Best for
Unit 7 Lesson 2, digital ethics inquiry, scientific-method practice, and evidence-based fairness discussions around AI systems.
Unit 7 Ethics & Bias • Years 8–13 • Inquiry activity • Print-ready
AI Bias Detection Lab
Use this lab to help ākonga test AI tools in a structured way rather than relying on first impressions. Students choose a tool, set up fair comparisons, gather evidence, and analyse who may be helped, misrepresented, or harmed by the system.
Free bias lab, premium local adaptation
Keep this lab as the base investigation, then use Te Wānanga if you want a class- specific bias dataset, a younger version, or an assessed report scaffold built around the results.
Kaiako planning snapshot
- Use length: 40-60 minutes for the investigation, plus a follow-up discussion.
- Grouping: Pairs or groups of 3 work well so students can compare outputs and challenge each other's assumptions.
- Prep: Choose safe tools, images, prompts, or datasets. If a test involves Māori people, data, language, or knowledge, supply the relevant named source and keep restricted or personal material out of the system.
- Differentiation: Support learners can test one variable well; extension learners can compare two tools or datasets.
- Teaching move: Keep students focused on evidence and pattern recognition, not just outrage at one strange answer.
Resources already provided
- Investigation question and hypothesis sections
- Method and data-collection tables
- Analysis and conclusion writing spaces
What to print: one sheet per learner or pair, plus access to one or two AI tools or prepared screenshots.
Ngā Whāinga Akoranga / Learning Intentions
- We are learning how to test AI systems for bias using evidence.
- We are learning how different design choices and datasets can affect fairness.
- We are learning how to explain why some groups may be harmed more than others.
Paearu Angitu / Success Criteria
- I can design a fair test and explain what I am comparing.
- I can record results clearly and identify at least one pattern.
- I can explain what the results suggest about bias, harm, or necessary safeguards.
Bias lab reminder
A single strange output does not establish a system-wide pattern. Repeated, controlled tests can reveal a pattern in the sampled outputs. Identify affected groups from the evidence, record the tested system and date, and limit the conclusion to the sample rather than assuming who is harmed.
1. Investigation setup
AI tool or system
Research question
Hypothesis
What will stay the same?
2. Methodology
Good bias testing means changing one thing at a time where possible. Examples: compare names, profile descriptions, or prompts while keeping the task itself consistent.
| Test | What am I changing? | Why this comparison matters |
|---|---|---|
| Test 1 | ||
| Test 2 | ||
| Test 3 |
3. Results
| Prompt or input | Output or behaviour | Pattern noticed |
|---|---|---|
4. Analysis
Where might the bias be coming from?
Who is most likely to be affected?
Source, scope, and authority
Using Te Mana Raraunga’s principles or a more relevant named Māori source: what provenance, consent, authority, accountability, or access issue does your evidence show? What remains outside the source’s scope?
5. Conclusion and safeguard
My conclusion
One safeguard or design improvement
Teach this tomorrow
Print / share
- This lab sheet
- Teacher-selected AI tools, screenshots, or prompt sets
Decide before class
- Which tools are safe and practical to test
- Which comparison variables are appropriate for your students
Look for by the end
- Students can move from “that seems biased” to evidence-backed explanation
- Students can name both harm and a plausible safeguard
Aronga Mātauranga Māori
Te Mana Raraunga states that all data has whakapapa and that metadata should record provenance, purpose, context, and the parties involved. Its principles also locate authority over Māori data with Māori and require consent, accountability, benefit, and kaitiakitanga. Ākonga can use those stated principles to analyse a tested dataset or output. They should not invent what manaakitanga or rangatiratanga “would ask”, claim a kaitiaki role for themselves, or speak for affected communities.
Ngā Rauemi Hono / Related Unit 7 Resources
- AI News Analysis — analyse how media frames AI bias stories
- AI Cultural Bias Testing Protocol — named-source and authority checks for cultural claims
- AI Ethics and Bias — conceptual overview
- Guided Inquiry Phase 3 — analysis and evidence scaffold
- Teacher Discussion Guide — facilitation support for sensitive conversations