History in the age of artificial intelligence

s a discipline, history is facing both an extraordinary opportunity and a serious challenge. We have entered an age in which a machine can search millions of pages, translate languages, recognise handwriting, arrange information, summarise books, suggest references and produce fluent prose within seconds. A student can ask a machine about the Mughal Empire, the Pakistan Movement, Partition, colonial Punjab, or the life of Jinnah and receive an answer almost instantly.

This represents a significant technological development. It also raises a fundamental question: if a machine can search, collect, translate, arrange and summarise - even write, what remains for the historian? My answer is simple: almost everything that matters most in history remains the work of the historian.

History is not the same thing as information. Information tells us that something happened. History asks why it happened; how we know that it happened; what the people involved thought they were doing; what alternatives were there; and how later generations understood the event. EH Carr described history as “an unending dialogue between the present and the past.” The past does not present itself to us as a complete and self-contained narrative. It comes in fragments: documents, letters, diaries, newspapers, government files, court records, photographs, maps, memoirs, buildings, objects and oral memories. The historian brings these fragments together and gives them meaning. AI may help us find the fragments, it cannot relieve us of the responsibility of understanding them.

Marc Bloch reminded us that ignorance of the past weakens our understanding of the present. He also taught us to question the evidence. Who produced this source? When and why? For whom? Under what circumstances? What did the writer know? What did the writer not know? What did the writer wish to conceal? What assumptions lie behind the language? What is missing? These questions are the foundation of historical method. They cannot be replaced by speed, however impressive that speed may be. We should therefore neither fear AI nor treat it uncritically; we should learn to use it responsibly.

Technology can help us search large archives, compare newspapers, transcribe difficult handwriting, organise material, identify patterns, prepare preliminary translations and locate possible sources. These functions can provide valuable scholarly assistance. The historian who refuses every new technology is not necessarily defending the discipline.

But there is an equally serious danger. The same technology that assists research can weaken the habits that make research possible. It can make students impatient with archives, dependent on summaries, careless about sources and less willing to read difficult books and documents. It can produce fluent paragraphs that contain wrong dates, false quotations, invented references or misleading interpretations. The greatest danger is, therefore, not that the machine will become the historian. It is that the historian will slowly become a machine operator. We must, therefore, return to the distinctive craft of the historian.

A historian is not simply someone who knows many dates or can produce many pages. A historian is trained to make responsible judgments about evidence. That requires patience, doubt, comparison, language, imagination, discipline and the courage to say, “The evidence does not allow us to say.” This sentence may become even more important in the age of AI. We are entering a world in which information will become abundant and certainty will become easy. History must defend the value of doubt. A machine can give an answer in five seconds; a historian may need five months to establish whether that answer is correct.

That is not a weakness. That is scholarship. These considerations become particularly significant in the context of Pakistan. The problem of history in Pakistan is older than artificial intelligence. The AI has simply arrived at a time when our historical discipline is already facing important questions. We have inherited colonial records and colonial categories, nationalist and ideological histories, official narratives, regional histories, family memories and competing interpretations of the Pakistan Movement and Partition. Our task is not to discard one narrative and replace it with another. Our task is to examine all narratives critically and place them against evidence.

The colonial archive is a good example. British rule produced an enormous record of South Asian society. Officials counted people, classified communities, surveyed land, mapped territories, prepared censuses, wrote administrative reports, and recorded political events. These sources are indispensable, but they are not neutral mirrors of society. They were produced for particular administrative purposes and reflected the concerns and categories of the colonial state. When a colonial officer described a community, he was not only describing that community; he was also revealing how the colonial government wished to understand, classify and govern it. The Pakistani historian must use the colonial archive without becoming captive to its categories.

The same rule applies to sources produced after independence. A government file, newspaper, political speech, memoir, textbook or official report may be extremely valuable, but none should be automatically treated as the final truth. Every source must be questioned and placed in context. This is particularly important in writing the history of the Pakistan Movement. The movement produced a vast record of speeches, letters, newspapers, resolutions, political correspondence, biographies, memoirs, British reports, provincial records and private papers. The historian must resist turning this complex history into a simple story. The Pakistan Movement was experienced differently in Delhi, Lahore, Karachi, Peshawar, Quetta, Sylhet, Aligarh, Lucknow and the Punjab countryside. Provincial politics, class, land, language, religion, institutions, electoral interests and local conditions all mattered. The history of the movement cannot, therefore, be reduced to the actions of a few great leaders, however important those leaders were.

We need to ask what students, traders, peasants, workers, women, religious scholars, minorities and ordinary families understood by Pakistan. We need to ask how political ideas travelled from speeches and newspapers into towns and villages. We need to move, in other words, from a history of political leadership alone towards a wider social history. The history of Partition makes this methodological need even more evident. The division of British India in 1947 was not just a constitutional event.

It was a human experience of enormous scale. Families were separated; homes were abandoned; cities and villages change; property changed hands; languages travelled with migrants; and memories were created and lost. Partition must be studied through political decisions and official records; but also through the experiences of refugees, families, workers, women, villages, towns, railway stations, refugee camps and local communities. Oral testimony is valuable, but memory must also be examined critically. We must respect memory without assuming that memory is always exact.

We must also recognise that historically Pakistan has never been uniform. Regional histories are not footnotes to national history; they are part of the national history. Histories of the Punjab, Sindh, Balochistan, Khyber Pakhtunkhwa, Gilgit-Baltistan and the history of Bengal and East Pakistan all contain experiences essential to understanding Pakistan. A national history that ignores regional experience will always remain incomplete. We must, therefore, study the village as well as the capital; the district as well as the state; and ordinary people as well as political leaders.

This also requires us to return to the languages of our past. Much of South Asian history was not written in English. Persian was for centuries a major language of administration and scholarship. Urdu became a major language of political and intellectual life. Punjabi, Sindhi, Pashto, Bengali, Balochi, Seraiki and other languages contain historical memories and sources that cannot always be fully understood through translation. A historian who depends entirely on translations is often dependent upon someone else’s interpretation of the evidence. Language is not merely a means of reading words. It carries concepts, meanings, assumptions, and historical worlds.

AI can help, but it cannot remove the need for language learning. Machine translation may open a door to an unfamiliar source, but it can also produce a sentence that is linguistically smooth and historically wrong. The machine may translate the word; the historian must understand the world in which the word was used. The question of historical training becomes even more important when we consider our present academic culture.

The writer is a professor in the Faculty of Liberal Arts at the Beaconhouse National University, Lahore.