More importantly, I think it ignores the reality in how the output is consumed. Writing is generally meant to be read; that is its use. When evaluating a text,

More importantly, I think it ignores the reality in how the output is consumed. Writing is generally meant to be read; that is its use. When evaluating a text, the qualities it has when read by a human are the primary criteria. When we evaluate code by reading it, our largest concern is how it will be interpreted by a machine. Clarity matters in comments, but can be justifiably sacrificed in code for an elegant or efficient solution. It might be more appropriate to compare the requirements of composing legal documents to that of coding. Testing for understanding of methods, patterns and common concepts is far more applicable in that case. Asking if a candidate knows the parameters for a work to be classified as work-for-hire or what kinds of indemnifications and disclaimers should be in a contract are arguable useful indicators of suitability.