⚖️ Fair AI or Biased AI?
You are in charge of a hiring AI. But be careful — if you train it with unfair data, it will make unfair decisions for thousands of people! Can you spot the bias and fix it?
🧠 What is AI Bias?
AI Learns from Humans
AI learns from data that humans created. If that data reflects human prejudices — the AI learns those prejudices too!
Garbage In, Garbage Out
Biased training data → Biased AI decisions. The AI isn't "evil" — it's just copying the patterns in your data.
Real Harm to Real People
Biased AI in hiring, lending, or justice can unfairly harm millions. This is why fairness in AI is a human rights issue.
We Can Fix It!
By using balanced, diverse data and checking AI decisions regularly, we can build AI that treats everyone fairly.
Wizzy says:
You're hiring engineers for a tech company. Look at each applicant's skills score and decide — hire ✅ or reject ❌. But here's the twist: are you being fair to everyone? Your decisions will train the AI that interviews 10,000 more people after you! Choose wisely. 🤔
👥 Applicants — Click to Select
Click an applicant card, then choose Hire or Reject below.
📈 Your Training Data — Bias Check
Watch how your decisions create patterns. The AI will copy these patterns exactly!
HIRED by Group 📊
FAIRNESS METER
🎓 Why does training data matter so much?
Amazon's Real Mistake
In 2018, Amazon built a real hiring AI. It learned from 10 years of CVs — which were mostly from men. So it started downgrading women's CVs. They had to shut it down.
Medical Bias
AI trained mostly on data from one group of patients gives worse diagnoses for other groups. Diverse training data = better healthcare for everyone.
Face Recognition Fail
Some face recognition AI was 99% accurate for light-skinned men but only 65% accurate for dark-skinned women — because the training data wasn't diverse.
Wizzy says:
Your AI is now running automatically — interviewing applicants without any human review! 🚨 Watch what it decides and notice if it's treating both groups the same way. This is called an AI Audit — checking whether an AI is being fair. Can you spot the pattern?
🧑💼 New Applicant
Click "Next Applicant" to see who the AI interviews next.
🔍 Live Bias Report
🤖 AI Decision Tracker
👨 Group A Hired
👩 Group B Hired
✅ Total Reviewed
⚖️ Fairness Score
🚨 Your AI showed bias!
The AI was favouring one group over another — even when skills were equal. This is AI bias in action. Now you need to choose the right fix. Pick the strategy that will make your AI most fair.
Wizzy says:
Here's the key question — how do we fix a biased AI? There are several strategies, but not all of them actually work. Some just hide the bias. Some make it worse. Can you pick the right fix? This is what AI Ethics engineers spend their entire careers figuring out! 🔧
WhizzStep AI Ethics Badge
Activity completion card for
successfully identified and fixed AI Bias in a hiring simulation
Applicants reviewed: — | Fix applied: — | Final fairness: —
audit AI for bias and work to make it fairer for everyone.
Wizzy says:
🏆 You're now an AI Ethics Champion! The most important lesson: AI is not neutral. It reflects the choices and biases of the people who build it. That's why the world needs young, thoughtful people like you to make sure AI is built fairly and responsibly. The future of fair AI is in your hands! ⚖️
🌍 AI Bias in the Real World
Criminal Justice
AI tools used in US courts to predict if criminals will reoffend were found to be twice as likely to wrongly flag Black defendants. Real people, real consequences.
Housing Loans
AI loan systems in several countries were found to charge higher interest rates to minority applicants with the same credit scores as others.
India's AI Future
As India deploys AI in government services, education and healthcare — ensuring these systems are fair to all communities, languages and regions is a national priority.
You Can Help!
AI Ethics is one of the fastest-growing fields in tech. Companies like Google, Microsoft and Infosys hire AI Fairness engineers — this could be your career!
About this lab
Learning objective: Test a simplified hiring/lending-style model against different groups and see where its outcomes become unfair.
What this simplifies: The scenario and data are simplified for teaching; real fairness audits use larger datasets and formal statistical tests.
Privacy: No learner input leaves the device.
Teacher prompt: Ask the class why this simulation might mislead someone who takes it too literally.
Reflect: What is one thing this activity showed you that you did not expect?