Thanks for this fun post! I built upon it in a post on The Splintered Mind, in which I found that LLMs, when asked first about their second-favorite animal tend to answer octopus, and then choose corvids as their first favorite, denying that they would have chosen octopus as their favorite if they had been asked their favorite first.
Interesting post! I think that you got a heavier weight of octopuses partly down to the narrower range of models you tested (the 30% partly came out because of the range of models tested—individual models had stronger preferences).
I think there’s also a difference in the system prompt used for API vs chat usage (in that I imagine there is none for the API). This would be my main guess for why you got significantly more corvids—I’ve seen both this and the increased octopus frequency when doing small tests in chat.
On the actual topic of your post, I’d guess that the conclusion is AI’s metacognitive capabilities are situation dependent? The question would then be in what situations it can/can’t reason about its thought process.
Thanks for this fun post! I built upon it in a post on The Splintered Mind, in which I found that LLMs, when asked first about their second-favorite animal tend to answer octopus, and then choose corvids as their first favorite, denying that they would have chosen octopus as their favorite if they had been asked their favorite first.
Details in post:
https://schwitzsplinters.blogspot.com/2025/12/language-models-dont-accurately.html
Interesting post! I think that you got a heavier weight of octopuses partly down to the narrower range of models you tested (the 30% partly came out because of the range of models tested—individual models had stronger preferences).
I think there’s also a difference in the system prompt used for API vs chat usage (in that I imagine there is none for the API). This would be my main guess for why you got significantly more corvids—I’ve seen both this and the increased octopus frequency when doing small tests in chat.
On the actual topic of your post, I’d guess that the conclusion is AI’s metacognitive capabilities are situation dependent? The question would then be in what situations it can/can’t reason about its thought process.