You wrote the following: "Most of the kinds of things I get asked about tend to follow similar general patterns, after all, even if the details are very different, and one of the default techniques for answering them is to try to subtly recast the question as something I already have a ready answer for. I’m not sure how one would intentionally build this into a curriculum, though."
To a large extent, this is the goal of my Test Question Template (TQT) framework. I try very hard to get students away from the memorize-a-million-flashcards approach by explicitly showing them PATTERNS of questions -- each defined by a Lesson Learning Objective directly linked to specific examples of how I might assess that LLO on an exam, along with my promise that the actual exam question they'll get won't be the exact examples shown but will follow the same general pattern. If you read the Introduction of Evans et al. JMBE 2023 (https://journals.asm.org/doi/10.1128/jmbe.00200-22) you'll get the basic idea. The overall strategy relates well to your concern that if we do too much specifying of expectations, there's no need to be an agile thinker. TQTs aim to provide enough structure to usefully constrain students' focus while still keeping things open enough that they can't just plug-and-chug, and thus must be able to think on their feet within the constraints specified. [end of TQT ad for now; please follow up if interested]
It's rather hard to teach people to think on their feet if making them think on their feet has a negative emotional loading. Lawyers have that "Socratic method" in law school in which the professor calls on a particular student for an answer then guides him or her to the answer. It is both effective, since lawyers do need to think on their feet, often literally, but it can also be traumatizing. There might be a way to unload the approach a bit by, as you suggest, decoupling it from grading.
You wrote the following: "Most of the kinds of things I get asked about tend to follow similar general patterns, after all, even if the details are very different, and one of the default techniques for answering them is to try to subtly recast the question as something I already have a ready answer for. I’m not sure how one would intentionally build this into a curriculum, though."
To a large extent, this is the goal of my Test Question Template (TQT) framework. I try very hard to get students away from the memorize-a-million-flashcards approach by explicitly showing them PATTERNS of questions -- each defined by a Lesson Learning Objective directly linked to specific examples of how I might assess that LLO on an exam, along with my promise that the actual exam question they'll get won't be the exact examples shown but will follow the same general pattern. If you read the Introduction of Evans et al. JMBE 2023 (https://journals.asm.org/doi/10.1128/jmbe.00200-22) you'll get the basic idea. The overall strategy relates well to your concern that if we do too much specifying of expectations, there's no need to be an agile thinker. TQTs aim to provide enough structure to usefully constrain students' focus while still keeping things open enough that they can't just plug-and-chug, and thus must be able to think on their feet within the constraints specified. [end of TQT ad for now; please follow up if interested]
It's rather hard to teach people to think on their feet if making them think on their feet has a negative emotional loading. Lawyers have that "Socratic method" in law school in which the professor calls on a particular student for an answer then guides him or her to the answer. It is both effective, since lawyers do need to think on their feet, often literally, but it can also be traumatizing. There might be a way to unload the approach a bit by, as you suggest, decoupling it from grading.