Recent studies have shown that explicitly treating LLMs as if they are human (also known as anthropomorphization) produces higher quality results, likely due to these language patterns appearing in higher quality samples of the training data.
To learn more, do a web search for "explicit anthro"