Digital Narrations: Fails and Errors
‘Digital Narrations’ is a series of different formats that has been exploring the influence of digital media on narration in the arts since 2016. In 2018, we invited Nishant Shah to participate in the event ‘Agents of Change.’ We stayed in touch, and the idea for this new edition of ‘Digital Narrations’ as part of APRIA Journal arose from this exchange.
Our curatorial approach is based on long-lasting co-operation, friendships, and the will to show different perspectives. This is why this editorial takes the form of an assemblage of four voices—Nishant Shah’s (in italic) and the individual points-of-view of us, the editors. We have cut up and compiled insights from our collaboration to give an oversight of ‘Digital Narrations: Fails and Errors.’
In this edition of APRIA Journal, five contributions tackle the concept of fails and errors in the digital space from different perspectives using diverse methods. When talking about digitality, the error is often ascribed on a human level—in us as biological beings —or in the old and slow analogue processes and structures.
On the other hand, there is a prevalent knee-jerk attitude of cultural pessimism that views the digital itself as suspicious. Too cold, too rational, not alive enough, and thus, regardless of its power of seduction, as one big fail. For us, this combination is a good starting position for taking a closer look. What other stories can be told on errors, or maybe even through errors? Where do gaps for subversion open up?
Anna Barham’s work ‘ZYX’ employs errors in Automatic Speech Recognition (ASR) software as a kind of productive disturbance. In so doing, Anna creates auditory hallucinations in machines and reflects on the ways these are different from human hallucinations.
Anna’s piece is vertiginous. She doesn’t just give us a thesis on hallucination but brings a landscape that makes us dizzy, showing the irrationality, paranoia, and the bits and pieces of messiness that go into the digital production. What is extraordinary is that the only way to detect these messinesses, these fails, these errors, is to have other kinds of technologies which can detect these errors. The human notion of what is a sound, of what is a text, and what hallucination can mean is severely reworked in order to take us back into the power and wealth of meaning-making and narratives where humans, too, can generate seeming meanings and narratives of a data dump of errors.
In her work, Anna used an audio filter that emphasises the discontinuities of human speech to force errors in speech-to-text programs, and used bits of pieces of those distorted sounds as material for an assemblage. With this, she creates an aesthetical experience where the disturbances are turned into a productive moment in the form of a sound pattern.
I see an interesting tension in fails and errors in regard to digitality. The fail can be investigated as a glitch and generate new forms of aesthetics. Likewise, the error can cast new light on the relationship between human and machine and the responsibilities we are confronted with. Consequently, dealing with fails and errors allows me to find new forms of narrations and uncover broader narratives.
It is the breaking of surfaces. Of disrupting. Of showing that there are these hidden mechanisms which are dismissed as noise, as fragments, as mess, as errors, but that these things have material traces, archaeological depths, and a presence that can’t be just dismissed. The aesthetics of foregrounding this also shows the politics of this process, where that which challenges the smoothness can be removed and the performance of digital smoothness glosses over those things which are removed from the interface surface.
For their collaborative work ‘Earthbnb, from Platform to Planet,’ international artist research collective DoriO used Covid-19 lockdowns create a 3D model of a fictitious collective apartment. They used this to create a listing on Airbnb. In their work, fails and errors are visible on multiple levels. On the one hand, they use the glitch aesthetic of their 3D rendering to show a defective—or enhanced—form of a reality. On the other, their work deals on a conceptual level with structural mistakes that are produced by large platforms such as Airbnb when they, for example, turn habitable infrastructure into unused ghost places.
DoriO’s piece is a reworking of how we even address the idea of fails and errors. In a world where the technological is given as too big to fail, and aesthetically the default, how does art infiltrate and reclaim the space within that structure? Both in the reworking of this space, but also in reorienting its objectives, the work shows the possibility of failure. By stretching it to its logical absurdity, by putting the earth up for sale, it is able to change our relationship and perspective on the issue, helping the transfer of the new-found insight into the one that we have already familiarised ourselves with.
The work by DoriO is a productive break with the aesthetic of digital platform economies such as Airbnb by appropriating their design, re-assembling it and, alienating it from the original context. This is supported by a link to the research work the group has done, which both provides additional context to the video and makes visible the artistic and social praxis of DoriO as a collective.
Particularly during the pandemic, when their homes functioned both as their workplaces and a schoolroom, it became clear for many people how scarce housing space can be. Thus, I see errors described in this work when it comes to gaps, housing scarcity and our handling of infrastructure. Who has access, who does not, who decides on this and which platforms reinforce and harden those structures?
In my research, I have focused a lot on two computational principles and how they shape the notion of the ‘fake.’ The first is GIGO (Garbage In, Garbage Out), which insists that if a computer program gives false results, then it is entirely because the human operators feeding it data or shaping its algorithms goofed up somewhere. The unquestioned idea that errors belong only to the human when it comes to computational processes becomes a critical point in understanding where errors are attributed. There is no such thing as a machine error, only the human failure to calibrate it correctly. And at the same time, with more self-learning devices and algorithms around us, it is clear that machines are creating entirely new conditions of failure—crashes, blackouts, shutdowns, and glitches that lead to profoundly interesting questions of responsibility and accountability when the materiality of a device takes on agency to cause collapse. For me, the question of how fails and errors are attributed, and who (or what) can afford to fail, and who bears the consequences of these failures and errors, remain critical questions that need continued unpacking.
In her essay ‘Democratise the Cyberspace! Storytelling in the Digital Era,’ Maria-Cecilia Quadri wonders how our society can develop a broader understanding of the errors it commits in its comprehension of digitality and which role cultural workers can play in fostering that understanding. In this respect, her work can be read as a framing of the contributions of this issue that propose different levels of exactly that.
Maria-Cecilia interweaves autobiographical experiences with an overview of the discourse on society’s handling of both general (supposed) errors and the blind spots specifically related to digitality. She draws on her own intersecting cultural backgrounds to illuminate structural fails and errors that perpetuate existing and prevailing balances of power, ethical values and systems of truth.
It is worth reflecting deeply on these errors and failures because they are so powerful, particularly against people who often are already disadvantaged, for example, migrants. Not only are they deeply nested in our psyche and provide us with feelings of appreciation or depreciation but they also have tangible effects on our lives. People who commit mistakes lose the right to participation—maybe they are no longer allowed to drive a car, won’t be invited to job interviews, or could lose access to education. But what is declared as an error or a failure is not always objective; it is instead located and contextualised in a negotiable field of moral concepts and opinions. As a result, that decision—which counts as an error—is now increasingly made by algorithms.
Fails and errors introduces an interesting teleology to the idea of digital narrations. Who is the recipient and the intended audience of digital narrations? As the human messaging and the machine information process this information differently, there is going to be a clash of values, ideologies and systems about the meaning of this information. The framework of fails and errors become critical to figuring out which one of these competing claims of truth and fake will be attributed to what systems of information processing. It helps us figure out where failure is attributed, where errors are placed, and where the interventions in accountability and responsibility, and also culpability and penalisation are made as we continue to build entwined digital narratives that shape our futures.
Nathan Jones’ essay ‘Experiential Literature? Comparing the work of AI and Human Authors’ examines the specific qualities and shortcomings of GPT-3, an AI text-generation software. He analyses the failure of this speech model by comparing it to contemporary literature, showing how the software is limited by the corpus of texts it is based on and the fact that those texts must always come from the past. The texts that GPT-3 generates show gaps and logical errors in their narration because the AI does not have a conceptual understanding of language and content; rather, it only simulates its outcomes on the basis of statistical and probabilistic models. His essay deals with the question of how this many-layered failure impacts language, literature, and eventually a possible future.
Nathan puts forward a powerful comparative framework to look at the production of literary text by comparing GPT-3 and experimental authors, both of which are invested in the polar ends of a text spectrum. GPT-3’s reliance on large-scale language corpus and its insistence in finding realism harkens to a prosaic twentieth-century modernity that mimics the exploitative nature of these databases in the production of writing itself—trying to develop efficient and coherent utterances without meaning. Using Ravn and Fosse [contemporary authors] as the counterpoints, the essay shows how these neural learning networks modelled on large scale data sets cannot be the measures of our language and meaning.
Using experiments that seed GPT-3 with works by contemporary avant-garde authors, Nathan illuminates the limits of the AI texts and proposes using them as a tool to learn more about what makes these authors’ writing special.
The idea that GPT-3 is not just a failure in complexity processing but is, in fact, a failure in comprehending language and its exigencies—and that the errors it produces are not at the level of merely syntax and processing but at the very heart of what makes meaning—is beautifully and poetically demonstrated.
There already is a seed for a story in the arc from failing as the starting point that leads to an error as the perceptible result. And mistakes in general have the potential to lead to fascinating stories that bring new insights—or just even the tragedy of seeing someone fail. In this regard, the topic is almost predestined for us as ‘Digital Narrations.’ Our focus on digital narratives and modes of storytelling entails a certain focus on phenomena of language.
But we want to accommodate other works as well, especially those that show how errors can be used to create, in one way or another, the potential for opening up a space for something new. We are interested in the whole range of errors that stems from digitality. That is to say, the failure of digital systems that continue the capitalist logic of reduction and exploitation, as well as the errors that can plague the tiny ‘building blocks’ of digital logic.
Based on his own experience of working as a programmer, Philip Ullrich looks at the phenomenon of the computer bug in his poetic text and subjective reflection, ‘The Bug (Up Close and Personal).’ His work proposes various categories of bugs that demonstrate different perspectives of these common errors and shows how they are produced.
Philip’s essay clearly shows us that computational systems can never fail because they are never complete, and they can never have errors or ‘bugs’ because bugs are designed, present, and even essential to the workings of the system. Perhaps playing on the internet meme, this is not a bug, but a design, one being intentional and the other an error, Philip gives us a multi-vocal theorisation that the bug is the design and that we need to think about both fails and errors as performing certain functions of valuation, of paranoia, and of control, thus shifting attention from pass/fail to why something—like a bug—is construed that way.
Philip explores the relationship between human and machine by foregrounding language and communication. In the process, he uses his programming experience to give unexpected takes on how to interpret the bug, which are humorous at times.
The essay is a great example of the artistic researcher as a synthesis point, bringing together multiple strands of deeply understood and practised concepts and practices to help create a new narrative which is more than the sum total of the parts. The reconstruction and appropriation of the bug as a divining tool, rather than something that has to be resolved, brings into relief the capacity of artistic research to reframe the questions, enchanting them and engendering them with new possibilities of meaning.
The series Digital Narrations deals with the influence of digital media on the form and distribution of stories and narratives within the internet and beyond. For me, the series is an open, experimental container for trying different formats, aesthetics and potentials of narration in the digital age. Since 2016, we have invited artists and guests for exhibitions, events, workshops and screenings to approach this topic from different angles and to enable the public to experience it in different forms. Which stories are told? How are they told? And who can and may tell their story?
For me, Digital Narrations is a provocation that opens up how we think about the very materiality of digital ontologies and the structure of narrative epistemologies. Digital technologies are fiercely non-narrative. Digital information processing is non-linear, does not have narrative causality, and is continually atomised into information sets which resist and overthrow the narrative impulse. The promise of digitalisation, in many conversations, has been to replace the interpretative, suggestive, symbolic, and ambiguous fickleness of narrative with the logical, decisive, definitive, and explanatory structures of digital data, in the quest of mythical objectivity and fixity.
Digital Narrations, then, defies these expectations of digitalisation and instead explores the potentials of narrative techniques, structures, and possibilities that emerge as the narrative framework queers the pitch of digital architectures. Digital Narrations emerge as cyborg forms, where the machine logic and the narrative possibility converge to mutually critique and co-constitute each other, in order to look at the creative, playful, disruptive, and imaginative ways in which born digital narratives are created. In the process, they critically question the ways in which we think of digital realities, or fakeness, and the new structures of making meaning within digital computational networks.
Our approach is to establish a fruitful mix of commonality and differences, of exact looking and blank spaces. The topics we deal with are too extensive to be described in complete detail by a limited number of works. But instead of restricting ourselves, we want to present a selection of different artistic positions that show the range of the extensive and complex phenomena we are dealing with. At the same time, the series fragmentary nature makes clear that these are individual standpoints that do not serve just to illustrate a thesis of ours. We prefer a multitude of voices that, through their limited number, keep their individuality over making absolute claims of truth.
My engagement with the group Digital Narrations has been more as a book-end. My first conversations with the curatorial team began with long public conversations and performance, around questions of fakeness, misinformation, machine narratives, and human possibilities. My initial prompt into thinking about digital narrations was to examine how human-machine intersections lead to new modes of telling, identifying, defining, and claiming conditions of truth-telling. The idea that the Turing Test was never meant to verify machines but to destabilise the human assurance of our capacity for recognising fakes resulted in different conversations around machine errors and human failures and how they coincide in digital narrations. I now come back to these conversations to revisit the idea of where fails and errors lie, when we think of digital phenomena and epistemologies. In many ways, the conversation has moved from figuring out how humans can measure up to machines in information processing to how we need to question the machine measures which are forced upon human scale narratives in identifying the fakes.
Often there is trust put into digital technologies because they operate with numbers and thus they suggest objectivity and precision. We think it is promising to offer a counter to this understanding. Not least because in a performance-optimised society such as ours, failure and errors are omnipresent and we are confronted with them anyway. The openness of the terms also allowed for the necessary freedom to establish an artistic process for the involved artists and ourselves. Because all contributions for the publication are originally produced for it and the artists participating practise a culture of exchange, we wanted to involve them in the conception of the issues as a whole, not just their own individual works. This allowed for insightful and revealing links and interconnections to be recognised and established.
What I learn from my engagement with the curatorial team is a commitment to not give in to conventional norms and measures of academic research. The idea that we do not begin with a definitive norm of artistic research, forcing different researchers to fit into outdated methodological moulds, but instead give people space to explore, experiment with and articulate their own de-disciplined practice of artistic research bears fruition. It makes the work of curatorial editorship that much more difficult because new controls for quality check, for research integrity, and for verifying positions have to be evolved, but the fact that this has happened, and that each contribution has pushed the notion of what is artistic research, how do we document it, and how artistic research intervenes in complex questions, informs the curatorial approach.
For me, there are countless formats that facilitate artistic research. Precisely because of the possibilities of the arts to find, refine and establish new methods, they can offer fresh insights and identify different perspectives. The formats presented in this issue originate from original and independent questions regarding the topic of fails and errors and the contributors have developed different means of answering them. The resulting works can be located in the field of artistic research because they produce knowledge beyond the methodological standards of conventional knowledge production in universities and scientific research institutes.
It is in the nature of research on fails and errors that comparisons are the default mode of address because something fails or an error is produced only in comparison with a baseline of the normative. All these essays, in their different artistic methods, are showing that the comparative framework, especially when it comes to human-machine interaction, are often skewed towards the machine as the tool and the human as the unit that has to be measured and has to measure up to the norms established by this tool. They all, instead, refuse this naturalised state of comparison and show us that machines themselves—in all their components and assemblages, need to be seen as flawed, mired in a political economy of fails, and cannot be accepted as the clean discreet things that shall govern the world. Instead, the attention of artistic research is in producing new kinds of scales, novel modes of imagining the production of irrational norms, and creating the possibility, through artistic practice and research, to show the making of fails and errors rather than just explaining the states of crises.
Because we are interested in errors in the digital world in relation to narrations, the formats represented in this issue are also dealing with forms of narration. The five contributions can be read as five narratives that show or expose different kinds of errors and allow new perspectives. Exercising different ways of telling stories is a way of practising artistic research for us.
Moreover, with our way of working, we want to present a contrast to the prevalent ways of production in the arts that are focused on clearly attributable authorships: a practice of close collaboration and exchange. We work in a collaborative manner, not with predefined roles and responsibilities. Everyone is equally responsible for everything. It is important for our attitude towards curatorial work to stress its individuality. This is why the communication and exchange with the artists and guests we invite has always been central for us. In each stage, the goal has been to establish a space for the contributors to get to know each other, to discuss the topic and to challenge and play with the format. The word stem of curare in curation has been an important concern of our group and has helped to build a network over the years and to take care of those friendships.
Another element of artistic research that we find recurring is an explanation of the process itself, showing both the evolution of the journey and the experimental set-up through which the artistic knowledges can be produced. The mode of artistic research, however, is not revelatory. It is not about instruction sets or manuals of making, but instead showing the iterative growth of an idea as it is taken through the different structures of making, thus creating a richer, narrative, and interpretative universe. It is a world-building exercise rather than context explanations and the Earthbnb piece is a great example of it.
My interest in fails and errors started with a fascination for fail-clips on social media channels. In these videos, humorous, thematically fitting clips of humans, animals or machines that fail in one way or the other are cut together. The failure is celebrated, sometimes even imitated and staged. The specific humour and the way that failure and clumsiness are celebrated were my starting point for thinking about the fascination of the faulty. The (sometimes deliberate) failure can mobilise and form new narratives. The generation of errors can also be the method of the mischief, the killjoy, the activist and the hacker. This is why in relation to fails and errors, I am interested in the intentional mistake and the generated glitch that can also be understood as forms of resistance.
Just before the world was shutting down because of the Covid-19 pandemic, there was a particular Silicon Valley venture capitalist mode that was taking centre stage—I call it failure porn. It largely involved privileged white men in North America going on world stages to announce the failure of their multi-million-dollar investments for an app that was supposed to change the world but just failed. This failure was announced not with regret but with a bravado, a wink-wink nod-nod, signalling the enormous capital and privilege of somebody to bear no consequences for their failures. In a similar movement around gender and sexual harassment in the work place, #MeToo advocates and survivors were pointing out that the predators who were called out often ‘failed upwards,’ protected by the structures that they embodied and manifested. Fails and errors is a strident framework that looks at who gets to fail and who bears the consequences of these failures, thus laying bare the invisible structural advantages and biases that continue to perpetuate and gloss over the inequities and inequalities of our everyday lives. I hope that the interventions around fails and errors continue to show us where the harms and entitlements are, and how they are coded in the narratives of our times.