For the specific example given there, I'd say it's most often called feature engineering. I'd also argue that it's a lot more necessary than most people think, but I'm probably just being stodgy and am biased by working in relatively narrow domains.

Calling it "feature engineering" implies it's still being fed into some sort of trained classifier to make the final decision, though.

What you're describing of your own work might better fall under the broad category of an "expert system".