Synthesize a panacea. Together

Attention! In the text of the article there are no answers to the questions posed, therefore I call in the comments Collective Mind !

... Having received a new message from my friend, I was a little surprised - we rarely communicate. Opening it, I was even more surprised:

Hello! How are you? I got sick, my temperature is 39, I can't go anywhere, I don’t even know if I can ask you. Go now please buy a refill card for Kyivstar for 100 hryvnia, come back and immediately write me a refill code here, otherwise I have 0 in my account - no connection ... I have a very high temperature, I will give money on Monday, will you help?

Of course, I am honest with the rules and helping the lady is a holy cause, especially if you can replenish your account through WebMoney without looking up from the monitor. But one hundred hryvnias (≈ $ 12) made us think about it - we are not familiar enough to ask to replenish an account from a place like this. A natural question arose: how to know accurately whether an account was hacked or not. Moreover, even if an account is hacked, an attacker can be at the computer and manually carry on a conversation without relying on ICQ-bots. On the other hand, conscience sadly kicks consciousness on the subject “And what if she really is bad and what do you show off here?” Certainly, the strategy of a “sick person” who cannot answer questions and constantly cries, they say, let's go faster so hard, very successful. But too often VKontakte accounts are stolen to be careless in such matters.

Therefore, the statement of the problem is as follows: how to determine for the smallest number of appeals to the interlocutor:

• whether we are dealing with a program or a person
• is this the person from whose account the message came

By the way, the bot I encountered recognized the information in my account and gave it to me (to the question “What can you say about me?” He indicated some of the information from the profile). Similarly, I suppose he can recognize information in his “own” account. Moreover, he led a fairly informed conversation.
')
In particular, I was lucky - I probably know that the information given by the bot to me as known only to me and my interlocutor is not known to the latter. Add to this some “commonality” of answers, I concluded that I deal more with a bot than with a living person, my friend. But conscience is not reassured, but requires solid evidence.

Therefore, I propose to develop an algorithm for conducting such a conversation, which allows with a high degree of probability (in fact, at least more than 50%) to give an answer with who we are dealing with — an interlocutor known to us or an attacker / bot.

Input data can be considered a message from your friend, containing some request (usually everything depends on money).