How can we know our last sentence was heard?

[DarkFalz]

When we are having a conversation, the response/comment from the other person to our last sentence acknowledges that the other person heard our sentence.

For instance, if person A and B are having a conversation and it goes like this

A - Do you want a beer?
B - Sure (at this point A knows B heard his sentence)

But B needs A to say something to know that he heard his response.

Since there is always one last sentence for which there won't be an answer, how do people know that the last sentence was heard?

Anyone willing to think about this with me? It's strange how it works even though this limitation exists.

 

[DaveC426913]

Yup. I have thought about this too. Even recently.

The key is that there are two types of message
- for some steps in the exchange, an answer is expected, and subsequent actions are determined by the reception of the response (such as a second attempt to establish communication).
whereas
- for some steps in the exchange, an answer is not expected, and - whether or not one is received - it does not change subsequent actions.

I have this kind of encounter when I text my wife to let her know I've arrived to pick her up from work.

Me: "I have arrived and I am waiting for you downstairs."
I require a response because it is quite possible she has not got my message, and I might be sitting there all night. My actions (such as texting again) are dependent on my reception and content of her message.

She: "I am on my way down."
She does NOT require a response, because - regardless of whether or not I got her message - she is going to perform the same action: to come downstairs. Her subsequent action is not dependent on the reception of a message from me.

 

[DaveC426913]

But B needs A to say something to know that he heard his response.

No. B does not need verbal confirmation that he's been heard. He merely needs to observe whether or not A gets him a beer.

 

[Vanadium 50]

But B needs A to say something to know that he heard his response.

Can't A just give him the beer?

 

[epenguin]

"Thank you"
"Thank you for saying thank you"
"Thank you for saying thank you for saying thank you"
"Thank you for saying thank you for saying thank you for saying thank you."
...

... That must be the answer - dot dot dot.

 

[Evo]

When we are having a conversation, the response/comment from the other person to our last sentence acknowledges that the other person heard our sentence.

For instance, if person A and B are having a conversation and it goes like this

A - Do you want a beer?
B - Sure (at this point A knows B heard his sentence)

But B needs A to say something to know that he heard his response.

Since there is always one last sentence for which there won't be an answer, how do people know that the last sentence was heard?

Anyone willing to think about this with me? It's strange how it works even though this limitation exists.

The response can be non-verbal. If you want to know if your last statement was heard, simply ask, did you hear - and repeat the last statement, this isn't something you need to wonder about.

 

[dipole]

I thought this was going to be a slightly more interesting thread - "How can we know our last sentence, i.e. dying words, can be heard?".

 

[collinsmark]

The OP touches on a problem very important to network communication theory.

It's usually introduced to computer science students in the form of a thought problem. I wish I had a reference to this (it's a classic, textbook problem), but my google-fu isn't cooperating today. Anyway, it goes something like this:

Two blue armies wish to defeat the evil red army. The two blue armies are on opposite sides of the red army. The blue armies can easily defeat the red army, but only if they attack together, simultaneously. If they fail to attack simultaneously, the red army would easily defeat one of the blue armies followed by the other.

The two blue armies can communicate with each other only by sending courier spies from one side to the other, by going through the red army. However, there is a small but finite possibility that a courier spy can be captured by the evil red army as it passes through. (And there's no going around the red army -- in order to pass a message, the courier must pass through the red.)

So the question is, "is it possible for the blue armies to defeat the red army with 100% certainty?"

And the answer, it turns out is "no." It can be done with a fair amount of certainty, but never 100%.

A good starting point might be for the right side to send a courier over to the left side with the message, "We attack Tuesday at dawn. Send a courier spy back this way with an acknowledgment that you received this message. We won't attack if we don't receive the acknowledgment."

Suppose that both couriers made it across successfully. The right side blue army is now ready to attack Tuesday at dawn. But the left side gets to thinking, "hmm. Ya know, I'm not sure if my acknowledgment courier spy made it across. If he was captured, the right side won't attack and we'll be defeated."

So instead of sending back a simple acknowledgment, the left side sends back this message, "We got your message to attack Tuesday at dawn. This is the acknowledgment message that you requested. Please send another courier spy back this way acknowledging that you received this acknowledgment. If we receive the acknowledgment of this message, we will also attack Tuesday at dawn."

And continuing that thought process, it's soon realized that in order to attack with 100% certainty, an infinite number of courier spy messages are needed.

[Edit: And what's more, since there is a finite probability that any given courier spy would be captured, and since it only takes a single capture to foil blue's plans, the probability of blue succeeding tends toward 0 with increasing number of acknowledgments. So a vast number of acknowledgment messages is definitely not the best approach.]

The real-world application to this is network protocols such as TCP-IP. Acknowledgement messages are built into the protocol. With single a acknowledgment it's reasonably certain that the desired information was transferred successfully. But it can never be known with 100% certainty.

 

[Evo]

The OP's post seems to be on a personal worry level of a question being heard. Such is life, it's to be expected.

 

[rootone]

It seems to me that this is all about having a reasonable level of certainty.
If I am in a bar and my partner says she will get the next drink, it won't worry me if I don't see her for say up to 15 minutes.
If I don't see her in 30 minutes I'll definitely be concerned, and I am sure she would as well if the situation was reversed,

 

[collinsmark]

The OP's post seems to be on a personal worry level of a question being heard. Such is life, it's to be expected.

I kinda figured. But it's a fascinating topic! And it's good to know that there are people out there that wrote graduate theses on the subject!

[btw, I didn't write a graduate thesis on this subject. I'm just saying that there are those who did.]

 

[Evo]

I kinda figured. But it's a fascinating topic! And it's good to know that there are people out there that wrote graduate theses on the subject!

Maybe that's why my thesis failed. "I sent a pigeon with a message, it didn't return. I'm stressed out. The end. "

 

[zoobyshoe]

Ground Control to Major Tom
Your circuit's dead,
there's something wrong
Can you hear me, Major Tom?
Can you hear me, Major Tom?
Can you hear me, Major Tom?
Can you...

 

[phion]

What does their body language say?

 

[OCR]

Such is life, it's to be expected.

"Copy that"... lol

 

[DaveC426913]

I propose we modify the question.

How can we know a thread was read by the opening poster of the thread?

 

[Ryan_m_b]

In most conversations if someone didn't hear the last thing you said it either doesn't matter or if they did you'd know because they repeat their question. E.g.:

Me: Are you going to the bar?

Flatmate: Yeah, why?

Me: Grab me a beer

Flatmate: Cheeky git. What do you want?

Me: I'll have that fruity ale

Now in this situation if I'm totally not being observant I won't know if my flatmate has heard what drink I want. Except I know that he, like many people, if they didn't hear would ask again. After all from his perspective he asked what I want and didn't hear an answer, he's not likely to just wander off and think I changed my mind because if I had I would have said. In short: we often know when the last message has received said because if it wasn't we'd get another message clarifying.

EDIT: For an easier explanation take a look at this flow chart. It details the decision making process of my flatmate asking me for a drink. We've never discussed this, we just know each other well enough to know this is what we would both do.

Following this if he hasn't heard my last message (asking what drink I want) he'll ask again. I know this so if I don't hear from him again it's probably because he's going to get me a drink, if not I can expect him to ask again. He won't ask indefinitely though because if I can't hear him and he can't hear me we're not going to send messages back and forth for hours (the bar will close!) so there's a reasonably cut off of "if I ask three times and get no answer you're not getting anything". If I failed to hear his question for what beer I want three times then I'm probably not paying enough attention and I don't deserve a beer.

 

[BobG]

The OP touches on a problem very important to network communication theory.

It's usually introduced to computer science students in the form of a thought problem. I wish I had a reference to this (it's a classic, textbook problem), but my google-fu isn't cooperating today. Anyway, it goes something like this:

Two blue armies wish to defeat the evil red army. The two blue armies are on opposite sides of the red army. The blue armies can easily defeat the red army, but only if they attack together, simultaneously. If they fail to attack simultaneously, the red army would easily defeat one of the blue armies followed by the other.

The two blue armies can communicate with each other only by sending courier spies from one side to the other, by going through the red army. However, there is a small but finite possibility that a courier spy can be captured by the evil red army as it passes through. (And there's no going around the red army -- in order to pass a message, the courier must pass through the red.)

So the question is, "is it possible for the blue armies to defeat the red army with 100% certainty?"

View attachment 84353

And the answer, it turns out is "no." It can be done with a fair amount of certainty, but never 100%.

A good starting point might be for the right side to send a courier over to the left side with the message, "We attack Tuesday at dawn. Send a courier spy back this way with an acknowledgment that you received this message. We won't attack if we don't receive the acknowledgment."

Suppose that both couriers made it across successfully. The right side blue army is now ready to attack Tuesday at dawn. But the left side gets to thinking, "hmm. Ya know, I'm not sure if my acknowledgment courier spy made it across. If he was captured, the right side won't attack and we'll be defeated."

So instead of sending back a simple acknowledgment, the left side sends back this message, "We got your message to attack Tuesday at dawn. This is the acknowledgment message that you requested. Please send another courier spy back this way acknowledging that you received this acknowledgment. If we receive the acknowledgment of this message, we will also attack Tuesday at dawn."

And continuing that thought process, it's soon realized that in order to attack with 100% certainty, an infinite number of courier spy messages are needed.

[Edit: And what's more, since there is a finite probability that any given courier spy would be captured, and since it only takes a single capture to foil blue's plans, the probability of blue succeeding tends toward 0 with increasing number of acknowledgments. So a vast number of acknowledgment messages is definitely not the best approach.]

The real-world application to this is network protocols such as TCP-IP. Acknowledgement messages are built into the protocol. With single a acknowledgment it's reasonably certain that the desired information was transferred successfully. But it can never be known with 100% certainty.

Send ten messengers, each with a portion of the message sliced vertically - kind of like a scytale, except each messenger would have only a tenth of the letters instead of all of them. Even if one scroll is missing because a messenger was caught, the message is almost certainly readable. It's impossible to reconstruct the message from one scroll.

Note, the scytale has to use multiple scrolls that wind around the scytale - not simply multiple cylinders. If you were lazy enough to use cylinders, the red army would know the diameter of the scytale and simply send one of the blue armies a bogus message in order to provoke an attack that would surely fail.

Send the acknowledgment of the message the same way.

In fact, the blue armies should be sending messengers back and forth each night even when they have nothing to say. Otherwise, the mere presence of messengers tips off the red army that something is up.

And if you're using this method to ask a girl in a bar if she wants a beer, it will usually elicit some unusual responses.

But it is somewhat similar to Reed-Solomon encoding, without which it would be impossible for CDs to last more than a few uses, as some scratches on the CD surface will definitely occur.

 

[russ_watters]

For many conversations that only convey information and don't end in action, saying "thank you" and "you're welcome" eliminates this issue.

 

[DaveC426913]

Except I know that he, like many people, if they didn't hear would ask again.

True, although this is adding another factor into the equation. Namely, a message is received but is corrupted.

That is qualitatively different than the OP's scenario where - more than not knowing what was sent/received - you don't know if something was sent received.

Getting a garbled message is a trivial scenario, since it is obvious what to do about it.

 

[russ_watters]

Getting a garbled message is a trivial scenario, since it is obvious what to do about it.

Yes, clearly you should bring him a cow.

 

[collinsmark]

Send ten messengers, each with a portion of the message sliced vertically - kind of like a scytale, except each messenger would have only a tenth of the letters instead of all of them. Even if one scroll is missing because a messenger was caught, the message is almost certainly readable. It's impossible to reconstruct the message from one scroll.

Note, the scytale has to use multiple scrolls that wind around the scytale - not simply multiple cylinders. If you were lazy enough to use cylinders, the red army would know the diameter of the scytale and simply send one of the blue armies a bogus message in order to provoke an attack that would surely fail.

Send the acknowledgment of the message the same way.

In fact, the blue armies should be sending messengers back and forth each night even when they have nothing to say. Otherwise, the mere presence of messengers tips off the red army that something is up.

I may have done a poor job at wording the problem. Gosh, I wish I could find an online reference to this. It's a classic and common problem and has been studied in depth; it's got to be out there somewhere. I just can't seem to google it.

Anyway, to clear up a couple of things (to keep things simple):

• Red does not have any double spies that could pose as blue spies. Red cannot not send any messages.
  
• If a blue spy courier gets captured, he gets captured and that's that. Red does not retaliate or react in any other way. It just means the message failed to get through.
  
• Red only retaliates once one of the blue armies attacks.

The doubt -- i.e., the less than absolute certainty -- comes from the fact that the side sending the most recent message never knows for sure if it arrived successfully. There is always at least a small possibility that a given side will attack, while the other side canceled planned attack, or didn't make plans to attack in the first place, due to a lack of acknowledgment when one was expected, or if the original message didn't get through and no acknowledgment was expected. [Edit: And if there is the case where blue never attacks, blue still fails to succeed in their goal.]

The "best" strategy invariably involves agreeing to keep the acknowledgments to a small number (if any at all), which means that one way or another, one side will have to just send a message and hope for the best that it got there, without knowing for certain.

But it is somewhat similar to Reed-Solomon encoding, without which it would be impossible for CDs to last more than a few uses, as some scratches on the CD surface will definitely occur.

Forward error correction is a lovely thing. It doesn't completely solve the problem here though, but it is a lovely thing.

[Edit: Yeah, I get it that sending redundant messages increases the chances of the particular message information getting through the channel. I do agree on that. But it still can't be done with 100% certainty and a finite number of messengers. Adding redundancy changes the numbers a bit. But it doesn't really change the nature of the problem.]

 

[collinsmark]

Found it!

http://en.wikipedia.org/wiki/Two_Generals'_Problem

Golly, why was that so difficult for me. ?!

 

[BobG]

I may have done a poor job at wording the problem. Gosh, I wish I could find an online reference to this. It's a classic and common problem and has been studied in depth; it's got to be out there somewhere. I just can't seem to google it.

You worded it fine. But it is a seemingly simple problem that truly is important to modern communications (even with extra variations added in).

I just wanted to actually use one of the words I heard on the National Spelling Bee...

... plus FEC really is a lovely thing.

 

[rootone]

If the last communication indicates that the message was received and no further information is required, that's a wrap.
It could be something like a simple 'thumbs up' or similar.
Both parties now expect things to unfold in an agreed manner, (not that it necessarily does).

 

[DarkFalz]

Thank you for all the answers so far. I am also a computer scientist, so my reasoning about this problem came from the fact that the two generals problem can be applied to everyday conversations. Regarding everyday conversations I believe it works like this:
- if A asks B something like X, B can be certain that A will hear the answer Y because A asked it! Otherwise it would be rude to ask something and then ignore the answer.
- In this situation, B can be sure that A heard the answer, and A can be sure that B knows that A knows because he knows that B knows this fact

This means that questions represent agreement points in a conversation, where both persons know that the other one knows X.

What do you think?

Still, the problem remains for sentences that are not questions. What do you say?

 

[Evo]

This thread has been answered, no need to go in circles.