Digital history,
digital futures
King's PhD student Jacob Forward is thinking about the future of History and applying AI Large Language Models to historical research, while drawing on his human connections to empower young people in the age of AI.
In the Autumn of 2021, as a recent history graduate from Keble College Oxford, I was
working as a research consultant for a project at the School of Advanced Study (SAS) called
‘Covid Rumours in Historical Context’. This project compared contemporary themes of mis- and disinformation about Covid on Twitter to historical examples of vaccine and epidemic related conspiracies as far back as the medieval period. It was then, working alongside members of the University of London’s Digital Humanities Research Hub, that I first saw the power and potential of advanced natural language processing (NLP) techniques.
I had encountered ‘digital history’ as an undergraduate, but through the prism of my tutors
who came from a generation in the field that focused lots on digitisation, library science, and referencing technologies. I had not found these topics engaging. It was also at the SAS that a fellow researcher and digital humanities legend, Marty Steer, first mentioned large language models (LLMs) to me. He told me that a model called ‘GPT-2’ (one year before ChatGPT was released) was capable of writing poetry. This was electrifying news to me. The more I read about LLMs the more I felt a creeping sense of existential dread. I could clearly see that this technology, which was still only in its infancy but was all about generating and interpreting language, was on a collision course with the humanities — as inevitable as gravity.
“A deeper understanding of digital methods has immeasurably
enriched what I can do in my historical research, and has caused me to value all the more strongly (...) what humans bring to the equation that cannot be
automated or delegated to AI.”
I realised that there were two possible avenues to take: one of fear, denial, and disengagement, the other was to lean in, to learn everything I could, to develop my programming skills, and perhaps play a small role in shaping this collision between LLMs and the humanities in a beneficial direction. I began to teach myself programming in Python, and read everything I could about LLMs, Natural Language Processing, and digital history more broadly.
When I arrived in Cambridge the following autumn and began my masters studies, I also started a pilot project to demonstrate the potential of LLMs in historical research, which I used to apply for a PhD. My supervisor and peers were kind, but a little sceptical, about my excited jabbering about LLMs. The feelings of vindication were astonishing when ChatGPT (released in November 2022) soared to 100 million users in the following two months — the fastest growth of any product ever in human history up to that point. For the remainder of my masters year, the PhD project that would follow began to seem more and more possible and timely. I haven’t looked back since. A deeper understanding of digital methods has immeasurably enriched what I can do in my historical research, and has caused me to value all the more strongly the importance of humanistic inquiry, and what humans bring to the equation that cannot be automated or delegated to AI.
As for my interest in post-Cold War politics, I have always been fascinated by America: a great experiment of a nation, a sort of Petri dish of violence and idealism, whose growth to global influence has indelibly defined the course of human history for the last century and a half. I wrote my undergraduate thesis on the debate over national security and civil liberties from the Patriot Act after 9/11 to the Freedom Act under Obama. I found it exhilarating to work on the contemporary period, which has received comparatively little historical attention, and to feel like I was at the ‘coalface’ of history, uncovering something new. I continued this theme in my masters work which examined the relationship between terrorism and the expansion of federal authority in the 1990s, a period where bouts of serious terrorist violence have been near completely overshadowed by the totemic significance of the 9/11 attacks.
Building on my findings in this research, I expanded the scope of analysis in my PhD to look at the discourse of crisis in American politics from the end of the Cold War to Trump’s first election. I wanted to understand the rising appeal of an authoritarian-style of politics in 21st century America, and how this has shaped, and been shaped by presidential rhetoric. I’m planning chapters on terrorism before and after 9/11, natural disasters such as Hurricane Katrina, the Global Financial Crisis, and Trump’s ‘immigration crisis’ rhetoric in the 2016 election. An epilogue on the current administration’s crisis discourse is growing longer by the day! I’m fascinated by the extent to which people are willing to give up their liberties in exchange for the promise of ‘security’, and how politicians have increasingly instrumentalised moments of genuine crisis, and even confected or amplified crises, for their political capital — their potential to mobilise support, silence resistance, and centralise executive power.
My main source of information is The American Presidency Project, widely considered the definitive, non-partisan, online source of presidential documents and an essential companion to my studies since my undergraduate thesis. It contains everything from press conference transcripts, speeches, written orders, and even ‘Truth Social posts’. The site is also an example of a pioneering digital humanities project from the late 90s. I remain inspired by, and in touch with, its founders.
I have long been interested in the evocative power of figurative language. In my PhD work I study metaphors as ‘thought-framing devices’, as seen in George Lakoff’s scholarship. Research from the fields of psychology and even neurology shows us that non-literal language such as metaphors, similes, allegories, and other forms of figurative speech are not mere stylistic flourishes, they are profoundly powerful and persuasive. By one estimate, English speakers deploy a new metaphor roughly every 25 words that we speak. Even neuroscientists are interested in metaphorical speech and have used fMRI machines to reveal the complex neural substrate to non-literal language.
As a historian studying American politics, there is immense practical and legal significance to a president’s choice to speak about terrorism with a crime metaphor of perpetrators, investigation, trial and justice, versus a war metaphor of national security, ‘anticipatory self-defense’, lethal force, rendition, etc. Even within the overarching crime metaphor, describing terrorists as parasites, beasts/ ‘lone wolves’, or a virus, all inform different strategies for dealing with the threat, whether hunting- targeting, or containment, or addressing the problem ‘at source’. The rich metaphors that fill presidential rhetoric are both symptomatic of the administration’s thinking on a given issue, and often become determinative of their response, especially when figurative language percolates down through Congress into the statute books.
At a 2025 Alumni Festival event, I presented a project in which I had built an AI Roosevelt and an AI Reagan with which one could hold historically accurate conversations based verbatim on their original speeches. I was sharing a debate between these presidential AIs on some of the perennial themes of American foreign and domestic policy.
Some questioned the ethics of working with inscrutable ‘black box’ AI systems. To this I pointed out that we don’t judge the reliability of the speech and ideas of our fellow humans by studying the neural networks in their brains, we are black boxes of hidden bias as well. I also highlighted some exciting new areas of research in computer science that study activation patterns in digital neural networks and are slowly helping make the ‘black box’ of AI systems more transparent and interpretable.
Most attendees were very excited about the potential of the technology, particularly ‘semantic search’, which enables us to move from character- matching with keyword search, to asking human like questions and receiving human like answers… searching by meaning not by letter matching.
After taking part in the King’s E-Lab residential programme I began to see how my academic research could deliver social good through the vehicle of a business. Perhaps most importantly, I came away with a more entrepreneurial mindset that has been invaluable in my research, emboldening me to apply for more research grants, with better articulated value propositions, and understanding the benefits of networking and collaboration in what can be a rather solitary life as an academic historian.
Noticing the parallel challenges in information search and retrieval between history and law I was inspired to apply my programming skills alongside a team of three friends at the 2024 LLM x Law hackathon, co-organised by the E-Lab.
It was an outstanding opportunity to learn from and compete alongside teams from 28 of the top universities around the world, and I’m delighted to say we won first prize for our tool ‘Precedent AI’.
Many good stories begin with the line ‘One evening in the King’s bar’. That is where, in my first term at Cambridge, I started a conversation with a brilliant computer science student named Xiaochen (Neo) Zhu on the interesting possibilities of applying computer science techniques to historical questions. A few months later we had co-authored a paper, and he remains a dear friend and collaborator.
The dizzying interdisciplinarity of graduate life at King’s, and all of these wonderful spaces and occasions where we can come together and learn from each other, is surely one of its greatest strengths.
One of the biggest challenges I believe society faces in the next few years is learning to collaborate productively with AI as the technology plays an ever-greater role in education and the workplace. Part of the solution to this challenge is to offer high-quality AI literacy education to 16-18 year-olds at this crucial juncture before they enter university or employment. I am building a company, drawing on my entrepreneurial skills from King’s E-Lab, my experiences as a teacher on the Cambridge Generative AI in Business course, and my deep conviction as a historian that the next chapter of the human story should centre human dignity and values. Aime Education will offer online course content covering all aspects of AI from the technology itself to its social, psychological, and ethical implications, giving schools the resources to help their students excel as humans in the age of AI. I will be drawing on my network of fellow PhD students who are at the forefront
of AI research in their respective fields, to compile a well-rounded, and hopefully thought-provoking course. I would be delighted to hear from anyone in the King’s community who has any thoughts on this project.
It is hard to avoid discussions of ‘artificial intelligence’ today. It is even harder to separate hype from reality. Still, Jacob Forward's illuminating reflections on his own academic journey through ‘digital history’, natural language processing, and ultimately AI, offer a valuable perspective from what may yet seem an unusual quarter in this discourse: the humanities. As a new Fellow at King's in Digital Humanities and English, I would like to offer some additional context to these reflections. Digital Humanities (DH) is a relatively new field that has grown rapidly in the last two decades; it combines the tools and methods of computer science with the questions and methods of the humanities. A key motivation for the field is the use of digital methods to conduct ‘distant readings’ of texts, allowing for the analysis of large corpora of texts as a complement to traditional ‘close readings’ of canonical works. Jacob's work in this field is an excellent example of the potential of DH to enrich humanities research. I agree with Jacob that, as theorists from Walter Benjamin to Ruha Benjamin have argued, humanists cannot afford to leave foundational technological developments to purely technical or market-driven research. If AI has the potential to reshape cultural production, including the ways in which we understand history and process cultural memory – key sites for humanistic inquiry – then we must take a more active role in shaping its development.
Ryan Heuser, Fellow in English and Digital Humanities
Built with Shorthand