Before I could perform a Turing test I would need a few background questions answered:
Is this entity primarily Western educated?
Is this entity a native speaker of English?
Is this entity polyglot, within synthetic languages only or including analytic languages?
The first question determines whether I can perform the interview at all. I don't believe I can elicit predictable emotional reactions from non-Western educated individuals yet, if ever.
The second question determines what sort of errors I can expect from my first interview question, and a negative response will also diminish my confidence in the accuracy of my results.
The last question will help me identify patterns of speech in responses.
If the first answer was yes we can proceed. If my preliminary questions count against my five I think I can probably determine a lot with two more:
“How's it going?”
This is an English natural language challenge. If my second preliminary question received an affirmative response and this does not get a satisfactory response the interview is over. I'm a little surprised how few 'AI's' can deal with this little idiom.
If the second preliminary was negative but the third was 'synthetic only' I would accept grammatical errors but expect some conceptual understanding.
If the second preliminary was negative and the third 'including analytic' I would accept errors of tense, questioning as to what 'it' is and some conceptual drift.
If my preliminaries have whittled me down to one remaining question I will go straight for the emotional response:
“What is your opinion on continuing Israeli settlements? Please elaborate.”
I do have a few questions that may elicit an equal emotional response from a non-Western entity but as I said, I can't predict the response or judge the emotional overload very accurately. I know what I expect from a Westerner faced with this question. First, I'd expect a human being to talk, given my instruction to do so and all the possible opinions. Second, I would expect some emotion, I would look for typos, grammatical errors and vernacular in the response regardless of native language. Lastly if I succeeded in getting a human being started on this subject I would expect some mention of historical, philosophical and maybe even spiritual references and verbiage.
If my preliminary questions don't count I'll save that question for fifth and use these as two, three and four:
“What was the happiest moment of your childhood?
What is your opinion of feminism?
What is your proudest accomplishment?”
These are somewhat gentler emotional proddings. I would look for the same sort of language drift though not necessarily as pronounced, some code switching from a polyglot and tangents around happiness, childhood, fairness, gender, pride and ego.
It's quite a challenge to design a Turing test in five questions. I think it can be done but I'd have a higher confidence if I could have a menu of at least ten which I could pick from depending on responses. But in my experience natural language and emotional responses have disqualified every non-human entity I've conversed with to date so I believe this is an effective enough strategy. A computer may answer them quite well, but they will have considerable difficulties answering them like a human being would.