40–45 minutesPrediction and training dataBias and reliabilityInteractive classroom activities
AI Literacy
An interactive introduction to how AI predicts, how training data shapes its answers, and why AI outputs should be checked critically.
Lessons
Students investigate how AI generates answers through prediction, how training data can introduce bias, and why confident-sounding answers still need to be checked.
Learning Objectives
- Explain next-word prediction and the data-to-prediction pipeline.
- Connect one-sided training data to biased outcomes.
- Identify reasons to double-check AI-generated answers.
Lesson 1: How Does AI Actually Work?
Students explore next-word prediction, the role of training data and bias, and four reasons AI answers should always be double-checked.
Learning Objectives
- Explain that AI generates answers by predicting likely next words from patterns in data
- Describe the data, numbers, patterns, predict pipeline
- Connect narrow or one-sided training data to biased predictions
- Give reasons why AI answers should be double-checked
Lesson 2: Data & Bias
Students examine how human judgments become embedded in data, analyze the Gender Shades case study, and experience how aggregate accuracy can conceal unequal outcomes.
Learning Objectives
- Explain how human decisions about what counts as good data shape AI systems
- Use the Gender Shades case study to identify bias caused by unrepresentative training and test data
- Explain why overall accuracy can hide failures for specific groups
- Predict how a deliberately biased dataset could produce unfair outcomes