Purpose: How bias enters an AI system through its training data, and why a confident answer can still be unfair or wrong.

Exercise

• Explore how bias can show up in AI-generated content, particularly when the system is asked to describe people or recommend things, by prompting a chatbot using Code.org's AI Chatlab. Click here to open the activity guide for this exercise. https://studio.code.org/courses/exploring-gen-ai-2025/units/1/lessons/5/levels/1?no_redirect=1

Solutions

• Link to view the explanation of bias in AI - what bias is, and how data bias and algorithm bias show up in the output.
• Link to view the solution for the Exercise - walkthrough of the Code.org AI Chatlab activity.

1. Check the availability of the AI tools used in this lesson and read the instruction documents for all the exercises before the class.
2. Emphasise the words recommendations, training data, bias, data bias and algorithm bias.
3. Create a class section on Code.org before the lesson so the students can complete the activity.
4. Points to Ponder:
   • Two different things can go wrong: the examples it learned from were unfair, or the way it picks an answer is unfair. From what you saw today, which one do you think you were looking at?
   • If a biased AI is used to shortlist people for a job, who should be held responsible - the people who built it, the people who chose to use it, or nobody?