Analyze · Grades 8-12
Bias Detective
Investigate model outputs to spot hidden bias in training data.
The AI idea behind it
An AI model learns from examples. If the examples reflect an unfair past, the model learns the unfairness too, and then repeats it at scale. That is algorithmic bias.
Bias usually comes from the data, not from the AI "deciding" to be unfair: who was included, who was left out, and what the historical decisions in the data looked like.
How it works in the app
- Work through five scenarios. In the first, a company trained an AI on ten years of hiring data in which it had mostly hired men, and the AI now rejects most female applicants.
- For each scenario, choose what went wrong from four answers and check your answer.
- Each scenario explains why the bias matters for the people affected.
Try this
Fix the data
- After each scenario, write one sentence on where the bias came from.
- Write down what data you would collect, or remove, to fix it.
- Pick one AI system you use and list who might be missing from its training data.
What it shows: Fixing bias starts with the data and the people checking it. "The AI is sexist" is the wrong diagnosis; "the training data was biased" is the one you can act on.
Where it fits in the curriculum
More Analyze tools
Try Bias Detective in the app
One of the creation tools in LittleAIMaster for Grades 6-12.