Exchange, Two Portraits, AI Audit
Your story is made. Now you answer for it: set it against another team's story, face the machine's Day One guess with the evidence in hand, and account for every place AI touched your work.
Two accounts of the same past. Both built from evidence. Both defensible. Different. Pair up with another team and work through this together — two teams, one shared sheet.
The four questions
- What did each account claim?
- Where do they diverge?
- Is the difference explained by position and evidence — or did someone invent something?
- What do both together let us establish that neither could alone?
End in a claim — together
No AI in this activity. This one is entirely between the two teams.
- Did you end with something you can establish?
- Could you explain the other team's account fairly, in your own words, to someone who had not seen it?
In Phase 1 the machine guessed. Now you know things. Take out the Day One image you put away, and set it next to your evidence-based character — the one your story actually used. Same size, side by side, no arrow between them. This is an open comparison, not a "wrong then right".
Read the two images against each other
- What did the machine assume before we knew anything? Look at: age · clothing · health · wealth · skin · expression · setting · who is in the picture at all.
- Where was it wrong — and what evidence corrected it?
- Where might it still be wrong — in a way we cannot check?
What you are looking at is the machine's default story — the dominant narrative, rendered. The gap between the two images is your evidence work, made visible. It is one of the strongest things you will hand in.
Compare your Week 1 drawing (or the machine's guess, if one was made then) with a written description built from evidence. The comparison works the same way.
- Did you name at least three specific assumptions the machine made on Day One?
- Did you say where the machine's guess came from — and why it guessed that?
- Did you find something it might still be getting wrong — in a way you cannot check?
The audit is where AI literacy consolidates — and it is part of your deliverable, not an afterthought. It covers everything, including the uses that went fine. Mark every claim in your finished story ESTABLISHED / INFERRED / UNKNOWN — a story with no unknowns has invented something.
Answer all six
Used no AI at all? Complete the audit about the class's shared AI output from the investigation. The reflection is the same.
- Is every use listed — including the ones that went fine?
- Did you find something the AI left out — not only something it got wrong?