Are We Losing the Vocabulary of Civilisation in the Age of AI?

One afternoon, sitting close to the window with my study table in front of me, I began trying to write a story. I was not attempting anything particularly ambitious. I simply wanted to put into words something I had been thinking about. However, as I began writing, I faced an unexpected difficulty. The English vocabulary available to me did not seem sufficient to express precisely what I wanted to say. I knew what I wanted to communicate, but the words seemed somehow inadequate. There were particular expressions I would naturally use when speaking in my native language, expressions that carried a certain emotional texture and cultural meaning, but I could not find their equivalent in English.

So I did what many of us now do. I opened Google Gemini and asked it to help me rewrite the passage using more Indian vocabulary.

That small incident made me wonder about something much larger: what happens to the vocabulary of a civilisation when increasingly large parts of our writing, communication and intellectual production are mediated by artificial intelligence?

At first, this may sound like a question about language. I think it is actually a question about civilisation.

We usually consider synonyms as interchangeable. If someone says, “I am tired”, we understand that they are not energetic. If they say, “I am exhausted”, we understand something similar, perhaps with greater intensity. But the two expressions do not necessarily belong in the same sentence, conversation or emotional context. Language does not simply communicate information; alongside it carries tone, intimacy, social position, historical experience and cultural memory.

This becomes even more apparent when we move across languages and civilisations. A word may have an apparent equivalent in another language without having an equivalent cultural meaning. The problem is not that translation is impossible. The problem is that something may disappear during translation precisely because we assume that words are containers of identical meanings.

Consider ordinary Indian forms of address. “Bhaiya”, “didi”, “behen”, “uncle” and “aunty” cannot always be understood simply as literal descriptions of biological relationships. They can express familiarity, respect, affection and social proximity. Similarly, a political leader addressing an audience as “bhaiyo aur behno” is not necessarily doing the same thing as a Western politician beginning a speech with “friends” or “fellow citizens”. The vocabulary establishes a particular relationship between speaker and audience.

Political language has always depended upon such cultural vocabularies. Leaders do not just communicate policies; they appeal to memories, identities and relationships through words. The choice between “citizens”, “people”, “brothers and sisters”, “comrades”, “friends”, or culturally specific forms of address creates different political worlds.

This is where artificial intelligence becomes interesting.

Large language models are trained on enormous quantities of existing human language. They learn patterns from the words, phrases, structures and associations that appear repeatedly in their training data. Their extraordinary strength comes partly from their ability to identify and reproduce these patterns. If a certain form of writing is dominant in the available corpus, it becomes statistically easier for the machine to reproduce it.

There is nothing inherently wrong with this. Indeed, it is one of the reasons AI can be so useful. It can help someone who struggles with English produce a polished academic paragraph. It can make an awkward sentence clearer. It can help a non-native speaker communicate with confidence.

But there may be a paradox here.

The better AI becomes at producing standardised, fluent language, the greater the possibility that our individual and culturally specific ways of expressing ourselves may gradually be replaced by statistically familiar forms.

The danger is not necessarily that AI will destroy languages. A more subtle possibility is that it may contribute to the erosion of linguistic difference within languages.

We may continue to speak Indian English, Nigerian English, Singaporean English or American English. But the deeper vocabulary through which particular communities interpret experience may gradually become less visible. Local idioms may be replaced by globally recognisable expressions. Unusual metaphors may be polished away. Culturally rooted phrases may be substituted with their closest statistically available equivalents. The emerging slangs among Gen-Z can be said to be one of the growing examples of this.

In that sense, AI may participate in a process that globalisation has already accelerated.

Globalisation has brought enormous benefits. It has connected markets, created supply chains and enabled societies to access products and technologies from across the world. There is nothing inherently undesirable about a shirt manufactured in Bangladesh, a phone assembled in Vietnam, machinery produced in China or another product coming from Cambodia. The extraordinary complexity of contemporary supply chains is itself an achievement of global economic organisation.

But alongside this economic integration, another form of integration is occurring: cultural and linguistic standardisation.

The local market now closely resembles the global market. The same brands appear in different cities. The same platforms shape our cultural consumption. The same visual aesthetics circulate through Instagram and YouTube. The same professional vocabulary is used in universities, corporations and governments. And increasingly, the same AI systems help us write our emails, articles, applications, speeches and stories.

The question, therefore, is not whether globalisation is good or bad. Nor is it whether AI is good or bad.

The more difficult question is: what should remain irreducibly local within a global system?

Can we say that a civilisation is not preserved only through monuments, temples, paintings, food or clothing. It also survives through its vocabulary, through the words by which people describe relationships, emotions, morality, nature, obligation, dignity, suffering and belonging.

From a sociological perspective, this raises a question similar to the importance sociology gives to social facts: can scholars of comparative civilisation studies take seriously the specific words and vocabulary through which a civilisation understands and expresses itself?

Prof. Greenfeld’s recent theory on civilisation emphasises first principles in the study of civilisation. I would distinguish the two rather than treating vocabulary as a first principle itself. First principles represent deeper assumptions about reality, morality, social relations and collective life, while vocabulary provides one of the ways through which these principles are expressed, transmitted and reproduced.

In this sense, vocabulary can become an important entry point into civilisational consciousness. When a culturally rooted concept is repeatedly replaced by an apparently equivalent global synonym, we may not simply be losing a word; we may also be losing access to part of the conceptual world, a social reality that each individual carries within the choices of their vocabulary.

Some civilisations do possess concepts that cannot be translated without loss. Indic traditions, for example, have developed extensive vocabularies around ideas such as dharma, karma, rta, moksha and seva. One can translate these words, but again post-colonialist and nationalists alleged that translation does not necessarily exhaust their conceptual worlds. The English word “duty”, for instance, may overlap with some dimensions of dharma, but it does not simply replace it. Neither does it seem totally aligned with the word religion.

This is why the preservation of vocabulary is not simply an exercise in linguistic nostalgia. It can become a form of preserving intellectual diversity.

Perhaps the greatest emerging threat of AI is therefore not that machines will stop us from speaking. It is that they will make it easy for us to speak in the same way.

A student in Delhi, a researcher in London, a consultant in Singapore and a political campaigner in Washington may receive assistance from systems trained on overlapping global corpora. Their circumstances will remain different, but the language through which they formulate their arguments may begin to converge.

We could then find ourselves in a strange paradox. Globalisation promised a connected world in which different cultures could communicate with one another. AI may make that communication even easier. But if communication requires the continual removal of linguistic particularity, we may eventually arrive at a world in which everyone can understand everyone else because everyone is using increasingly similar vocab.

The challenge, therefore, is not to reject AI or return to an imagined linguistic purity. Neither would be possible nor desirable. The challenge is to make AI capable of preserving difference instead of simply smoothing it away. Few would disagree that difference is, in itself, beautiful.

This is also where a broader reading of Hegel’s dialectical thought becomes relevant. If thesis and antithesis encounter one another, they generate the possibility of synthesis. Difference, disagreement and even opposing forces can therefore become productive rather than merely divisive. When difference is continually flattened into uniformity, however, the possibility of such creative synthesis may be weakened. A society in which contrasting vocabularies, ideas and cultural perspectives are progressively reduced to a single common form may gain efficiency, but it may also risk intellectual and cultural stagnation.

Then, we should ask a little more from our machines. When a word does not have an exact equivalent, perhaps the machine should tell us so rather than quietly replacing it with the nearest familiar word. Sometimes the unfamiliar word may be carrying something with it – a memory, a relationship, a way of seeing the world– which the translation cannot quite carry.

This brings me back to that afternoon by the window. I could not find the English word I wanted. At first, I thought my English was insufficient. But it seems now that was not the problem. Infact I was trying to find a word which could do the work of another language.

And this leaves me wondering about AI. We speak so much about technological independence, about who controls data, technology and strategic resources. We worry about nuclear weapons; we worry about economic dependence. Should we also begin to think about the language of the machines through which we will increasingly write and communicate?

I do not mean that every country needs its own AI, or that technology should become another reason for the world to divide itself. But perhaps we should be careful about allowing one vocabulary, one style of expression, and one way of making sense of the world to become the invisible standard.

Maybe an AI trained to understand the languages and cultural vocabularies of different societies would be a more humane AI, particularly for those of us who do not think in English first.

I began this piece because I could not find the right word. I am still not sure whether I have found it.

That is the point.

Some words may be difficult to translate precisely because they belong to a particular place, history or civilisation. And perhaps we should not always be in such a hurry to replace them.

I leave the thought here, rather than trying to resolve it. I would genuinely like to know how others see it.