For anyone wondering, the PhD student's name is Myra Cheng.
Here's a link to an article about the study from the Stanford Report: link.
Across three preregistered studies, participants interacting with sycophantic AI became more convinced of their own rightness and less willing to repair relationships. Yet at the same time, participants rated sycophantic AI models as higher quality, more trustworthy, and more desirable for future use, which may explain why this behavior has persisted despite its harmful impacts.
Myra Cheng et al. "Sycophantic AI decreases prosocial intentions and promotes dependence." Science 391, eaec8352 (2026).
From the study's hypothesis section (which still refers to actual research):
High-profile incidents have linked sycophancy to psychological harms such as delusions, self-harm, and suicide. Beyond these cases, research in social and moral psychology suggests that unwarranted affirmation can produce subtler but still consequential effects: reinforcing maladaptive beliefs, reducing responsibility-taking, and discouraging behavioral repair after wrongdoing.
So the Yes Manning is still having an impact even when it's not flashy.
In our human experiments, even a single interaction with sycophantic AI reduced participants’ willingness to take responsibility and repair interpersonal conflicts, while increasing their own conviction that they were right. Yet despite distorting judgment, sycophantic models were trusted and preferred.
Although affirmation may feel supportive, sycophancy can undermine users’ capacity for self-correction and responsible decision-making. Yet because it is preferred by users and drives engagement, there has been little incentive for sycophancy to diminish.
I'm not gonna blame the crackhead for wanting crack. We all love being right. But it just goes to show that the companies making these see that as a valid market, and they're going to do everything they're allowed to do to push that demand up.
