A number of experiments would be performed using an Alpha Interface Unit and the responses recorded.

During the experiment the operator asks a series of prepared questions one at a time, allowing an interval of 10 seconds between each question for responses.

The questions are pre-selected from a list of questions known to have an above average probability of producing a response.

All responses are recorded - but any phrases occurring outside a (nominal) 6-second window extending from the exact end of the operator completing the last word of a question are discarded.

An administrator who is a tested skilled listener transcribes all the words heard during that period with the exception of the operator's words. Each such set of words is called, here, a Response.

Each Response is placed at a randomly selected position within a group of nine other possible responses. The other nine responses have been made-up and should be genuinely possible responses to a wide variety of questions, except for the question asked. And none of the responses should be substantially similar to the actual Response received must be eliminated.

So what you have is the actual Response embedded in a group of 10 responses only one of which is relevant. Here is an example: -

(Question: "What is your name?")

Answers:

1. I find it colorful
2. But not as good as yesterday
3. Horrible horrible
4. Cats or dogs
5. Peter Smith
6. Krakhov wins
7. Tomorrow is another day
8. I cant see anything
9. Help yourself
10. Heaven's Door

We now give the questions and sets of responses to a panel of analysts, who are not required to know anything about the experiment and who could be simply a number of English-speaking students on a remote campus. Or the whole thing could be done on-line.

What the analyst has to do is choose which of the 10 responses shown to him/her matches the particular question asked.

And this is repeated for each of the ten questions.

Thus the subject has a 1 in 10 chance of getting any one response correct by chance.

With 10 such tests the odds against getting all responses right are 10 to the power 10 or 10 billion to 1.

Here, it is not a case of explaining some itsy-bitsy little physical event - the task that is before us is to explain: -

1. how someone heard a question that was asked,
1. when there was no one present, and
2. no means of communicating that question.

And,

2. who received and understood the question sufficiently to respond to it with a relevant answer.

With our interface unit we are docking right into extended reality.

We have connectivity.

Ideally, we would also perform the experiments in a Faraday Cage - but in doing so we are simply compensating for Cambridge University's dropped ball of thirty years ago.

What we are doing is showing that the question was received and that whoever received it was possessed of intelligence. In doing so we are in effect meeting Turin's criterion - although not in application to computers of course.

To get there some more development of the Alpha system will be useful and a new series of current experiments carried out to optimise responsiveness as a factor.

This paper just states the goal and a possible modus operandi.

Any comments will be welcome.