Yet despite distorting judgment, sycophantic models were trusted and preferred. All of these effects persisted when controlling for individual traits such as demographics and prior familiarity with AI; perceived response source; and response style. This creates perverse incentives for sycophancy to persist: The very feature that causes harm also drives engagement.

AI doesn't have to be sycophantic, but it is. If people really believe it can be used to change the world for the better, they're going to have to start by acknowledging what human traits have made the world so fucked up in the first place.

It reminds of one of the best parenting tips I was ever given by somebody who was raised by a parent with a maternal instinct in the negative range. She basically said something like when I'm not sure what to do as a parent, I start by thinking about what my mom would have done in the situation. "Step one: Ok, don't do that..."

Even if there can be no consensus on how AI should be used to improve the world, it does seem like we should at least be able to agree on the maladaptive traits we know we want to avoid passing on since they definitely aren't doing us any favors as a species.