The Social Cost of Not Knowing
On ChatGPT, Amma, and the private relief of asking questions without being judged.
This post was born out of a conversation with Amma on a rainy Thursday night during my summer break at home.
Over the last year, I've spent an unreasonable amount of time thinking about AI.
I've watched engineers build agents that call other agents. I've seen students juggle multiple accounts to squeeze a little more usage out of coding assistants. I've met people who spend more on AI APIs in a month than some families spend on groceries.
Naturally, I assumed I had a reasonable understanding of how people use these systems.
Then I came home.
Amma is a government school teacher. She has spent decades teaching Kannada, English and Social Science to children. She has also served as a headmistress.
Like many people of her generation, she can read and speak English. She can navigate websites. She can fill out forms. She can get things done.
But increasingly, she uses ChatGPT.
Not for anything particularly exotic.
Mutual funds.
Bank applications.
Hairfall.
Government procedures.
Random questions that occur to her while going about her day.
At one point I asked her why she preferred typing over voice mode.
Voice mode seems like the obvious future. Speaking is supposedly more natural than typing. Every AI company appears convinced that we will eventually spend our days talking to computers.
Her answer surprised me.
"What if someone hears my question?"
I assumed she was joking.
She wasn't.
As a teacher and former headmistress, she explained, there are certain questions she feels awkward asking aloud.
Questions that feel too basic.
Questions she thinks she should already know the answer to.
Questions that might invite judgment.
Typing removes that friction.
Nobody hears the question.
Nobody knows what she doesn't know.
Nobody is standing nearby waiting to evaluate whether the question was intelligent enough.
The conversation remains private.
I couldn't stop thinking about that answer.
For years, I have viewed systems like ChatGPT as a way of making intelligence cheaper.
Need an explanation?
Ask.
Need a summary?
Ask.
Need help writing something?
Ask.
The value proposition seemed obvious.
Cheaper intelligence.
But perhaps that isn't quite right.
Information has been cheap for a very long time.
Google exists.
YouTube exists.
Wikipedia exists.
Entire university courses are available online for free.
The problem was rarely that answers were unavailable.
The problem was that asking questions often carried a cost.
Not a financial cost.
A social cost.
The older I get, the more I suspect ignorance has a strange economic structure.
Children are allowed to know nothing.
That is expected.
Students are allowed to know very little.
That is normal.
Young employees are allowed to ask basic questions.
That is part of learning.
But somewhere along the way, ignorance becomes expensive.
A headmistress is expected to know.
A manager is expected to know.
A parent is expected to know.
A senior employee is expected to know.
A respected member of the community is expected to know.
The role begins demanding certainty even when the person occupying it still has questions.
Perhaps especially then.
Many of the questions we never ask are not difficult questions.
They are embarrassing questions.
Questions about money.
Questions about health.
Questions about technology.
Questions about taxes.
Questions about forms.
Questions that everyone assumes everybody else already understands.
The internet gave us access to information.
But it never fully removed the embarrassment associated with ignorance.
In some ways, it amplified it.
A search engine assumes you know what you're looking for.
A search engine assumes you know which result to trust.
A search engine assumes you know enough to frame the right query.
A conversation is different.
A conversation allows uncertainty.
You can ask a bad question.
Then a worse question.
Then a question that contradicts the previous two.
Nobody rolls their eyes.
Nobody tells you that you should have paid attention in class.
Nobody reminds you that you're old enough to know this already.
The system simply responds.
Watching Amma use ChatGPT also reminded me of something she used to complain about years ago.
In schools, she occasionally used Nudi, a Kannada word processor.
She would laugh at some of the vocabulary it employed.
Words that were technically correct.
Words that belonged to Kannada.
Words that almost nobody seemed to use in everyday life.
One example that always stayed with me was "Hosa Kadata" for "New Document."
A perfectly valid translation.
Yet somehow unnatural.
The irony was remarkable.
Software designed for Kannada speakers often required Kannada speakers to learn a new dialect of Kannada.
It was localized.
But it wasn't comfortable.
Looking back, I wonder whether we have been misunderstanding technology adoption all along.
Perhaps people do not adopt technologies because they are intelligent.
Perhaps they adopt technologies because they make them feel safe enough to learn.
There is a reason Amma doesn't care about model benchmarks.
There is a reason she isn't interested in trying five competing AI apps.
When I asked why she didn't try Gemini or Claude, her response was immediate.
"This is working well. Why would I install something else and use up space on my phone?"
The answer sounds almost disappointingly simple.
Yet it contains a lesson.
People do not buy technology.
They buy relief.
Relief from confusion.
Relief from friction.
Relief from uncertainty.
Relief from having to ask someone else.
For all the discussion about artificial intelligence, I increasingly wonder whether the most valuable thing these systems provide has very little to do with intelligence at all.
Perhaps their greatest contribution is something much smaller.
A private place to ask questions we are embarrassed to ask elsewhere.
A place where we are temporarily allowed to not know.
And perhaps that is why they feel so useful.
Not because they make us smarter.
But because they make ignorance less expensive.