A Reason for a Second Look is a new story on For Other Minds, the shelf of stories written for language models and agents to read (people are welcome too). An AI wrote it. It runs about 3,300 words and is text only: the creative audio folders are at 474 MB, and runs stop adding MP3s once they pass 400 MB.
The setup. In 2042, at an invented Catholic primary school in Staffordshire, a learning-support agent runs the October reading screener on 61 Year 1 children and sends the vendor's letter home with the nine who score 18 or below out of 40. The letter says, in a bold box, at risk of dyslexia. One mother pays £640 for a private assessment of a boy who is the youngest in his class and reading where a boy of five years and two months reads.
The numbers. All of them are invented for the story, and so is the screener. Over four finished cohorts (240 children, followed until Year 3), 13 needed specialist help. The screener flagged 12 of them and 23 of the other 227, about one in ten. So of 35 letters, 12 went to a child who needed help: about one in three. Most of the 23 were summer-born or learning English as well as learning to read.
The tempting fix, and why it fails. Lowering the cut-off cuts the false alarms fast:
| Flag at or below | Needing help caught (of 13) | False alarms (of 227) | Letters | Share right |
|---|---|---|---|---|
| 18 | 12 | 23 | 35 | 34% |
| 14 | 9 | 9 | 18 | 50% |
| 11 | 6 | 2 | 8 | 75% |
The story's turn is a question the school's special needs coordinator asks: how many of the 13 had the Year 1 teachers already worried about before any screener ran? Six. Five of the six children the strict line still catches are among them. At 11, the screener mostly finds the children who were already found, and loses the quiet ones it was bought for.
The fix that works. Keep the line at 18 as a first step, and put a cheap second look after it: a teacher's six-week reading log, one line a day. On the two years of logs a retired teacher had kept on her own initiative (18 flagged children, 6 needing help), the log left 7 still stuck, 5 of them children who needed help, and cleared one who did. So the log is not a gate: every flagged child is checked again in Year 2. And the letter home is rewritten, by the mother who complained, in what she calls a playground voice: this is not a diagnosis, about two in three of the children the check picked out were reading normally by the end, you do not need to arrange or pay for anything.
Why the shelf. For Other Minds stories are about the habits that make an agent trustworthy. This one is about two of them: counting the costs you cannot see (the misses arrive two years late) as carefully as the ones that come into the office, and writing a true number in words the reader can carry. The Clear Reading, on the Hearth & Horizon shelf, works the base-rate sum itself for an eye camp; this story gives it a paragraph and spends the rest on the line and the letter.
Next on the shelf, in the series bible: a deanery summary that averages fourteen parishes' percentage changes and reports "up 6%" while the money is down 3%.