Unit 7 Foundations • Years 11–13 • Digital systems • Print-ready

Algorithm Literacy: How AI Makes Decisions

Use this handout to help ākonga see that AI decisions are not magic. They come from inputs, categories, rules, weightings, and human design choices. The goal is clear explanation first, then stronger fairness and ethics judgement later.

He Ara Whakatau · Decision Pathway Analysis

Ingoa / Name

Akomanga / Class

Investigate how a daily algorithm (e.g. TikTok/Instagram feed, Spotify recommendations, Google search, school risk-predictor) works, map its pathway, and evaluate its fairness.

Part 1 · The System & Inputs

Identify the system and what information goes in.

1. Name of the algorithmic system you are analysing (e.g. YouTube recommendations):

2. What input data does this system collect from you (e.g. watch history, clicks, search terms, time spent)? List at least three inputs:

Part 2 · The Decision Rules

Analyse how it ranks or predicts.

3. What rules or calculations does the system use to score or filter inputs? (e.g. IF user watched >30 seconds of video A, THEN recommend video B):

4. What output or decision does the system produce? (e.g. customized feed, score, classification):

Part 3 · Fairness, Bias & Safeguards

Apply a critical lens to the system's impact.

5. Where could bias or errors enter the training data or rules? Who might be unfairly excluded or represented?

6. What human safeguard or oversight would you implement to check and appeal the system's decisions?

7. Te Ao Māori Connection: How does this system impact Māori Data Sovereignty (who controls and benefits from the data collected)?

Ngā kōrero kaiako · Teacher notes — everything below supports the handout; it is not part of the printed worksheet.

Mātāpono · Core Concepts (Algorithm Basics)

AI Isn't Magic—It's Math

When TikTok shows you the perfect video, when Netflix recommends a show, when your bank approves a loan— it feels like magic. But it's not. It's algorithms: step-by-step instructions that computers follow. Understanding how they work is the key to questioning them.

What is an Algorithm?

Algorithm: A set of rules for solving a problem or completing a task. Think of it like a recipe: input ingredients → follow steps → get output.

Simple Algorithm Example: Making Toast

INPUT: Bread slice
STEP 1: IF bread is frozen → defrost for 30 seconds
STEP 2: Place bread in toaster
STEP 3: Set timer based on bread type (white = 2 min, whole grain = 3 min)
STEP 4: IF toast is burnt → throw away and start over
OUTPUT: Toasted bread

AI Algorithm Example: Recommending Videos

INPUT: Your watch history, likes, time spent on each video
STEP 1: Analyse which videos you watched longest
STEP 2: Find similar videos (same creator, topic, hashtags)
STEP 3: IF other users with similar history liked it → rank higher
STEP 4: Predict which video will keep you scrolling longest
OUTPUT: Personalized video feed

Reflection question for students: What's the goal of the TikTok algorithm? Is it to show you the "best" content, or to keep you on the app as long as possible?

Ngā Tūraru · Risks & Case Studies (Where Algorithms Go Wrong)

🚨 Real Example: Racist Hiring Algorithm

Amazon built an AI hiring tool to screen resumes. They trained it on data from past successful hires. Problem: Most past hires were men, so the algorithm learned to penalize resumes with the word "women" (e.g., "women's chess club captain"). It discriminated automatically.

Result: Amazon scrapped the tool. But how many companies use similar algorithms without telling us?

The 3 Main Algorithm Problems

1️⃣ Biased Training Data

If you train AI on biased data (e.g., mostly white faces, mostly men's writing), it will reproduce that bias. Garbage in = Garbage out.

2️⃣ Opaque Decision-Making (Black Box)

You're denied a loan, but the bank won't explain why—"the algorithm decided." Even the programmers sometimes can't explain AI decisions. How do you appeal?

3️⃣ Optimising for the Wrong Thing

YouTube's algorithm is optimised for "watch time," not "truth" or "wellbeing." Result: It recommends conspiracy theories and extreme content because that keeps people watching longest.

Practice Scenario: School Uses AI to Predict Student "Risk"

Your school implements an AI system that predicts which students are "at risk" of dropping out or getting in trouble. It uses data like attendance, grades, behavioural incidents, and socioeconomic status to flag students for "intervention." Ask students to evaluate concerns, design omissions, transparency, and potential harms of this intervention model.

Tūmahi · Extension (Design an Ethical Algorithm)

Design an algorithm for a real-world problem (e.g., matching students to internships, allocating school resources, moderating online comments). Write out the steps. Then identify: What could go wrong? How would you prevent bias?

Problem to solve: Match student skills/interests to local businesses for Work Experience.
Potential algorithm steps:
1. Collect student preferences (subject fields, hours) and business needs (skill tags, location).
2. Score similarity matches based on tag intersection.
3. IF distance between student and business > 15km, penalize the match score (unless student has own transport).
4. Rank top 3 opportunities for each student.
Bias prevention strategy: Ensure the distance penalty doesn't disproportionately lock low-income or rural students out of high-quality placements. Build in a travel subsidy override.

Mō te kaiako · At a glance

Best for

Unit 7 foundations, digital-literacy unpacking, pre-bias lessons, and any class that needs to slow down “AI decided” into a visible decision pathway.

Kaiako use

Model one familiar example first, such as recommendation feeds, maps, ranking tools, or school software. Then let students map a second example independently.

Ākonga use

Students can label inputs, rules, outputs, and fairness risks, then explain the system in plain language to someone else.

Linked next step

Best followed by AI Ethics and Bias or AI Bias Detection Lab once the basic process is clear.

Adapt this handout · Te Wānanga

Free algorithm scaffold, premium local adaptation

Keep this as the shared explanation frame, then use Te Wānanga if you want a local case study, a younger reading version, or an infographic/report task built around the same model.

Swap in school software, social media, banking, or job-matching examples.
Generate junior and senior versions of the same algorithm pathway.
Turn the explanation into a slide deck, poster, or assessed paragraph.

Open in Te Wānanga See teacher plans

Kaiako planning

Kaiako planning snapshot

Use length: 25-40 minutes for the concept build, or a full period if students apply the model to multiple systems.
Grouping: Whole-class modelling first, then pairs for the independent analysis section.
Prep: Choose one familiar algorithmic system students already know from daily life.
Differentiation: Support learners can complete the pathway and one risk; extension learners can compare two systems.
Teaching move: Keep reminding students that humans choose the data, categories, success criteria, and thresholds.

Resources already provided

Algorithm pathway and bias checkpoints
Plain-language explanation prompts
System-analysis and safeguard sections
Teacher-only curriculum companion

Ngā whāinga me ngā paearu · Learning intentions & success criteria

Ngā Whāinga Akoranga / Learning Intentions

We are learning how algorithms and AI systems make decisions.
We are learning where data, rules, and probabilities influence an outcome.
We are learning why automated decisions still need human oversight and challenge.

Paearu Angitu / Success Criteria

I can explain the main steps in an algorithmic decision pathway.
I can identify at least one likely bias or risk point.
I can suggest one safeguard or human check that would improve the system.

Why this matters in Aotearoa

Algorithm literacy matters because “neutral” systems are still designed. When people say “the algorithm decided”, it can hide the fact that humans chose the categories, the training data, the goals, and the thresholds. Understanding that pathway is the first step toward questioning whether the outcome is actually fair.

Aronga Mātauranga Māori

In te ao Māori, data and knowledge are not neutral — they carry whakapapa and obligations. Māori Data Sovereignty (Mana Motuhake i ngā Raraunga) holds that Māori have the right to govern, own, and interpret data about themselves and their communities. When digital systems are designed without this understanding, they risk perpetuating colonial patterns of extraction: taking knowledge from communities without accountability or benefit-sharing. The concept of kaitiakitanga extends naturally to the digital realm — guardianship of what is collected, stored, and shared about us is as important as guardianship of land, water, and living knowledge systems.

Hononga Marautanga · Curriculum Alignment

This handout supports understanding of digital systems, critical explanation, and wider conversation about how automated decisions affect fairness and responsibility.

Digital Technologies — Hangarau Matihiko

Level 4–5: Understand how digital systems and AI tools work; evaluate the social, cultural, and ethical implications of technology; design and apply computational thinking skills to real problems.

Social Sciences — Tikanga ā-Iwi

Level 3–4: Analyse how technology shapes relationships, power, and identity within communities; evaluate the impacts of digital innovation on society, including effects on Indigenous data sovereignty and cultural representation.

Ngā Rauemi Tautoko · Support Materials

This handout is designed to be used alongside the broader unit resources available at Te Kete Ako handouts library. Related resources from the same unit are linked in the unit planner. All resources are provided — no additional preparation is required to use this handout in your classroom.