Firstly, I'm 100% against torture and agree its ineffective. But your rejection of the logic here is incorrect.
> In this situation there's no reason the person isn't going to lie for an extended period of time about A, B, C and D. So why wouldn't they then lie about E?
You seem to be thinking about this in chronological sequence, as if we ask about A then B then C then D then E. If that was the case, then yes, your logic would work. They might have lied about A for a long time, then B for a long time, then C for a long time, then D for a long time, then E for a long time and you'd never know if they were telling the truth about E.
But that's not what is being proposed. We ask about A, B, C, D, and E altogether. As in you have conversation and asked them about A, B, C, D, and E before doing any torture. If they lied about A, B, C, or D we torture them. Then we have the same conversation again, and torture them again. Eventually, they start telling the truth about say C. Why C? They are just choosing something at random to tell truth about in hopes of making the pain stop. But nothing changes, we keep torturing them because we know they are still lying about A, B, and D. Eventually they also start telling the truth about A, then E, then B, then D. Then we stop torturing them because they are being honest about everything we can verify. They can escape without divulging the information only if they keep E until last.
The key assumption here is that the person being tortured reveals the truth in a random order. It is improbable that they happen to keep the one thing we cannot verify until the end.
This is not to say that this procedure would be effective, but its not ineffective for the reason you suggest.
On the theory side, the probability of keeping E until the end is 1/5, which isn't that low.
On the practical side, this requires that the person being tortured has no idea what you can and can't verify. If, instead, they have some idea of what you can and can't verify, they'll be able to keep the hardest to verify information till the end preventing this scheme from working.
> In this situation there's no reason the person isn't going to lie for an extended period of time about A, B, C and D. So why wouldn't they then lie about E?
You seem to be thinking about this in chronological sequence, as if we ask about A then B then C then D then E. If that was the case, then yes, your logic would work. They might have lied about A for a long time, then B for a long time, then C for a long time, then D for a long time, then E for a long time and you'd never know if they were telling the truth about E.
But that's not what is being proposed. We ask about A, B, C, D, and E altogether. As in you have conversation and asked them about A, B, C, D, and E before doing any torture. If they lied about A, B, C, or D we torture them. Then we have the same conversation again, and torture them again. Eventually, they start telling the truth about say C. Why C? They are just choosing something at random to tell truth about in hopes of making the pain stop. But nothing changes, we keep torturing them because we know they are still lying about A, B, and D. Eventually they also start telling the truth about A, then E, then B, then D. Then we stop torturing them because they are being honest about everything we can verify. They can escape without divulging the information only if they keep E until last.
The key assumption here is that the person being tortured reveals the truth in a random order. It is improbable that they happen to keep the one thing we cannot verify until the end.
This is not to say that this procedure would be effective, but its not ineffective for the reason you suggest.
On the theory side, the probability of keeping E until the end is 1/5, which isn't that low.
On the practical side, this requires that the person being tortured has no idea what you can and can't verify. If, instead, they have some idea of what you can and can't verify, they'll be able to keep the hardest to verify information till the end preventing this scheme from working.