The main mistake in communication is assuming understanding has been reached too early
Two people are speaking the same language.
One asks a simple question. The second hears all the words perfectly, understands every single one, and answers immediately.
The answer is reasonable.
Yet completely useless.
Because the first person meant something else.
This happens all the time. Between strangers, between loved ones, at work, in families, in education. Sometimes the result is a few lost minutes. Sometimes—a wrong decision. Sometimes interlocutors start arguing and even conflicting, even though they could initially have agreed with each other.
With the advent of artificial intelligence, the problem has become even more noticeable.
AI is capable of giving an extremely detailed and persuasive answer a few seconds after receiving a question.
But a fast and good answer makes sense under only one condition:
the question was understood correctly.
We do not transmit thoughts
When one person says to another:
I need a new website.
a ready-made thought does not travel from their head into the interlocutor's head.
Several words travel.
Behind them, the first person may have a big picture.
The company's old website looks bad. Customers sometimes can't find the phone number. Prices have to be changed by asking a programmer. The director saw a competitor's nice website. They don't want to deal with technology. They just want the problem to disappear.
But the interlocutor receives:
I need a new website.
And begins to independently reconstruct the missing part.
If it's a web developer, they might think about design, the CMS, the server, development deadlines.
A marketer—about customer acquisition.
A fellow entrepreneur—about cost.
AI—about everything at once and offer a detailed website creation plan.
Everyone heard the same words.
But everyone built a slightly different thought behind them.
Meaning is not only in the words
Imagine two old friends.
One says:
That history with Sergey is happening again.
That may be enough for the other.
He knows Sergey. Remembers previous events. Knows the friend's character. Understands why he is annoyed. Perhaps even capable of roughly predicting what happened.
Five words reveal a huge amount of information.
Now say the same phrase to a stranger:
That history with Sergey is happening again.
There is almost no information.
The words are the same.
The difference is in the heads of the speakers.
Therefore, the effectiveness of communication strongly depends on how much they already have in common.
The better people know each other, the more they can shorten their messages.
Sometimes a glance is enough.
With a stranger, this method easily breaks down.
They do not have the necessary part to reconstruct the message—and they replace it with something of their own.
Humans almost always fill in the missing pieces
And this is not a mental flaw.
Without this ability, communication would become almost impossible.
If a person says:
Put the kettle on.
it is usually not necessary to clarify:
Which kettle specifically? Where to put it? Do I need to pour water? How much water? Do I need to turn on the electricity? What temperature should the water be?
Shared experience allows us to reconstruct the most likely meaning.
Most of the time this works great.
The problem arises when probability is mistaken for knowledge.
The interlocutor does not think:
I think I understood.
They think:
I understood.
And they continue the conversation, relying no longer on the other person's thought, but on their own reconstruction of that thought.
The smarter the interlocutor, the more interesting the problem
The ability to neatly fill in missing information seems like an advantage.
It is.
But it has a flip side.
A smart person can very quickly construct a plausible explanation for someone else's words.
AI can do this even better.
Having received a short query, a modern model is able to guess the context, intent, desired result, and even the format of the answer.
This allows it to answer surprisingly vague questions.
কিন্তু the same ability makes it possible to understand a person incorrectly with high quality.
And the better constructed the erroneous picture is, the harder it is to notice the problem.
A meaningless answer immediately causes an objection.
A very good answer to a misunderstood question can look convincing.
You can perfectly solve the wrong problem
Imagine a small company director.
He asks:
I don't understand programming at all. How long will it take me to learn how to make websites?
You can figure out his current knowledge.
Make a list of necessary topics.
Pick materials.
Break training into stages.
Make a personal plan for a year.
This will be a great answer to the question asked.
But you can ask first:
Why do you need to learn how to make websites?
And get:
I don't need to become a programmer. The company has an old website, I want to replace it.
Now the task looks different.
Perhaps modern tools already allow him to describe the desired website in plain words and get most of the result automatically.
He will only need to figure out a few things.
The annual training plan turned out to be a high-quality solution to a problem that didn't actually exist.
The person confused the desired result with the method of achieving it that was familiar to him.
And the interlocutor decided too early that they understood the request.
People ask based on the world they know
This is one of the main causes of the problem.
We cannot offer a solution whose existence we cannot imagine.
Therefore, a person often comes with the question:
How to do X?
Although their real thought is:
I want to get Y and assume that to do this I need to do X.
The difference is huge.
If you immediately start explaining X, the whole conversation will remain inside the person's worldview.
But perhaps there is Z, which they have never heard of.
And Z solves their problem in two days instead of a year.
This problem is especially pronounced now.
Technologies are changing so fast that a person's idea of possible ways to solve a problem can become obsolete in a few months.
What really needed to be learned yesterday can sometimes simply be delegated to a machine today.
Therefore, the question:
How to learn X faster?
increasingly deserves a counter-question:
And what result do you want to achieve with the help of X?
This applies not only to tasks
The same mistake arises in ordinary conversations.
A person says:
You never listen to me.
You can immediately start proving:
That's not true. Yesterday I listened to you for half an hour.
Formally, this can be a completely correct objection.
But what did the original phrase mean?
Perhaps:
It seems to me that my opinion doesn't matter to you.
Or:
I'm upset right now and want you to pay attention.
Or indeed:
You constantly interrupt me.
These are three different conversations.
But if you immediately choose one interpretation and start answering it, within a few minutes people might already be arguing about something neither of them intended to assert.
Then an unpleasant effect appears.
Each new utterance responds to the previous erroneous interpretation.
The distance between the speakers' real thoughts increases.
After some time, they are no longer arguing with each other.
Each is arguing with the image of the other person they have built in their own head.
Identical words do not guarantee identical meanings
People have shared dictionaries.
But internal meanings still differ.
A simple word like "expensive" for one person might mean:
I don't have that kind of money.
For another:
I can pay, but I don't think the purchase is worth that money.
For a third:
The competitor is cheaper.
For a fourth:
I'm afraid of making a mistake with a big purchase.
If you hear only the word and immediately classify the problem, it's easy to suggest the wrong solution.
And the more complex the concept, the greater the differences.
"Success."
"Safety."
"Justice."
"Love."
"A good job."
"A normal life."
Two people can argue about justice for an hour and only then discover they are calling different things by the same word.
Sometimes even the meanings of words are not what differs
The whole picture of the situation can differ.
One person knows event A.
Another knows B.
Both know C.
The first draws a completely logical conclusion X.
The second—equally logical Y.
And each considers the other's reasoning obviously flawed.
You can argue about X and Y for a long time.
Or you can discover:
Wait. Do you even know about A?
And the whole conflict will disappear in a minute.
Therefore, mutual understanding cannot be reduced to the correct interpretation of words.
You need to understand, at least approximately, from what worldview these words emerged.
Motive is also part of the meaning
There is another thing that is especially easy to lose.
Why is a person saying this at all?
The same question can have different reasons.
How much does a new phone cost?
A person might want to buy it.
Might be comparing it to their own.
Might be deciding how much money to gift their son.
Might just be surprised by the price.
Might be arguing with a friend.
The answer to the literal question is one.
The useful answer can be different.
The same thing happens with people's actions.
In the same external situation, the same person is capable of making different decisions because today money is more important to them, tomorrow time, and the day after peace of mind.
Therefore, it is impossible to fully understand a person's decision solely from external circumstances.
You need to take into account what they want right now.
But you cannot clarify everything
From all of the above, it's easy to make the opposite mistake.
Asking twenty questions before every answer.
What do you mean?
Why?
What is your ultimate goal?
What are the constraints?
What have you already tried?
How do you define success?
It is impossible to communicate like this.
Most human communication works wonderfully thanks to assumptions.
If a person asks:
What time is it?
you don't need to find out their life goals.
The task is not to stop making assumptions.
The task is to better feel the cost of a wrong assumption.
If an erroneous interpretation costs almost nothing—it's reasonable to act immediately.
If a year of work, a large sum of money, an important decision, or a potential conflict depends on it—a few seconds to check understanding can turn out to be extremely cheap.
Good communication is constant minor checking
Sometimes it's enough to rephrase the thought in your own words:
Do I understand correctly that you don't necessarily need to know how to make websites yourself—you just need a new company website?
And the person answers:
Yes.
Or:
No, I actually want to learn myself. I'm interested in it.
One short question divided two completely different tasks.
In another case:
So what upsets you is not the action itself, but the fact that you weren't consulted in advance?
Yes, exactly.
This is not a formality.
The speakers have just checked how similar thoughts are behind their words.
A good interlocutor is not obligated to agree
Understanding is often confused with agreement.
These are different things.
You can understand someone else's position very well and consider it erroneous.
Moreover, only after sufficient understanding does the opportunity arise to disagree with it substantively.
Otherwise, a person is arguing not with someone else's thought, but with their own version of that thought.
Therefore, the phrase:
I understand why you think so, but I disagree right here
can be a sign of a much deeper disagreement than an emotional argument.
In the first case, it is at least clear what exactly the disagreement is about.
For artificial intelligence, the problem is especially important
Modern AI is trained to be helpful.
A person asks a question—the system tries to answer.
A person asks for a plan—it builds a plan.
Asks for a comparison—compares.
Asks to write—writes.
This is natural behavior.
But as AI capabilities grow, the cost of misunderstanding increases.
If the system only tells, the error is limited to a bad answer.
If it is capable of making plans, making decisions, buying, writing programs, controlling other systems, and performing long sequences of actions, the situation changes.
Before very good execution, another question becomes increasingly important:
is the task understood correctly at all?
The more powerful the executor, the more expensive the magnificent execution of the wrong intent can turn out to be.
AI should not automatically consider the user's words the complete task
It is more useful to view the prompt as part of the information about the task.
Sometimes it is precise enough:
Translate this text into Spanish.
Sometimes not:
Teach me how to make websites.
In the second case, it is useful to understand:
why;
what result is needed;
why this particular path was chosen;
what the person already knows;
how much time and effort they are ready to spend;
what other options exist.
You don't have to ask about everything.
Sometimes a single "why?" question completely changes the solution space.
But AI shouldn't make decisions for the person either
Here lies another danger.
From the idea:
You need to understand the user's real goal
it is easy to derive:
AI knows better than the user what they actually want.
This is even worse.
The system also sees the person only partially.
It also makes assumptions.
And it can also be wrong.
Therefore, a more reasonable approach does not look like this:
You are asking for X, but actually you need Y.
But like this:
If I understood correctly, your ultimate goal is Y. If so, perhaps X is not needed at all. There is a shorter option. Would you like to consider it?
Do not replace the person's desire with your interpretation.
Check the interpretation together with the person.
Unfamiliarity should increase caution
Old friends have a huge amount of shared information.
A new interlocutor has almost none.
This also applies to AI.
If the system is talking to a person for the first time, a short phrase leaves a huge amount of unknowns.
What words do they use in an unusual way?
What is their experience?
What is obvious to them?
What do they consider a good result?
What constraints did they not name because they consider them self-evident?
As communication progresses, part of this becomes clearer.
Therefore, the exact same phrase from an old acquaintance and a complete stranger objectively contains a different amount of the unknown.
The less shared context there is, the more cautious one should be about the feeling:
I understood.
Perhaps understanding is not a "yes" or "no" state at all
We often say:
I understood.
As if there is a clear boundary.
But practically, understanding rather gradually increases.
First, the general meaning is clear.
Then the goal.
Then the reasons.
Then the details.
Then it becomes clear why the person chose those exact words.
And different tasks require different depths.
To answer:
Where is the nearest bus stop?
very little mutual understanding is enough.
To decide:
What business should I build for the next five years?
much more is required.
Therefore, the question is not always:
Did I understand the person?
Sometimes another is more useful:
Do I understand them well enough for the action I am about to take?
This is much more practical.
The more expensive the next step, the more important the check
From this follows a simple principle.
Before a small and easily reversible action, you can afford more assumptions.
Before a large and hard-to-reverse one—fewer.
You can answer a simple question immediately.
To make an annual plan—it's better to check the goal first.
To spend money—check the constraints once more.
To start a big project—make sure the result and reasons are understood the same way.
To accuse someone of something—perhaps check first whether their words actually meant what it seemed.
This applies to both people and machines.
Sometimes the most useful answer is another question
A good question is not a refusal to help.
Sometimes it's the shortest path to help.
Why?
What do you want to get in the end?
Did I understand correctly that...?
When you say "expensive", what exactly doesn't suit you?
Do you want to learn to do this yourself or just get the result?
What is most important to you here?
A few words can destroy a huge wrong branch of reasoning.
And after that, the answer often becomes much simpler.
Understanding takes time, but misunderstanding sometimes takes much more
The desire to move on to the answer faster is understandable.
Clarifications seem like an unnecessary delay.
But this is saving a few seconds at the cost of potential hours, months, or conflicts.
The more complex the world becomes and the more powerful our tools become, the more important the ability to first make sure of the following becomes:
are we even solving the same problem?
We will never be able to fully peer into the inner world of another person.
AI won't be able to do this perfectly either.
We will always have to fill in the missing pieces.
Without this, communication is impossible.
The problem does not begin when we make assumptions.
The problem begins when we stop noticing that this is just an assumption.
The main mistake in communication is assuming understanding has been reached too early.
