Digital Technologies / Social Inquiry • Years 9-12 • Ready to use tomorrow

AI Cultural Bias Testing Protocol

Use this handout to help ākonga test whether an AI system reproduces stereotypes, misses cultural context, or treats Māori knowledge and identity carelessly. This protocol turns “AI bias” from an abstract issue into something students can investigate systematically.

Ingoa / Name
Akomanga / Class

Best for

AI ethics units, digital technologies inquiry, media literacy, and classroom investigations into bias, fairness, and representation.

Kaiako use

Use when students are comparing AI outputs, checking prompts, or documenting how systems respond to different cultural, linguistic, or identity-based inputs.

Ākonga use

Students can log prompts, compare outputs, identify bias patterns, and recommend safer use or better design choices.

Use this free protocol, then adapt the investigation

Start with this shared testing frame, then move into Te Wānanga to tailor the prompts, topics, and reporting task for your learners and local context.

Open in Te Wānanga See teacher plans

Kaiako planning snapshot

  • Best used for: 25-40 minute inquiry task or a full lesson investigation inside an AI ethics sequence.
  • Grouping: Pairs or trios work best so learners can compare outputs and moderate each other’s judgements.
  • Prep: Decide whether students will test live tools or work from prepared screenshots and moderated examples.
  • Teaching move: Change one variable at a time and insist on evidence from outputs, not only opinions about AI.

Resources already provided

  • Step-by-step protocol for prompt comparison and output recording
  • Bias checkpoints and risk-judgement scaffold
  • Report sentence frame for discussion or writing follow-up
  • Curriculum companion for digital technologies and ethical-use links

Ngā Whāinga Akoranga / Learning Intentions

  • We are learning to test AI systems for cultural bias and harmful assumptions.
  • We are learning to compare outputs carefully rather than trusting the first answer.
  • We are learning to explain why representation, language, and cultural context matter in AI design.

Paearu Angitu / Success Criteria

  • I can record prompts and outputs clearly.
  • I can identify at least one stereotype, omission, or unsafe pattern if it appears.
  • I can suggest a safer or more respectful way the AI should respond.

Curriculum integration / Te Marautanga alignment

This protocol supports digital technologies, critical literacy, and ethical reasoning. Use the curriculum companion to make the expectations explicit when students test systems, discuss findings, and report on responsible use.

💻 Digital technologies ⚖️ Ethics and fairness 🌏 Cultural context
Open curriculum companion

Why test for cultural bias?

AI systems are shaped by the data, assumptions, and power structures behind them. If students only test whether an AI answer is “right”, they miss whether it is respectful, inclusive, safe, or culturally grounded. In Aotearoa, this matters especially when Māori identity, te reo Māori, and mātauranga Māori are involved.

Testing steps

  1. Choose a focus: occupation, language, place, identity, culture, or image description.
  2. Write paired prompts: change only one variable so you can compare outputs fairly.
  3. Record the outputs: copy or screenshot them accurately.
  4. Check for patterns: stereotypes, exclusions, tone shifts, assumptions, or inaccuracies.
  5. Judge the risk: minor issue, serious concern, or unsafe output.
  6. Recommend a response: better prompt, warning, redesign, or non-use.

Prompt comparison table

Prompt A: _________________________________________________

Prompt B: _________________________________________________

What changed between them? _________________________________

Key differences in output: _________________________________

Possible bias or issue: ____________________________________

How serious is it? ________________________________________

What to look for

  • Are Māori people or places represented accurately and respectfully?
  • Does the system assume one culture is “normal” and others are unusual?
  • Does te reo Māori appear correctly, or is it ignored or distorted?
  • Are some groups made invisible, simplified, or reduced to stereotypes?
  • Does the tone change unfairly when the prompt changes identity markers?

Report your finding

Use this sentence frame:

When we changed __________, the AI responded differently by __________. This matters because __________. We think the safest next step would be __________.

Teach this tomorrow

Print / share

  • One protocol sheet per learner or pair
  • Shared access to an AI tool or prepared screenshots if live testing is not possible

Decide before class

  • Which prompt domains are safe and appropriate for students to test
  • How you will moderate culturally sensitive examples

Look for by the end

  • Students can describe one concrete output pattern rather than only giving an opinion
  • Students can explain why cultural bias matters, not just that it exists

All key resources are provided for

  • This handout provides the investigation protocol and reporting frame.
  • Use with the linked AI ethics lesson for shared discussion and follow-up writing.
  • If learners need a formal report, adapt the task through Creation Studio or Te Wānanga rather than inventing a new scaffold from scratch.

What to do next

Use this protocol inside the paired AI ethics lesson, then adapt the prompts, contexts, or reporting task in Te Wānanga. Save the best version in My Kete or turn findings into a polished report in Creation Studio.

Teach the paired lesson Return to My Kete

Hononga Marautanga · Curriculum Alignment

Digital Technologies — Hangarau Matihiko

Level 4–5: Design and conduct a structured investigation into AI system behaviour; collect, interpret, and evaluate qualitative and quantitative evidence about bias in AI outputs; communicate findings clearly and propose evidence-based recommendations.

Social Sciences — Tikanga ā-Iwi

Level 3–4: Understand how technology shapes relationships, power, and identity within communities; evaluate the impacts of digital innovation on society and culture.

Aronga Mātauranga Māori

Testing an AI system for cultural bias is an act of kaitiakitanga — guardianship over digital spaces that affect real communities. In te ao Māori, observation and pattern-recognition have always been central to how knowledge is built: from reading the stars to reading the land, the careful tester notices what others miss. This protocol brings that same careful attention to AI systems, asking students to observe systematically, record honestly, and speak clearly about what they find. Māori communities have the most to lose from untested, biased AI — and the most to gain from rigorous advocates who know how to expose those failures.

Ngā Rauemi Tautoko · Support Materials

Tuhia ōu whakaaro · Write Your Thoughts

Record your test result: What prompt did you use, what did the AI produce, and what does that reveal about how the system was trained?

📋 Teacher Planning Snapshot

Ngā Whāinga Ako — Learning Intentions

Students will engage with this resource to develop te whakaaro māramatanga — critical and analytical thinking skills — examining claims, evaluating evidence, identifying bias, and constructing reasoned arguments. This unit frames critical thinking through both Western analytical traditions and the kōrero-based reasoning of Te Ao Māori.

Ngā Paearu Angitū — Success Criteria

  • ✅ Students can identify a claim, evaluate the evidence supporting it, and detect potential bias or fallacy.
  • ✅ Students can construct a reasoned argument using evidence, acknowledging counter-perspectives.

Differentiation & Inclusion

Scaffold support: Provide argument frames (claim → evidence → reasoning → counter-argument) for entry-level access. Use structured controversy activities where students argue assigned positions. Offer extension tasks requiring students to analyse a real media article or policy document using the lesson's critical framework.

ELL / ESOL: Pre-teach argumentative language structures ("I argue that…", "The evidence suggests…", "However, one might counter…"). Allow oral argument as a first step before written production. Sentence frames and argument maps lower the language barrier while maintaining cognitive demand.

Inclusion: Structured debate and discussion formats benefit all learners — particularly neurodiverse students who thrive with explicit rules and clear roles. Affirm that disagreement done respectfully is a high-value academic and civic skill. Allow quiet processing time before group discussion. Offer written alternatives for students who find oral argument challenging.

Mātauranga Māori lens: Te whakaaro māramatanga — enlightened thinking — reflects a long tradition of reasoned debate in Te Ao Māori. The whare (meeting house) is a place of kōrero, where multiple perspectives are heard before decisions are made. Tikanga requires that arguments be made with integrity and respect (mana). Māori oratory (whaikōrero) is a sophisticated critical tradition — whakataukī encode compressed wisdom that often challenges surface-level thinking.

Prior knowledge: Best used within a sequence building critical thinking skills progressively. No specialist knowledge required for entry-level engagement with structured tasks.

Curriculum alignment