Using AI Effectively & Ethically in Course Assignments

Session Context & Purpose

  • Workshop led by senior tutor (second‐language English speaker with 27 years in NZ) to prepare students for assignments that allow AI.
  • Key goals:
    • Clarify what is permitted / expected when using AI in the course.
    • Encourage students to ask other lecturers for explicit AI rules.
    • Reduce fear of accidental misconduct (some students avoid even Grammarly).
    • Demonstrate free AI research tools and compare their outputs.
    • Explain how to integrate AI without losing personal, authentic scholarship.

Course Assessment Overview

  • Deliverables: 2 written reports, 2 videos, 2 oral presentations.
  • Today’s focus = written reports; video tutorial may follow.
  • Reports limited to ≤ 2 pages plus references; reflective paragraph (~100 words) and optional prompt appendix.

Why (and How) AI Can Help

  • Time-saving for:
    • Brainstorming & outlining ideas.
    • Summarising complex literature / papers.
    • Improving academic tone, grammar, spelling.
    • Generating visuals (e.g., PowerPoint scaffolds) — tutor used AI for presentation structure.
  • Learning aid:
    • Re-phrasing difficult concepts until understood.
    • Practising prompt writing (analogous to early keyword-based Googling).
    • Rapid spell-check for non-native writers (tutor’s personal example).
  • Potential industry use-case shown (Dubai “AI chef” restaurant):
    • Bot creates recipes from huge biochemical flavour datasets.
    • Humans still required for sensory judgement (taste, colour, plating) → metaphor for AI-drafted text needing human refinement.

Major Risks & Limitations

  • Loss of authentic voice; AI often defaults to American English & generic phrasing.
  • “Hallucinations”:
    • Entirely fabricated references, authors, or data.
    • Plausible-sounding but false statements due to mis-linked training data.
    • Detected previously: 3 invented citations in a student project.
  • Incomplete or geographically irrelevant results (e.g., Greek study returned for NZ query).
  • Over-reliance undermines critical & creative thinking.

Prompt Engineering Essentials

  • Precise, iterative questioning critical to quality output.
  • Include specific keywords (e.g., “native New Zealand”, “North Island”, “commercial apple orchards”).
  • If output unsatisfactory:
    • Refine location, species, date range, etc.
    • Request alternative wording or deeper explanation.
  • Save every prompt for traceability and reflection.

Demonstrated Free Tools (Research-oriented)

ToolCore StrengthsNoted Weaknesses / Findings
ChatGPT (free tier)Drafting, brainstorming, re-phrasing sentences.Limited factual depth; must verify all claims.
ElicitSearches ∼126 000 000\sim126\,000\,000 papers via Semantic Scholar; auto-generated abstract & report.Initially returned only 1960s NZ data → later drifted to non-NZ studies; quality varied day-to-day.
SciSpace (formerly SciSpa)Table view: objectives, methods, conclusions, research gaps; “Deep Search” in paid tier.Some gaps; insects/pollination example partly incomplete.
SciSpace “Notebooks”One-click research report draft.Coverage shallow for niche entomology topic.
ConsensusEvidence-based short answers; good starting summaries.Supplied Greek, Kashmir data despite NZ filter; needed repeated prompt revision.

Verification Workflow Suggested

  1. Use AI tool to discover candidate papers.
  2. Copy titles into University Library catalogue (Learn ⇢ Te Puna Mātauraka link) to access full texts.
  3. Cross-check methodology, geography, species, dates.
  4. Evaluate for hallucinated or outdated content.

Assignment Integration Guidelines

  • Do NOT submit unedited AI content.
  • ALWAYS verify facts, numbers, claims.
  • Maintain personal analytical voice; minor grammatical imperfections acceptable.
  • Reference AI per APA-style example:
    • In-text: “According to ChatGPT (OpenAI, 20252025) …” (optional if space tight).
    • Reference list: “OpenAI. (20252025, March 1010). ChatGPT response to prompt ‘Pros and cons of bridge NAAC’. https://chat.openai.com/ …”.
  • Keep running log (can be Appendix):
    • Prompt text.
    • Tool used.
    • Stage of workflow (brainstorm, structure, source search, language polish, etc.).
    • Need for peer-reviewed sources? (Y/N)
  • Reflective paragraph (≤ 100 words) inside 2-page report should cover:
    • Strengths AI provided.
    • Weaknesses / failures encountered.
    • Key lessons for future projects.

Ethical, Philosophical & Practical Implications Discussed

  • Universities still drafting comprehensive AI policies → responsibility falls on individual lecturers & students.
  • Authenticity more valued than ever in graduate recruitment; AI-generated CVs risk uniformity.
  • Human creativity & critical evaluation remain irreplaceable (parallel to chef tasting the AI recipe).
  • Rapid AI evolution means current advice may change within a year; stay adaptive.

Tips for Maintaining Authentic Voice

  • Retain some quirky / region-specific language instead of fully AI-polished US prose.
  • Insert personal anecdotes (e.g., placement experience, farm upbringing, fieldwork observations).
  • For videos/orals: treat yourself as consultant advising real farmer; emphasise individuality.

Future Support & Resources

  • Potential follow-up tutorial on AI for video/graphic design (e.g., DALL·E, Midjourney, Synthesia).
  • Assignment instruction sheet on Learn includes:
    • Direct link to University of Sydney’s AI resource hub (discipline-specific tool lists & guidelines).
  • Email / office hours: Tutor invites questions or demo requests; willing to learn new tools from students’ discoveries.