So how would a scientist prove conclusively that other humans are conscious and self-aware and not just very advanced automata that behave in ways which mimic consciousness and self-awareness?

No matter how much data you collect, when you're assessing behaviour there's always a subjective interpretation of how behaviours differ and what they mean in context.

Experiments like these simply disguise and hide the subjectivity.

Questions of subjective experience are absolutely unprovable scientifically.

You can collect correlations between stimuli and responses, but when you're done you have a table of correlations between stimuli and responses - and that's all.

In practice we assume that our own states correlate with behaviours, and this also applies to other humans in a straightforward way. (It often doesn't, but it's a nice thing to believe.) From there it's an easy step to making the same mapping for certain animals, albeit with more of a stretch.

But this is all unprovable, even as a hypothesis. All you can say objectively is that stimuli either match or don't match expected behaviours.

That's all that's ever on the table.