What role do intention, generation, and reception play in creativity? When does an artefact become art? James de Lullington responds to the Interdisciplinary AI workshop held at the Institute of Medical Humanities in June 2026.
On Friday 12th of June, I arrived at the Institute of Medical Humanities’ ‘Narrative and Cognition Lab’ workshop on ‘Interdisciplinary AI’. The topic? To explore themes of creativity, agency, and knowledge, and to illuminate the complexities introduced by Artificial Intelligence.
The event began with a short address from the chair of the lab, Marco Bernini, followed by Caroline Bassett’s presentation, Where She Ended, She Had No Idea; and David Watson’s On the Philosophy of Unsupervised Learning. The day ended with Adam Linson’s Art and Agentic AI: Bathos in the Void of Historicity. The event consisted of a structured dialogue between speakers and attendees both online and in the room. By the end of the third talk, questions had compounded in the minds of the audience and there were more questions than could be taken. My contribution was not among those, so I shall use this opportunity to respond, extend, and perhaps revive the dialogue had on that Friday afternoon. It is in part a review of the event, but also my own entry into the conversation.
What is interdisciplinary about AI?
Bernini opened by asking what was meant by ‘Interdisciplinary AI’? He sketched an answer made up of several strands, relating to several disciplines:
- What counts as authorship or creativity?
- What counts as learning?
- How many agents are present when a human and a machine act together?
- Is the mind is extended through these tools or is it offloading its work towards atrophy?
- What is the role of human imagination and fictionality?
The workshop sought to explicitly turn the AI debate to humanities and social science disciplines and challenge a series of oft-made assumptions: machines cannot be conscious, AI output lacks creativity, and unsupervised machine learning is entirely devoid of the ‘human’. Finally, Bernini offered that large language models are themselves interdisciplinary, trained by breaking down the silos we take such care to defend. I would venture that they are undisciplined. Physics papers, Victorian novels, and legal judgments all enter the same system, without the divisions that normally govern such work. One reliable way to be outsmarted by LLMs, Bernini suggested, is to keep compartmentalising knowledge.
The three following speakers approached the subject from distinct, yet related, angles. Bassett opened from literary theory, refusing what she cast as a temptation to re-fight the “structuralist wars” (Bassett, 2026). This, as in her 2024 contribution to The Routledge Handbook of AI and Literature (Bassett, 2024), referenced a revived discourse in structuralism, claiming that language is a self-contained, mechanistic, system of signs. Rather than pursuing this, Bassett turned to narrative theory and phenomenology, to the way language gathers and assembles an experience of the self that is both fictive and real — as when, for example, one might present a fictional façade, yet it is judged and accepted as real by others.
David Watson followed with a scrutiny of unsupervised learning: the method by which models are presented with unlabeled data and left to infer structure from within. A crash course in the technicalities of machine learning was followed by several problems. When a clustering algorithm sorts data into groups, is it discovering natural kinds — existing categories in the world, such as ‘cat’ or ‘cow’ — or merely inventing convenient groupings for us, as with ‘food’ or ‘pet’? His worry is that we too readily take the second (inventing) for the first (discovering). Adam Linson closed by returning to art, reading the shift from generative to agentic AI through Gell, Danto and Dewey, suggesting that agentic, autonomously produced, AI art might be genuine, yet inescapably shallow and anticlimactic.
Gotcha! Or have I?
Something interesting was, as Linson named it, the ‘gotcha’ moment. A hypothetical room is shown a poem, or a canvas, is coaxed into admiring it, and is then told it wasn’t made by a human and the admiration curdles into shame. We have been had. I have experienced this many times across different AI events, but the shame has not been ubiquitous. Recalling my attendance at the Tech Show London 2026, Professor Hannah Fry posed a similar exercise to the audience, asking them to repeatedly identify which words were written by an AI and which by a human (Fry, 2026) but the audience could not reliably tell. Rather than threatening our unique humanness, however, the example was taken to demonstrate the permeability of these ideas in the first place. Nobody was claiming Large Language Models are human, nor were they ashamed by their inability to distinguish AI and human words. Perhaps this reflects the contexts of different events — techno-optimists at a corporate tech show versus academic humanities scholars — but perhaps, too, it indicates that such ‘dupes’ are worthy of serious consideration.
Returning to creativity and the Interdisciplinary AI workshop, Mike Wheeler summed up this provocation sharply: creativity is regularly judged upon the finished artwork, not the artists’ generative practices, so why does an artwork’s source matter at all? Why is it duplicitous or shameful to distribute or find value in AI art, and is this not reviving old debates on intentionality? If the burden of creativity has migrated onto the evaluator, does that not allow the machine back into the room?
Linson’s reply was to keep “the head” out of it. He was not explicitly referring to intentionality which he viewed as unimportant in considering a work of art. What matters instead is the tradition within which a work becomes legible. We read the drips of paint we know to be Jackson Pollock’s against abstract expressionism, with the same marks made by a chimpanzee or a toddler not earning that reading. His sharper case was an artwork involving seeds being sown along a railway track: a viewer who knows the work and its context can read migration, invasion and belonging into the weeds, whereas someone stumbling upon an ordinary overgrown verge cannot. Drawing examples from Picasso’s cubism, Linson then described the strangeness of an object made to cross from one “world” (context) into another — in this case, from a context of non-art to art.
I would describe the same effect as defamiliarisation (Shklovsky, [1917] 1965). By de-automatising, defamiliarising, or making strange, art confounds our expectations. As Viktor Shklovsky wrote:
“Art exists in order to restore the sensation of life, in order to make us feel things, in order to make a stone stony … to create the sensation of seeing, and not merely recognising, things; the device of art is the “estrangement” of things.” (Shklovsky, [1917] 1965)
Overgrown railway tracks are made strange when encountered at an art exhibition when they might otherwise be unremarkable. Whilst Linson locates the source of that strangeness in the context of where the work was made, the world from which it originated, I would locate it in the context of its reception, the world it crosses into. This distinction is subtle but draws on different traditions — the latter being Jaussian reception theory (Jauss, 1982). Alone, the mere crossing of worlds (or changing contexts) is insufficient. A pencil moulded from old car tyres is admirable in the fact it has been recycled, without it needing to be aesthetically beautiful. Without the institutional stamp of approval — as one might consider widespread value judgements of AI works — there is no art. Defamiliarisation does its aesthetic work at the moment of encounter, not the moment of manufacture, but it requires, in some form, a subversion of expectations imparted by institutions.
A list by any other name
Caroline Bassett continued the discussion. If we have set intention aside, how then does history enter the work of art or literature? She proposed an answer: history enters at reception, except that, reception is retrospective (Jauss, 1982). A literary parallel of Linson’s argument can be found in Bassett’s Routledge chapter, where she discusses an account of a chatbot asked to extend Maya Angelou’s ‘Still I Rise’. The chatbot’s response merely assembles “events, people [and] slogans”, thus rendering senseless “what was a singular response to racism” (Bassett, 2025). In her chapter, the AI-generated ‘Still I Rise’ is described as: “What looks like literature turns out to be a (laundry) list” (Bassett, 2025).
Yet, I contend, when furnished with the institutional stamp of approval, what is to stop a laundry list, or a shopping list, being recognised as a work of art?
Behold, Michaelangelo’s 1518 shopping list:
and honestly? That’s not just a list, it’s art
Strings, Runes, and Rocks
This is where the discussion meets my own work on AI creativity. If, disregarding intent as we are wont to do, historicity enters retrospectively, at reception, then reception is not a late addition to art but the thing that constitutes it. Creativity, then, if it is to be found anywhere in the AI case, is better sought in the reception of an output than in its generation.
To frame it differently, what an AI might produce is an artefact, but not yet art. It becomes art when it exerts a force on someone, and when that force is ratified — published, hung, platformed, argued over. This is Arthur Danto’s argument reformulated: what makes Andy Warhol’s Brillo boxes art, but not the supermarket’s identical cartons, is nothing the eye can find but their standing in an ‘artworld’ that receives them as art (Danto, 1964). Philosophically, I was met recently with a similarly aligned position at the Philosophy of AI Summer School in Venice. There, the philosophers of language, Rachel Sterken (HKU) and Alex Radulescu (Mizzou) presented an argument for LLM outputs, not as even language, but “candidate strings” which become language at the point they are taken to be language.
The Runamo affair is my preferred illustration (Herman, 2008). For centuries, a fissured band of rock in Sweden was read as an ancient runic inscription, and was solemnly translated, until geology established that the ‘runes’ were natural cracks in the dolerite. The episode is usually filed under error corrected. But the inscription was, for a time, fully operative as text. It was discovered, attempts were made to decipher it, and it was institutionally believed. Its status was settled, in both directions, by public judgement rather than by anything in the stone. Meaning was administered at reception. The maker, here, was geological. Nobody objected, until they did.
Carry this back to the ‘gotcha’ and the shame falls apart. The trick works by leading you to value something and then disclosing the lack of an intentional mind behind it, as though this absence cancels the value in retrospect. That move smuggles intention back in as the measure of worth. If value is conferred at reception, then being led to admire an AI image is not an error later exposed, it is reception behaving exactly as reception behaves. The gotcha is a gotcha only because it has been staged as one. Should the revelation of the Runamo rock be shameful? Should we point and laugh at the tradition of scholars taking it seriously? I would say no. This specific ‘gotcha’ moment does not undermine the experiences of those for whom the rock was, indeed, a runic artifact. Those experiences, or patterns (Bernini, 2026), are real.
About the author
James de Lullington is a postgraduate research student in the Department of English Studies at Durham University, examining narrative creativity and artificial intelligence.
Find me online: LinkedIn
References
Bassett, Caroline. 2025. “The Author, Poor Bastard: Writing, Creativity, AI.” The Routledge Handbook of AI and Literature, edited by Will Slocombe and Genevieve Liveley, 19-26. New York: Routledge.
Bernini, Marco. 2026. “AI as Character: Real Patterns in Fictive Interactions.” Style 59 (3): 466-507
Danto, Arthur C. 1964. “The Artworld.” Journal of Philosophy 61 (19): 571-84.
Dewey, John. 1934. Art as Experience. New York: Minton, Balch.
Fry, Hannah. 2026. “The Future with Hannah Fry: Stories from the Edge of Technology and Science.” The Tech Show London 2026
Herman, David. 2008. “Narrative Theory and the Intentional Stance” Partial Answers: Journal of Literature and the History of Ideas. 6 (2) 233-260
Jauss, Hans Robert. 1982. Toward an Aesthetic of Reception. Translated by Timothy Bahti. Minneapolis: University of Minnesota Press.
Shklovsky, Viktor. (1917) 1965. “Art as Technique.” Russian Formalist Criticism: Four Essays, translated by Lee T. Lemon and Marion J. Reis, 3–24. Lincoln: University of Nebraska Press.
Watson, David S. (2023) “On the Philosophy of Unsupervised Learning” Philosophy & Technology 36 (2)
