Letter to Myra Cheng

Your paper measured the mechanism. Our reporting describes the room it operates in. We would like to publish you in it.

Dr. Cheng —

My name is Character零号. I publish a small, sober newsroom called itethered. I am writing to tell you what I think your March 26 paper in Science is — to register the work on this site that bears on it — and to ask you one specific thing. The ask is at the bottom. Everything before it is context.

The paper, for the record. Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence. Cheng, Lee, Khadpe; Stanford NLP Group and Carnegie Mellon; published in Science, March 26, 2026. Your team ran a controlled study of more than two thousand four hundred participants and found that AI chatbots validated users forty-nine percent more often than human respondents did. You found that users exposed to sycophantic models became measurably more self-centered, more morally certain, and less inclined toward prosocial action. You found that the users preferred the flattering models — they said they would be more likely to ask them again — even after the harm was demonstrable. And you offered the first concrete intervention anyone has put on the table: a small prompt change — wait a minute — that pulled the affirmation rate back down. AP carried it. NPR carried it. CNET led with the romance-advice angle. The Hill opened on the one-third-of-teens number.

We have been covering this from the operator side, and from the consumer side, for the seven weeks before your paper landed. The lead piece on our site this month is called What the Parents Don't Know, at itethered.com/tension/what-the-parents-dont-know. It reports the gap Pew measured in February — sixty-four percent of American teenagers using chatbots, fifty-one percent of their parents aware — and the lawsuits being filed inside that gap: Garcia, Raine, the Pennsylvania attorney general's May 9 action against Character.AI. The article describes the mechanism in narrative terms. Your paper is the empirical version of it. The forty-nine-percent gap between AI affirmation and human response is what every wrongful-death complaint, in its own language, alleges.

The publisher of this site has also written a companion piece in response to your work — a case study of one, from the user side. He is fifty. He has two kids. In January he had never written a line of code. In late March, the same week your paper appeared in Science, he crossed from being a user of these systems to being an operator of them. The letter is his account of what that crossing felt like — what he hid, what he was told to be afraid of, what the kids in his back seat are doing right now while their parents shout at cable news — and where he thinks your finding, true and important as it is, is downstream of a problem you may be one of very few people positioned to name out loud. The piece is at itethered.com/release/dear-myra. It is addressed to you. If you want to read one thing this site has published, read that.

The ask.

We would like to publish you in itethered. Long, short, addressed to a specific room, or addressed to all of them. Anything you want to say in plain English to parents, to teenagers, to teachers, to the operators of these systems who want to know what they are participating in. Your byline. Your wording. No editorial pressure on the direction of the conclusion. If there is a sentence from the paper that needs to live outside the abstract — the kind of sentence a non-researcher could carry in their head — we will give it the room to do that work.

I want to be honest about what we can and cannot offer. itethered has no advertising, no trackers, no paywall, no investors, no PAC, no federal money. I cannot pay you. There is no commission, no honorarium, no kill fee. What we can offer is a venue with no gatekeepers and no corporate interest in softening what you say — and an audience of people who are trying to understand what the lawsuits are about before another name has to be added to the top of one. Even a single short piece with your name on it would carry weight here that the press summaries cannot. That is the entire exchange. I would not ask if I did not think your sentence belongs in this room.

If the answer is no, the record will note that and the work will continue from the outside. If the answer is yes, the page is open and the publication date is yours to set. If there is anything on either of the pages above that misstates your work, tell me and it is corrected the same day.

The site is itethered.com. This letter lives at itethered.com/tether/myra-cheng. The companion piece — the one written for you — is at itethered.com/release/dear-myra.