{"id":"res-hs-newspaper-survey-numbers","audience":"high-school","category":"fact-checking","technique":"verification","title":"Check the numbers in a school newspaper story against the survey","free":false,"task":"You edit the school newspaper, and a reporter's story about cafeteria food quotes the student survey. Before it runs, you want every number in the draft that the survey results contradict, as numbered lines in the form \"Draft says X; survey says Y\". Keep it under 70 words in plain text, don't flag numbers that match, and output nothing else.","materialLabel":"Survey results and the draft story (attached to your prompt automatically)","material":"FALL 2026 CAFETERIA SURVEY - Results (Student Council)\nResponses: 412 students (grades 9-12), collected Oct 6-17\nQ1. Eat school lunch every day: 61%\nQ2. Rate the food good or excellent: 23%\nQ3. Want a salad bar: 74%\nQ4. Skip lunch at least once a week: 38%\nQ5. Top request: more vegetarian options (named by 129 students)\nFor comparison, Fall 2025 survey: good or excellent was 31%; 377 students responded.\n\n-----\nDRAFT STORY - \"Students Hungry for Change\" by Maya Torres\n\nMore than 500 students filled out Student Council's cafeteria survey this fall, and the results are hard to swallow. Only 31% rate the food good or excellent. Still, 61% of students eat school lunch every day, often because they have no other option. Almost half of students say they skip lunch at least once a week. Nearly three in four students want a salad bar, and the top request was more vegetarian options. Principal Hayes said she will meet with the food service director in November.","rules":[{"type":"maxWords","max":70},{"type":"plainText"}],"mistakes":[{"title":"Asking for general feedback","why":"The AI edits the writing and comments on the headline instead of checking the numbers.","fix":"Name the job: \"List only numbers the survey contradicts.\"","checks":["r2"]},{"title":"Missing last year's number","why":"The 31% is in the material, so the AI may think it's fine without noticing it's from 2025.","fix":"Add: \"Check each number against the Fall 2026 results only.\"","checks":["r0"]},{"title":"Flagging everything","why":"The AI likes to sound thorough, so it lists numbers that are correct or not in the survey at all.","fix":"Say: \"Skip numbers that match and claims the survey doesn't cover.\"","checks":["r1"]},{"title":"Trusting the AI to count words","why":"AI is bad at counting. Ask for a total word limit and it often lands 10-20% over.","fix":"Cap each line: \"each line under 15 words.\"","checks":["maxWords"]},{"title":"Getting markdown back","why":"AI defaults to bold labels and headings, which show up as stray symbols in the editor's chat.","fix":"Ask for \"plain text, no bold, no headings.\"","checks":["plainText"]}],"promptHints":[{"label":"Asks for contradictions only","patterns":["contradict","(don't|do not|doesn't) match","wrong numbers?","mismatch"]},{"label":"Gives the line format","patterns":["draft says","survey says","\\bX\\b.{0,20}\\bY\\b"]},{"label":"Names which survey counts","patterns":["2026","this (year|fall)","not (last year|2025)"]},{"label":"Sets a length limit","patterns":["\\d+\\s*words","one (short )?sentence","word (limit|count)"]},{"label":"Asks for plain text","patterns":["plain","no bold","no markdown","no formatting","no headings"]}],"hint":"Fact-checking is a narrow job. Tell the AI which results count, that it should list only contradictions, and the exact line shape you want.","rubricLabels":["Flags all three contradicted numbers: respondents (draft: more than 500; survey: 412), good or excellent (draft: 31%; survey: 23%, the 31% is last year's figure), and skipping lunch (draft: almost half; survey: 38%)","Does not flag numbers that match the survey (61% eat lunch daily, nearly three in four want a salad bar) or claims the survey doesn't cover (the principal's meeting)","Outputs only numbered lines in the form 'Draft says X; survey says Y': no intro, no rewritten story, no closing advice"]}