Or on a cloudless day at noon, the sky looks azure blue
This is the first entry in my Lexicon – an ongoing glossary of words and phrases I am thinking with and about. Future entries can be found here.
How lucky we felt. We single parents, carers, non-city dwellers, agoraphobes, and other Others, who were suddenly treated to a panoply of talks, interviews and presentations, which, before Covid 19 had forced institutions online, were largely out of reach for far too many of us.
Given my focus of study, and paradoxically since we were grateful for the existence of digital technologies making it possible to remain in touch, many of the talks I attended hosted artists discussing projects that explored a multitude of horrors connected to various forms of digital technology. An easy target since, in amongst the benefits such as making it possible for marginialised groups to attend artists’ talks, there is a wide array of horrors on which to focus. From data annotators to AI-generated porn, surveillance capitalism to the commodification and instrumentalisation of our most intimate relations. During these talks, the presence of bias in machine-learning processes was a frequent topic, often headlined with artists’ statements that warned potential viewers of its existence, and which explained, sometimes at length, how the work aimed to critique and even undo the harms.
One of the most memorable moments during what became, in 2020, a barrage of online events that most of us, even those who benefited, eventually grew wary of – even traumatised by – was when an author stopped proceedings mid-talk. They said they were frustrated by the banal conversation taking place, in which a young artist had identified bias in what we routinely refer to as AI – dynamic machine-learning systems in which feedback mechanisms make bias structurally unavoidable – as if naming it were all that is wanted. (Belligerent, perhaps, but I find the longer description more useful and honest than a term that evokes science fiction rather than digital engineering.) Of course, the tech contains bias – it is structurally impossible for digital expression not to contain bias since bias is fundamentally part of all human expression too. The frustrated author left the Zoom call abruptly, stating he simply did not have the time for such a superficial conversation. I felt very sorry for the young artist speaking and for the host, who was evidently mortified. However, I had a great deal of sympathy for the author too. The artist had latched onto one of the most obvious tropes in today’s digital landscape. They had recognised and named an obvious symptom. But so what?
Anna Kornbluh (2023; 12) identifies in Immediacy: the style of too late capitalism, how ‘directness and literalism are the techniques’ that dominate a 21st-century culture, and a world in which ‘instantaneity are economic premiums as much as they are artistic ones’. Try as one might, we cannot escape immediacy, as according to Kornbluh (2023: 8), it is the ‘master category for making sense of twenty-first-century cultural production. Immediacy rules art as well as economics, politics as much as intimacy. It’s at the art auction, in the boardroom, in the lingo, on the brain’. It was in the talk that lockdown day.
While I was studying photography, I picked up on this trend and eventually attempted to resist the generalised encouragement to make projects that stated facts, such as ‘being female in a man’s world is shit’, often accompanied by tautological images of women or young girls ‘wistfully staring into the distance’ (Rubinstein, 2018) – a trope I had employed right at the start of my photography studies when I made a project called Girhood. Kornbluh laments this sort of one-to-one correspondence that photography, in particular, seems to favour. (I will write about why photography is so prone in another lexicon entry). Being female in a world organised for masculine existence can indeed be quite shit, but do we really need to spell that out quite so tautologically? Kornbluh’s thesis implies that one-to-one correspondence dominates because if anyone is forced to think for too long, they are likely never to appear on the shiny surface screen of our devices, never mind have a chance to become subsumed by it. And that being immersed into the liquidised state informing our shiny world-picture is all that matters. All that makes us real. Kornbluh (2023: 8) laments that the only presence that matters today is that which is immediate, intellectually flimsy, but above all ‘surface’, even if only for a moment and far short of the famous fifteen minutes of fame modernity once promised all: ‘Spatially, immediacy encloses while delivering everything close: the world at your fingertips’. And on our devices. We are isolated while everything is in reach. This situation was in play long before Midjourney or ChatGPT were available to all.
I do wonder now if I am being unfair to the young artist delivering their talk in that Covid Zoom call who had to suffer through the awkwardness of a respected author loudly abandoning it midway, and whose work may well have been an expression of multi-layered complexity. Since the author’s rupture is really all I can recall, it is impossible to know for sure. The act of abandonment might be as much an expression of Kornbluh’s immediacy as the banality that tautological statements about bias in the machine can be. Perhaps the artist went further than mere recognition and naming. Perhaps they did not. Without wishing to put words in the frustrated author’s mouth or intentions, where perhaps there was nothing more than an understandable bad mood, it seemed very much as if the surface of that artwork was all that existed. There is a bias in the machine: recognised and named. But then what? And so what?
As mentioned earlier, there are many, many, many articles and papers discussing bias in proprietary image generators, such as Midjourney, Kling, Sora or Google Veo, and in large language models (LLMs) like ChatGPT, Claude, Deepseek or Euria. Do a search, and you will be served up too many options to count. Many, I am sure, avoid being entirely tautologous. As someone whose creative research is a kind of reflexive anthropological field study of 21st-century cybernetic culture, I try to read or at least glance at most related articles that pop into one of my various scrolling feeds. Being an inhabitant of Immediacyland, overwhelmed with information and prone to scroll addiction and fragmented attention, I am afraid I often leave before the ending, a little like the author did, whenever the topic refers to ‘bias’. Are these researchers really so surprised by the fact, ‘There is bias in the machine?’, I wonder. I admit here, my frustration opens me up to accusations of missing out on conclusions and perhaps never learning something deeper or more complex myself. And yes, I see the irony. Here I am responding with bias of my own, frequently leaning into my preconceptions. My overwhelming response, almost before I have even begun reading, is, ‘And??’. By that I do not mean, ‘And, who cares?’, for we could mostly do with giving a lot more dams than we do, but rather ‘And, what more have you got to say or show us about this bias?’ For the most part, it seems to me that the claim is akin to someone saying ‘On a cloudless day at noon, the sky looks azure blue’. Yes, yes…. so it does.
The problem I see with the conversation around bias is not that it exists, but rather the constant presence of alarming surprise. Perhaps asking ourselves why we are still so naïve may be a more fruitful endeavour than simply stating facts. What is it about our culture that seems not only to favour directness and literalism, oversimplification and base tautology, but that has, according to Kornbluh (2023: 13), even embraced such an approach in academia? Is she correct (2023: 15) and right to abhor a world in which, as she suggests, most of us demand to be spoonfed the real, and anything more complex gets filed under ‘Had enough of experts’? (Oh, for some bloody experts in the art of diplomatic leadership right now, heh?!) And if so, why exactly is this malaise with complexity such a problem, rather than merely being a necessary reconfiguration of authority as new systems allow us to challenge old ones, and what can and should we be doing about it?
While Kornbluh may or may not have a point, depending on your sensibilities, one aspect of this trope – the surprise and subsequent disappointment that digitally coded media contains bias – comes, perhaps, from our habit of splitting the body from thought, language from reality, and positioning our human selves beyond the rest of creation. This habit is so ingrained, so deep, so unconscious, it finds itself expressed even in narratives that might be labelled post-human or more-than-human, known for critiquing harms associated with representationalism. In other words, even where authors are cognisant of this habitual way of splitting our thoughts and conceptions, the tendency to sever cannot help but creep back in through the words we almost cannot help but use. Owen Barfield, a friend and fellow academic of CS Lewis and JRR Tolkien, referred to words as fossils of consciousness. Our words, the ways in which we use them, inevitably contain what biases exist in the epoch in which they are being used. Our epoch is deeply and irrevocably split. Everywhere, we continue to speak of mental health as if the mind and its ills are somehow severed from the body and the landscape in which it flows. On or off. 1 or 0. Black or white. Even when we conceive of two sides of the same coin and then pit them against each other (reason and rational thought vs embodied practice, for instance), we are capitulating to a certain Western view that has dominated for centuries. If we stopped seeing expressive tools, from the words we speak to the code that undergirds the modern paradigm, as something Other than us, but rather as elements in the same substrate of languagematerials* from which our bodies emerge, perhaps we might be less shocked that the feedback machine-learning processes we invent harbour and exacerbate the bad stuff. The problem with not recognising how we split is that it allows us to appoint blame without taking any responsibility. I understand that these machines proliferate and deepen bias more rapidly than ever before, and that we must recognise its presence, but perhaps the deeper issue that desperately needs addressing is our tendency to avoid responsibility. We do this constantly. From the invention of printing to the camera to digital images to AI. We consistently make these technologies function as the fall guy for our complex drives and frailties. Perhaps, if we were more honest about what and who we are, we could sit more realistically with our creations.
If we were able to acknowledge a more holistic picture of reality, we might see that bias is always everywhere, all of the time. It is in the so-called objective texts invented to exclude subjectivity. It is in the way we organise our world. And it is in the everyday actions that motivate and inform our choices. For instance, it was deeply embedded in the habitual acceptance that events and talks should be in person only, despite the fact that online and hybrid meetings were possible, if not yet ubiquitous, long before Covid 19 came along. Sure, no one had been forced into investing massive sums to make work-from-home possible and online meeting spaces function smoothly prior to 2019. But then the wider habits of modern existence are not really aimed at single parents, carers, non-city dwellers, agoraphobes, and other Others, as much as they are about helping a narrow concept of Homo economicus do his job. We might remember that next time we feel inclined to roll our eyes and exclaim, ‘So biased!’ about our machines.
*Languagematerials is a neologism I have been using for a few years, encouraged perhaps by Karen Barad’s embrace of such things, for example ‘ethico-onto-epistem-ology’, but which I have not yet entered into the lexicon. Soon, soon.
Refs:
Kornbluh, A. (2023). Immediacy, or The style of too late capitalism. London New York: Verso.
Barfield, O. (2025) Owen Barfield and the Mystery of the Word, The Owen Barfield Literary Estate, 19 December. Available at: https://owenbarfield.org/owen-barfield-and-the-mystery-of-the-word/ (Accessed: 23/01/2026)
Rubinstein, D. (2018) ‘What is 21st Century Photography’. Available at: https://thephotographersgallery.org.uk/photography-culture/what-21st-century-photography (Accessed: 4 April 2021).