a short course about judgement, not hype. The teaching goal is to help learners understand how AI changes research work and to give them practical habits for using AI more selectively, with clearer boundaries, stronger validation, and better records.

The course works best when learners leave with:

  • a clearer sense of where AI actually helps
  • better prompting habits
  • stronger source-checking instincts
  • a lightweight approach to documentation and disclosure
  • more confidence saying “this task should stay manual”

Delivery Pattern

The introductory lessons each allocate 60 minutes and can run as two linked sessions or one combined session with breaks added separately. The applied lesson is modular; allow approximately two hours as a planning guide and confirm the chosen agenda before advertising a delivery time.

The applied lesson’s metadata currently allocates 115 minutes of teaching and discussion, while its outline totals 120 minutes without a separate break row. Select coverage and record breaks in the session plan rather than promising that every block and a break will fit into two hours.

A reliable rhythm is:

  1. Short concept framing.
  2. Example or comparison.
  3. Brief pair discussion focused on one judgement.
  4. Concise debrief before returning to the lecture.

Learners usually understand the material better after they have had to defend a prompt, identify a weak claim, or explain what should be checked next.

Session-Level Emphasis

Introduction to AI for Research

Use Introduction to AI for Research to establish why AI changes research work and why opportunity and risk now travel together. This lesson should reduce simplistic positions like “AI is obviously good” or “AI is obviously cheating” and replace them with more useful judgement.

Prompting for Research

Use Prompting for Research to move from principle to practice. Focus on prompt design, source boundaries, validation, documentation, and the question of when a repeated task deserves a workflow or agent.

Applied AI for Research

Use Applied AI for Research to connect tool selection and prompting to the research lifecycle. The lecture is the focus. Eight brief pauses ask pairs to make one evidence, design or boundary judgement before a short debrief; the lesson metadata allocates roughly 15 minutes to these pauses.

  • Use the fictional study extracts, interviews, dataset and funding call included in the lesson if learners cannot supply permitted material.
  • For literature work with real papers, confirm access and upload permissions in advance; public availability alone does not settle licence or institutional requirements.
  • For the analysis pause, let pairs choose the qualitative or quantitative example, or select the route that best fits the audience.
  • Check the quantitative debrief against five complete pairs and a mean change of 0.4 hours/week. In qualitative discussion, expect justified differences in interpretation rather than one correct set of themes.
  • Keep grant preparation distinct from externally assigned assessment. Recheck the linked tool documentation and funder/publisher policies before delivery.
  • Use the final boundary discussion to reinforce permission, evidence, method and responsibility rather than asking for a lengthy written output.

Common Sticking Points

  • Learners confuse confidence in output style with confidence in factual reliability.
  • People underestimate how much checking work AI creates.
  • Tool questions get mixed up with policy questions.
  • Some learners assume disclosure rules are obvious when they are often context-dependent.
  • Others assume every repeated prompt should become an automation, even when the task is too unstable.

Facilitation Moves That Usually Help

  • Ask “what would you check next?” whenever a group accepts an AI claim too quickly.
  • Ask learners to rewrite vague prompts rather than only criticising them.
  • Turn open discussions into ranking tasks when the room is too abstract.
  • When groups describe a benefit, ask what responsibility or risk moved elsewhere.
  • When groups describe a risk, ask what safeguard or habit would reduce it.

Managing Policy and Tool Questions

Learners will often ask for definitive answers about local policy, approved tools, or disclosure requirements. When the answer depends on context:

  • be honest about uncertainty
  • distinguish principle from local implementation
  • signpost the relevant institutional service or policy route
  • avoid improvising a confident answer when escalation is more responsible

This matters especially for:

  • sensitive or unpublished material
  • authorship and disclosure
  • external sharing and cloud tools
  • funder- or publisher-specific requirements

Room Setup and Activity Design

The learner practice pack provides a self-contained fictional source and authored AI-style response, so a five-minute source check still works without accounts or a live tool. Its prompt and use-record exercises each include a room response and a debrief. Use these as replacements or follow-up tasks; add time explicitly if you include them in a session.

  • Mixed-discipline groups usually work well because assumptions differ.
  • If confidence levels vary a lot, start with individual reflection before plenary.
  • If tool access is inconsistent, frame activities so they still work as paper-based prompt or workflow design tasks.
  • Encourage each group to capture one prompt improvement and one validation habit they want to keep.

Good Evidence of Learning

You are looking for learners who can:

  • explain where AI is useful and where it should be constrained
  • write prompts with clearer task definition and source boundaries
  • identify what needs verification before reuse
  • describe a lightweight way to document AI use
  • distinguish between a one-off prompt, a reusable workflow, and a risky automation idea

Course Navigation

Course Overview · Instructor Notes · Additional Material · Extra Exercises · Learner Setup