It's useful to have a typographical way of distinguishing two types of utterance which are otherwise graphically identical, e.g.:
Type 1: Can you raise your hand? (Inquiring about ability, perhaps a physiotherapist to a patient).
Type 2: Can you raise your hand. (Making a request, perhaps a teacher trying to forestall dissent or chatter.)
Type 1: Would you pass the test? (Asking about the consequent of some
condition.)
Type 2: Would you pass the test. (Actually requesting the
completion of a physical action on a physical object.)
In Type 1 sentences, what the speaker wants is a verbal response; in Type 2 sentences, the speaker is asking for a non-verbal action, conformity with a rule etc. Type 1 sentences are distinguished by a rising intonation in speech, and it would be useful to have a mark to do the same in writing.
Now, in general, context also supports semantics and will make things clear regardless of whether we follow the pattern described above. Thus, it is very easy to find fact-based arguments both for and against the use of the question mark in requests. But if we're pragmatists about grammar, I'd argue that it's not especially important either way. If a sentence is clear, it's good.
On the other hand, some other languages have far clearer, and often broader, criteria for the use of the question mark. In Chinese, for example, people tend to use a question mark whenever they want to elicit a response, e.g.:
"I don't know when you'll be coming tomorrow?"
I.e. I want you to tell me when you'll be coming, and I use the mark to show this.
Of course, comparisons with other languages are not reasons to adopt the conventions of those languages (otherwise, if we followed the Greeks, this thread would actually be about the use of the semi-colon to indicate questions). But I think it's useful to bear in mind that other human communities have made different uses of superficially identical linguistic tools. Language varies across space and through time. Whatever conclusion we come to is provisional and contingent.