01
FACTUAL CORRECTION
A published claim, figure, quotation, attribution or source reference is wrong. The text is repaired and the change is logged in public with the date and what changed.
CORRECTIONS & CONTACT
Errors are corrected in public.
A publication that asks readers to check its evidence has to be correctable. If something here is wrong, tell us. The record shows the repair, not a cleaner version of the past.
TWO KINDS OF CHANGE
A correction and an update are not the same thing.
Logging every edit makes the record meaningless; logging none makes it dishonest. The line sits here:
02
EDITORIAL UPDATE
Nothing was wrong. A link moved, a typo slipped through, or wording was clarified without changing the claim. Ordinary maintenance, not logged as corrections.
WHAT THE PUBLICATION COMMITS TO
- 01FROZEN MEANS FROZENA published essay is never quietly rewritten so that it reads as though it had been right all along. If a claim inside it fails, the correction appears with it. The error stays visible.
- 02CORRECTIONS ARE DATED AND SPECIFICEach correction records what was published, what it was changed to, when, and why. No unattributed edits, no silent deletions.
- 03THE SOURCE NOTE IS PART OF THE ESSAYAn error in a source note is treated exactly like an error in the essay it supports.
- 04A DISAGREEMENT IS NOT A CORRECTIONReasoned objection to an interpretation is welcome and answered as argument, but it is not logged as a factual correction, because it is not one.
THE RECORD
How a correction is recorded.
A confirmed error is fixed where it appears, with a dated notice attached to the essay or source note stating what was published, what it now says, and why. Every correction is also gathered here, so the whole record can be read in one place.
CORRECTION LOG
Essay 6 — What Changes When AI Enters the Room?
C-1 · REGISTERED 8 September 2026 · PUBLISHED 9 September 2026 · source note 6
What was published. …those who were touched agreed to the robot’s course suggestion significantly more often (63 percent against 37) and laughed more, with some rise in positive feeling.
What it now says. Six participants gave no clear yes-or-no answer and were set aside. Among the forty-two who answered, the touched group agreed significantly more often, 17 of 21 against 10 of 21, and laughed more and reported less negative feeling. The sentence that followed was bounded to no statistically significant difference in how they rated the robot.
Why. The “63 percent against 37” figure was the split of all yes answers across the two groups, 17 of 27 and 10 of 27, not a within-group compliance rate. The correct within-group counts are 17 of 21 against 10 of 21, with six unclear responses excluded from 48 participants. On feeling: the study found significantly lower negative affect after the interaction, while the difference in positive affect afterwards was not statistically significant. Source: Hoffmann and Krämer (2021), PLOS ONE 16(5): e0249554.
This log stays here permanently, including entries that are no longer flattering. An editorial or actuality update, where something was incomplete or where a live matter has moved, is disclosed by a dated notice on the source note it belongs to rather than recorded here as an error.
CONTACT
Reach the publication.
Corrections, source challenges, permissions and press all go to the same address, and corrections are read first.
To report an error, quote the sentence, name the page it is on, say what is wrong with it, and include a source where one is relevant. Written by a person, answered by a person.
Before writing about a source, it may be worth reading the evidence layer — several common objections are already answered in the limits each essay publishes about itself.