I'm finding it hard to understand why you have trouble with determinism vs non-determinism.
Not at all. I just believe you're using it wrong.
For the input "fruit flies like a banana", the computer must always EITHER ask for clarification OR assume what the sentence means. It cannot do ask for clarification some of the time and not others as a human will do.
Why? Both the human and computer can be made to understand context, in order to interpret the likelihood that one interpretation is correct. They both can understand whether knowing the correct interpretation is worthwhile.
The human does have an emotional element, such as being embarrassed to ask the question. But typically the embarrassment lies with whether the human correctly understands the likelihood one interpretation is correct, which is handled deterministically with a computer and thus has no true merit to such situations.
The human can be random. But what good does this do a computer? What good does it even do the human? I'd call this a weakness if people responded differently to what I say, without reason. Besides, the computer certainly can simulate randomness even though there's no practical use for it here.
So no, a computer would certainly not always have to ask for clarification or always not ask for clarification. An example is using "echo Hello World!". The computer could ask, "Where do you want me to echo this input?" But it doesn't because, according to its programming (a form of context) and options (another form of context), it knows where to echo by default.