The Digital Futures of History: Can AI be superstitious?

This is an approximate text of the talk I gave at King’s Digital Futures Institute’s event on the Digital Futures of History on 15th October 2025.

In 1937, H G Wells imagined an interconnected worldwide information system of the future, which he termed the “Permanent World Encyclopaedia”. This carefully curated and presented summary of the world’s knowledge would supplement, and eventually replace, both what he saw as the outmoded eighteenth and nineteenth century concept of the encyclopaedia, and the knowledge traded in by traditional universities. These, he said:

“[did] not perform the task nor exercise the authority that might reasonably be attributed to the thought and knowledge organisation of the world”.  Never again, in Wells’s view, would there be misinformation, false narratives or unevidenced claims because an army of professional librarians, curators and archivists would have custody of the sum total of human knowledge, in a “a well-ordered scheme of reference and reproduction”. (World Brain, 1937)

Fast forward to 2001, and the closest manifestation of the Permanent World Encyclopaedia is, of course, Wikipedia. We may tip a wry smile in H G Well’s direction over his faith in how the permanent world encyclopaedia might be overseen and run, but he was not as far off as might be seen on the face of it. Rather than a professional staff of curators and librarians, Wikipedia, and its offshoots such as Wikidata, has evolved a set of processes and models through which volunteer effort is channelled systematically and, to an extent, reliably. In 2002, after just one year’s operation, Wikipedia had over 48,000 articles. Today’s estimate puts that number somewhere north of 65m articles in 340 languages. A world encyclopaedia indeed.

2002 was also the year that what has to be my favourite ever Onion headline appeared, professing shock that a factual error had been found on the Internet. Looking back at nearly a quarter of a century’s distance, the idea that this was a subject of satire even then is telling.  From the early days, the need for the kind of stable, reliable information the Permanent World Encyclopaedia promised was needed by anyone wanting to take advantage of the world’s knowledge opened up by digitisation, including historians and folklorists. But it wasn’t quite there yet. A response to this, as noted in the main description of this event, was a rush of projects in the 1990s and early 2000s to digitise the sources and resources which the humanities rely on. In those heady days before the 2008 financial crash, this meant significant financial investment, with funders like the AHRC paying large sums, into the many millions according to this list, for the AHRC’s Resource Enhancement Scheme (which I am old to remember) alone.  

Inevitably however, centrally planned and funded efforts to digitise existing resources  replicated existing structures and existing content. Even at the time however, it became clear that the effort and the ambition to digitise masses of historical content would eventually push both its form and the way it was used in new directions. In 2000, the renowned digital humanities scholar Susan Hockey noted:

The humanities scholar of the future will have access to digital objects of many different kinds on the network. It is not yet clear exactly what these objects will look like, but the need for a common interface and set of software tools is obvious. It is also clear that the future environment will be mixed … much remains to be done to work towards a common methodology which allows for the many different and complex types of material that are the subject of research in the humanities and can also leave open the possibility of new and as yet undiscovered avenues of exploration

Posterity has shown than one such kind of humanities information object, which could not have been fully anticipated at the time, was digital vernacular expression on, for example, social media, and how this took hold as the WWW evolved from a publishing environment to a conversational one – the advent of was called Web 2.0 – in the mid noughties. Facebook was founded in 2004, Google Maps in 2005, Twitter, as was, in 2006. The very word “crowdsourcing” was coined by the tech journalist Jeff Howe in the same year.  However, as Trevor J Blank points out in the introduction to his landmark edited volume of 2009 Folklore and the Internet, while the fields of anthropology, sociology and communication studies began to engage with the ever-expanding and evolving WWW in this period (and this interdisciplinary collaborative is very much manifested in the present-day Department of Digital Humanities at King’s), folklore studies largely stayed behind. It has certainly done some catching up in recent years -m and Blank’s volume must be acknowledged as a key moment on this – but it is worth thinking a bit about why this might have been.

Conventional definitions of key ideas such as superstition stresses the informality and fuzzy nature of both the information they rely on and, more importantly, the evidence we have for them – not generally acknowledged or officially recognised. Quoting Peter Opie, Iona Opie and Moria Tatem state in the 1989 Dictionary of Superstitions, that ““[f]olklore in general ‘consists of all the knowledge traditionally passed on from one person to another which is not knowledge generally accepted or “officially” recognized’ (p. viii). Blank, however, stresses both the creative and the shared significance of folklore beyond the archive. He states that “[f]olklore should be considered to be the outward expression of creativity – in myriad forms and interactions – by individuals and their communities” (p. 6). In the context of this definition, Blank goes on to argue that printing and the distribution of printed material promoted and enabled the study of folklore in the nineteenth and twentieth centuries – despite William Wells Newall’s claim that genuine folklore is the “lore which escapes print” – and that the Internet was proving to be the “new print”  which was bringing folklore into a new “electronic vernacular”.  Blank and others since 2009 acknowledge the continuous presence of physical metaphors in the online space (which is, itself, a metaphor) – web page, surfing, portable document etc – and states that the Internet has an inherent base in the real world – that “there is a human behind everything that takes place online”.

That, I think, is the key statement that we need to consider when thinking about the digital futures of folklore. The digital culture theorist Guro Flinterud recently argued that the algorithms which sort content on platforms such as X (Twitter as was) shape the behaviour of those contributing content to them by exerting a priori influence on which pieces of information get widely circulated and which don’t.  Flinterund views the algorithm as

a form of traditional folklore connector, recording and archiving the stories we present, but only choosing a select few to present to the larger public [p. 451].

To the digitized archive therefore, we must add the algorithm, a decidedly non-human actor in the construction of knowledge. It is the very nature of algorithms that they build connections according to the logic with which they are programmed, not to any vernacular or human-driven scheme.. Add to this the interacting powers of spin, misinformation and commercial appropriation, where the “ground truth” of the (digitized) archive is overtly relegated.

Does truth even matter as a result? In a 2023 interview where he faced criticism of the historical accuracy of his movie Napoleon, Ridley Scott is quoted as saying “When I have issues with historians, I ask: excuse me, mate, were you there? No? Well, shut the f*** up then.”  And in 2022, Graham Hancock’s Netflix series, which claimed that there are previously unknown pre-Ice Age connections between the emergence of complex societies in different parts of the world. Whether this can be evidenced or not hardly matters,  the reach of the show – some 25 million viewing hours in one week, according to one report – drove great impact and algorithmic success as a “folkloric connector”, in Flinterud’s terms.

The phenomenological mediation of the screen can make it very hard to tell these different kinds of connector apart. In a recent post, the historian of belief Francis Young makes an intriguing argument which expands on this. Young argues that the reception of AI, across the public and the scholarly spheres, has shifted society towards a new epistemology of knowledge, where content produced by LLMs is perceived as literal ground truth. For example,  that a LLM can internalise the extant works of Livy, and then produce the 107 books of his that are lost – that it can literarily produce the lost works of Livy.

This is the context of my question – can AI be literally superstitious? This is meant to be more of a provocation than a literal question. What I want to provoke us to consider is a comparison between AI generated content -the logic of the algorithm, driven by theoretically limitless computational power over billions of training datasets across the Web – with the complete absence of logic inherent in superstition. Is there a meaningful difference between the two? This is what I think the Digital Folklorist of the future will have to figure out.