I think the first automated unit would be one that could assess lists....you know, 53's, LIC's, even custom lists created specifically for that user. Then, depending upon what "read", it could shunt them over to an appropriate call center\session body.
Well, an L1C is done to handle a sad (say) pc. What's the point of the pc going into an automated booth and finding out that he read on an L1C? He's now had a machine tell him that unless there were false reads he's probably sad.
Let's see whether this would work with automated metering, and why. Let's take a look at it. Guy has to be hatted as a pc in the first place. Let's say the booth has a screen that asks questions, complete with audio. I have a lot of experience with this, of course. Let's try and do a relatively standard Scn version of an LIC. Metered. Ahem.
Guy picks up the cans. Screen is relatively smart and responds to audio prompts, simple yes/no. Maybe there's a footpedal control since his hands are tied up. Various screens check he's sessionable and TA is in range. Screens have to accept his data that he's sessionable. We're gonna do this with existing relatively-simple technology, not pretend we've got software that notices from his skin colour or eye motions whether he's sessionable or lying or something fancy like that. Sensitivity could probably be set automatically. All OK so far, start of session.
An automatic meter can NOT recognize an F/N sufficiently well to be practical. It is feasible it could recognize an "instant read", in terms of a small fall or fall at the exact sonic end of an audio command. Personally, I have no idea at all if a meter will read instantly like this with automated commands. I do know that automated audio commands in a well-designed module work much better than one would think, and that such a session feels like a regular dual session and does not feel like a solo session. That is maybe counter-intuitive, but that's the reality of it. I'm very familiar with it as it's the basis of PaulsRobot online sessions. I've never tried metering on that basis and I don't know if it would work or not. Haven't a clue.
Let's take a big leap and say it will read OK. Let's try a simple L1C, done Method 3. We'll skip the first couple of questions for now. The screen shows the words at the exact same time as the audio voices them, or rather, at the end of the word or maybe a
very short time before. So question 3 appears alone on the screen, no other questions to distract the pc.
3. Has some emotion been rejected? Meter doesn't react. Pc doesn't squawk.
Next question is read out the same way. The previous question has disappeared off the screen. One question only at a time is visible, and even then not all at once until the very end of the line.
4. Has some affinity been rejected? Meter reads. Software notes the read, and pauses, effectively looking expectantly at the pc. Pc either notes something live in the bank and starts yapping, or says "NO" or hits the "next question" switch or whatever. If pc is yapping, pc keeps yapping until he feels all has been said. This could be automated in terms of if there is a silence for 5 (or 10, or whatever) seconds, the software prompts with "Is that all?" It wouldn't really be annoying as the pc would understand the limits of an automated program (it really does work like this re expectation. A pc does NOT get ARCXen over the "robotness" of a Robot Auditor if it is programmed well. IF.)
5. Pc tells the software he has said all. Note the software has no way of telling if the question is EP'd or not as it can't read pc indicators or F/Ns. The only way to do it is to have the PC decide if it is or not. Tricky, tricky, as the pc is not going to be a fully trained solo auditor. If he was, he wouldn't be screwing around with this in the first place. He may not be a very experienced pc either, so isn't able to recognize EPs. But this could hopefully be worked out on the basis of whether he feels any attention left on the question/area. If so, check earlier similar. If no obvious answer springs to mind, drop the question and go on to the next one.
6. Continue down the list in the same way. If the software doesn't take up a question and the pc feels he wants to answer the question, he has to be able to override the program and answer the question. The software must accept the answer as if it had read.
7. If the TA starts climbing, the software could pick that up, although maybe not as fast as an auditor would, and ask for an overrun. Good luck with that!
-----
I'm not going to continue with this. I just wanted to give the flavour of what would be needed if you wanted to automate something like this. Personally I wouldn't touch it at all, as there are far better ways of handling upsets automatically than to try and do a metered L1C!
Paul