This is a good observation that I think is also helpful as a top-level reply to the analysis itself by adding further context. Both because 1. it provides a non-sycophancy explanation for why Claude Fable adjusts its answer so drastically depending on the inferred occupation of the asker (an academic would be likely to want sources published in a credible academic journal), and 2. raises an interesting point that Claude's apparent preference seems durable enough that reasoning chain of thought wouldn't make it second guess itself to fall back to the "safe"... (read more)
This is a good observation that I think is also helpful as a top-level reply to the analysis itself by adding further context. Both because 1. it provides a non-sycophancy explanation for why Claude Fable adjusts its answer so drastically depending on the inferred occupation of the asker (an academic would be likely to want sources published in a credible academic journal), and 2. raises an interesting point that Claude's apparent preference seems durable enough that reasoning chain of thought wouldn't make it second guess itself to fall back to the "safe"... (read more)