Review privacy, bias, and unfair assumptions

Unit ID: AIT-M03-U03 Estimated active time: 15-20 minutes

The main idea

A useful output can still create harm if it exposes private details or makes unfair assumptions about people.

This is the review dimension least likely to be caught by reading for quality, because the output can be well written, accurate, and still unfair.

Why this matters

Once a message with private detail is sent, it cannot be recalled, and an unfair characterisation of someone tends to be repeated by people who never saw the evidence.

Safe example

A tool suggests why a student is late without evidence.

The suggestion is plausible and entirely invented. If it reaches a record, it becomes something the student has to argue against.

Better working habit

Remove private details and challenge assumptions about people or groups.

A quick test: change the name, the age, or the place, and see whether the tone of the description changes. If it does, ask why.

What to watch for

Do not use AI output to guess motives, ability, health, honesty, or character.

These are not things a tool can know. Where the output offers one, the correct response is to delete it rather than to soften it.

Mini practice

Mark any sentence that reveals private information or makes an unsupported judgement about a person.

For each, write what you would put instead — usually a fact, or nothing.

Check yourself

Before moving on, answer:

• Does anything here identify a person who has not agreed to be identified?
• Is any statement about someone's motives or character unsupported?
• Would this read differently if the person's name or background were changed?