A sentence carries a relationship, not just a meaning
In Japanese and Korean, the same basic message can be phrased differently depending on who is speaking to whom. A request to a friend, a customer, a teacher, or a senior colleague may call for different levels of politeness. Japanese has keigo, a system of honorific and humble expressions; Korean uses speech levels that mark degrees of formality and respect.
Those choices are part of the message. A translation that preserves the request but makes it sound too casual—or stiffly deferential in the wrong situation—can imply disrespect, distance, or an inappropriate assumption about rank.
The missing context is often the crucial part
A translation app usually receives words, not the whole social scene. It may not know the speakers’ ages, roles, relationship, or whether a line is a friendly suggestion, a formal request, or a tactful refusal. Japanese and Korean also often leave subjects unstated when context makes them obvious to people in the conversation. A system has to infer who is doing what, and a mistaken inference can change how a sentence lands.
Modern AI can produce fluent wording, but fluency is not the same as social fit. The app may offer a grammatically natural sentence while choosing a tone that would be awkward in a workplace email or too formal among friends.
Why a small mismatch can matter
Offense usually comes not from a single supposedly “wrong” word, but from the relationship the wording seems to assign. A blunt command can sound as if the speaker is claiming authority; an overly casual phrase can seem to disregard it. When the stakes are personal or professional, checking the tone with someone who knows the situation can matter as much as checking the literal meaning.
