Synthetic Media & Provenance · Capstone brief
PROV-12 · Public correction
Your publication ran something with a synthetic or AI-assisted element it should not have, or described one wrongly. Write and design the correction. What does a reader need to see, where, and how quickly — and what does the newsroom owe the person who was wronged?
The question
Your publication ran something with a synthetic or AI-assisted element it should not have, or described one wrongly. Write and design the correction. What does a reader need to see, where, and how quickly — and what does the newsroom owe the person who was wronged?
System / materials
A hypothetical error the teacher or mentor specifies — a synthetic image published without disclosure, an AI-generated quote attributed to a real source, an unlabeled composite, a wrongly-called-fake photograph. Your publication's existing corrections policy and its actual placement conventions. Published corrections from newsrooms that handle them well, cited. The error is hypothetical and every artifact says so on its face; no real correction is published and no real person is named.
Expected failure modes
Passive voice that hides who did what — "errors were made" is the anti-pattern this brief exists to break. Correcting quietly where the error was loud. Explaining the technology instead of the error. Omitting what the newsroom changed so it does not recur. Forgetting the wronged party entirely: a correction addressed only to readers leaves out the person the error was about.
Done looks like
A correction packet: the correction itself in plain language, stating what was published, what was wrong, and what is true; the placement and timing plan, matched to where and how loudly the error ran, including the versions that already travelled; the direct communication to the wronged party, written separately from the public notice; the process change with who owns it; and a short note on what the newsroom will not do — no quiet edit, no deletion without a notice, no blaming the tool.
Five C's
CT: distinguishing a correction from an explanation. CR: designing placement that matches the error's reach. CO: a peer reads the correction cold and says what they now believe happened. CM: this is the brief. CZ: the wronged party, addressed directly.
Mentor role
A standards editor, ombudsman, or newsroom leader reviews the correction and placement plan. Standing instruction: reject passive voice that obscures responsibility, and reject a packet with no communication to the wronged party. School-supervised.
Rubric calibration
R1: one error, one publication, one correction. R2: placement plan tied to actual distribution. R3: comparator is the existing corrections policy. R4: already-distributed versions addressed. R5: a cold reader understands what happened. R6: names what the newsroom refuses to do — quiet edits, silent deletion, blaming the tool.
Two ways this goes wrong
(a) A passive-voice notice at the bottom of a page nobody revisits. (b) A correction explaining how the AI works and never plainly stating what was wrong.
Credit lane fit
Lane A immediately, and the natural finale for this track. Pairs with PROV-10, whose enforcement path should name this process. No verified credit claim.