The Words Our Machines Cannot Find

Yoruba speakers, to the front, please.

I have an AI question for you about an expression.

Not for me, exactly.

For Dr. McMillan Cottom.

Dr. Tressie McMillan Cottom, the sociologist, writer, and recipient of the MacArthur “Genius Grant,” recently described a problem on her Instagram account involving a door, several bags, and a person trying to help.

Her hands were full. She needed to open the door without dropping what she was carrying. She worked out how to position the bags, reach the handle, and pull.

Then, just as she was about to open it, someone came through from the other side.

The person saw a woman with her hands full and did what a good Samaritan is supposed to do. They opened the door for her.

They also tripped over the bags.

Everything spilled.

The door locked.

The help made the situation worse.

Dr. McMillan Cottom wanted to know if there was a word for that.

Not simply bad help. Not an accident. Not malicious interference disguised as kindness.

Something more precise.

The kind of help offered by someone who sees your difficulty but not the solution you have already constructed. The kind that interrupts your competence, produces the very problem it intended to prevent, and leaves you carrying the consequences of someone else’s good intentions.

I knew the experience before I knew what to call it.

So did many of the people responding to her video.

Recognition came quickly.

The word did not.

Everyone Recognized It. No One Could Name It.

Some commenters offered familiar phrases.

“Intent doesn’t negate impact.”

“The road to hell is paved with good intentions.”

Others called it “inconvenient politeness,” the social pressure to accept or perform a gesture of courtesy even when the gesture disrupts what would have worked better.

Someone else described it as “unsolicited help,” which identifies the absence of a request but not necessarily the particular damage caused by the intervention.

One response pointed toward the judgment we expect from a “fully developed frontal lobe.”

Each phrase catches something true.

The helper had good intentions. The impact was still harmful. The assistance was not requested. The politeness became inconvenient. Better judgment might have prevented the whole thing.

But none of these expressions quite contains the situation.

“The road to hell is paved with good intentions” is a warning about intention without consequence. It is broad enough to apply to personal relationships, public policy, and almost every institutional failure ever explained as an attempt to do the right thing.

“Intent doesn’t negate impact” shifts moral attention toward the person who was harmed. It tells us how to evaluate what happened.

“Inconvenient politeness” captures the way a social expectation can turn a supposedly helpful gesture into an obstacle. “Unsolicited help” identifies the intervention’s lack of invitation, but it could describe assistance that is merely unwanted rather than assistance that actively makes things worse.

None names the entire mechanism Dr. McMillan Cottom described.

A person intervenes without understanding.

The intervention disrupts an arrangement that was already working.

Manageable difficulty becomes actual failure.

A phrase can explain the lesson without naming the experience.

That distinction matters.

Then Yoruba Entered the Conversation

Then one commenter reached into Yoruba.

The expression they offered was atọ́rọ̀ṣe bọ̀rọ̀jẹ́.

It was presented as a way of understanding what happens when an intervention intended to help leaves a situation worse than it was before.

I wanted to know what the expression carries beyond that English explanation.

Does it name the helper, the act of intervening, or the damaged result? Is it used seriously, humorously, or as a form of admonishment? What kinds of relationships and situations make it feel appropriate?

Finding an expression is not the same as understanding it.

Atọ́rọ̀ṣe bọ̀rọ̀jẹ́ has proved difficult for me to investigate beyond the original comment. Even reproducing it became part of the problem, one that points toward the constraints of North America’s innovation ecosystem.

At first, I entered it without the tonal marks and underdots used in written Yoruba. The expression became “atorose boroje.”

Search produced almost nothing.

I had assumed the search system would recognize what I was trying to spell. I had not understood that the tonal marks and underdots help determine how the words are pronounced and what they mean.

Without them, I had not simply entered an incomplete version of the expression.

I had entered something different.

Once I returned to the comment and copied its actual form, I could see what my transcription had failed to preserve:

atọ́rọ̀ṣe bọ̀rọ̀jẹ́

The marks were critical to the expression’s lexical significance.

They were part of the language.

I still want to know how Yoruba speakers understand the expression. Does it name the person who offers the help, the behavior itself, or what happens as a result? Is it colloquial, regional, or more common in speech than in writing?

I cannot yet say with confidence that it is the precise Yoruba expression for what happened at the door.

But my inability to retrieve it clarified the question at the center of this article.

The meaning seemed able to travel.

The expression almost did not.

An experience can cross languages before the expression naming it does. The distance between the two tells us something about what language preserves and what our systems are prepared to recognize.

The Experience Wasn’t Missing. The Word Was.

Linguists sometimes describe the absence of a direct equivalent as a lexical gap.

A lexical gap appears when one language has conventionalized a concept in a word or compact expression while another language needs a phrase, explanation, or approximation to communicate it.

That does not mean speakers of the second language cannot understand the experience.

English speakers did not need a single English word to recognize what happened at the door. The comments made that clear. People understood the social pattern immediately, even as they reached for different language to describe it.

What English seemed to lack was compression.

One phrase emphasized intention. Another emphasized impact. “Inconvenient politeness” emphasized social obligation. “Unsolicited help” emphasized the absence of invitation. Each assembled part of the meaning, but none retrieved the entire situation at once.

A conventional word or expression can do something different.

It can bring a pattern into the conversation before anyone has to reconstruct it.

This is one reason lexical gaps are revealing. Languages do not divide experience in identical ways. A distinction that remains diffuse in one language may have acquired a stable boundary in another. Research examining lexical gaps across hundreds of languages describes them as evidence of linguistic diversity, not proof that one language possesses an idea another cannot comprehend. Languages simply lexicalize different parts of human experience. Read: Research on lexical gaps across 699 languages

The gap is lexical.

It is not necessarily cognitive.

People routinely understand experiences for which they have no singular name. We improvise descriptions. We use analogy. We tell a story and watch recognition move across someone’s face.

Human understanding exceeds the dictionary.

Still, the presence of an expression can tell us something.

A Word Can Be a Form of Memory

Words are not only labels attached to experiences.

They are forms of social memory.

A conventional expression preserves a distinction that people have found useful enough to repeat. Once a pattern has a name, it becomes easier to notice again. It can be taught without recreating every detail. It can travel between conversations. People can dispute whether the expression applies, refine its boundaries, or warn someone that they are about to enact it.

Naming does not create the experience.

It changes what a community can do with it.

This is why the exchange underneath Dr. McMillan Cottom’s video became more interesting than a hunt for clever vocabulary. The commenters were not merely supplying synonyms. Together, they were showing how many ways there are to organize the meaning of a single event.

English offered moral principles and descriptive approximations.

A Yoruba commenter offered an expression that appeared to gather the experience more compactly, although understanding exactly how requires more than matching it to an English definition.

The expression might locate the meaning in the person who intervenes.

It might name the behavior itself or the damaged result.

It might carry humor, judgment, or a familiarity that does not survive translation.

Its force might change depending on who says it, to whom, and under what circumstances.

That uncertainty does not weaken the question.

It reveals why the question is difficult.

No language has to provide the definitive word. The more interesting possibility is that languages draw the boundaries of an experience differently. Each may preserve a different answer to the question of what matters most about what happened.

Was the problem the helper’s failure to understand?

The decision to intervene without being asked?

The destruction of an arrangement that was already working?

Or the final condition of something made worse by an attempt to improve it?

English speakers recognized all of these dimensions.

We needed several phrases to hold them.

Meaning Doesn’t Always Travel Word for Word

When an expression travels, we often ask what it means in English.

The question assumes that the meaning already exists intact and only needs a new label. Translation becomes a search for the corresponding item on the other side.

But languages do not always offer corresponding items.

A translator may have to choose which dimension of an expression to preserve: its literal meaning, emotional force, social judgment, humor, or relationship to the people involved.

Something becomes clearer.

Something else may disappear.

This is especially important when an expression appears to fill a lexical gap. The temptation is to celebrate it as “the phrase English does not have,” detach it from the community that uses it, and circulate it as an elegant piece of trivia.

The borrowed expression becomes a collectible.

Its meaning becomes a caption.

I am more interested in what the encounter reveals.

The English suggestions did not fail because their speakers misunderstood the event. They distributed its meaning across intention, impact, interference, competence, consequence, politeness, and consent.

The Yoruba response suggested that another language might gather those elements differently.

Not more truthfully.

Not more intelligently.

Differently enough to help us see what English makes us assemble piece by piece.

That is what cross-lingual discovery can offer at its best. Another language does not simply give us a more elegant name. It shows us that the boundaries around a concept were never inevitable.

Someone else may have drawn them somewhere else.

And once we see the distinction, we may begin noticing the pattern everywhere.

The colleague who “fixes” a process without asking why it works the way it does.

The institution that provides assistance by replacing the arrangements people built for themselves.

The product that removes friction while also removing control.

The policy that solves the problem visible to the person intervening and creates a new one for the person expected to live with the result.

The door is small.

The mechanism is not.

There is also a technical way to think about what happened in that comment section.

Multilingual machine-learning systems attempt something almost inverse to what I watched unfold. Rather than holding several possible descriptions beside one another and asking what each preserves, they try to abstract away from language itself and represent similar meanings in a shared semantic space.

That abstraction makes cross-lingual search and translation possible.

But it raises another question.

If an expression carries evidence of what a culture has learned to notice, what happens when a system is designed to treat its meaning as language-independent?