Lesson 1.4 - Hallucinations, Bias, and False Confidence

Learning objective

Understand AI risks that directly affect testing quality.

Chapter 4 / Objective 4

One of the most dangerous AI behaviors is hallucination.

Hallucination means:

• AI confidently gives incorrect or made-up information

For testers, this can look like:

• Invented test steps
• Incorrect API behavior
• Made-up error messages
• Non-existent configuration options

AI is also biased toward:

• Common use cases
• Happy paths
• Popular frameworks

That's a problem-because testers focus on what goes wrong.

How testers should respond

• Ask AI why it suggests something
• Request assumptions explicitly
• Cross-check against requirements

Example follow-up prompt:

"What assumptions are you making about this system?"

Key takeaway

Confidence is not correctness. Always challenge AI output.