Mike Pepi: Bad Data

This text was originally developed as remarks for Coding the Institution: Noya Kohavi and Mike Pepi, part of Criteria, curated by Natasha Matteson as part of the requirements for the master of arts degree at the Center for Curatorial Studies, Bard College.

“Thanks to the gorging abundance of our reproductive devices, we deface the symbol and debase it, treating it contemptuously, negligently, only half-believing that its employment makes any difference. By contrast, we overvalue the technical instrument: the machine has become our main source of magic, and it has given us a false sense of possessing God-like powers. An age that has devalued all its symbols has turned the machine itself into a universal symbol: a God to be worshipped. Under these conditions, neither art nor technics is in a healthy state.”

• Lewis Mumford, Art and Technics

You might recall that the ideology we call techno-utopianism reached a peak just about 10 years ago. At this time I found myself neck-deep in both the world of tech (as a new data scientist) and the world of art (as a critic). I wasn’t very good at either of these jobs. But I knew enough to pose an important question. I figured the database (a digital storage unit for information on cloud-based servers) and the art museum ( to use a brutal oversimplification - a Western invention tasked with storing and displaying art) could be constructively compared.

The main reason: the myth of the Museum was withering - perhaps even more so than usual. Institutional critique and a general exhaustion with the financialized nature of the art world had shattered its illusions. I had no idea what would unfold in the years to come.

Digital utopianism was infecting every facet of culture. The museum was not immune. In fact, it was among the most vulnerable. The ideology of silicon valley - sometimes called the California ideology - found in the museum the perfect foil. What is a museum, other than a terribly functioning database?, they asked. Open source the collections! Free the art - put it online! Join social media and live tweet the experience! Museum websites were the first battlefield, but eventually it bled into the physical museum space. In a short span of time the entire museum field became colonized by the mindset that believed that software was set to “eat the world” - the gospel of innovation and disruption enchanted curators and museums boards alike.

But underneath the surface, the humanistic endeavor would also come under attack - this new digital class viewed every fiber of cultural production as liable to datafied optimization. No one seemed to want to be an institution any more. Everyone wanted to be a platform.

In the time since the rise of platform capitalism I have made three basic challenges to the broad efforts to digitize the museum. In that roughly 15 year period, those lines of inquiry have grown in urgency. With some time to reflect, I’ve summarized them in three parts:

One. While the western museum is not free from sin, namely “imperial cultural extraction, epistemological violence, and stratified access to knowledge/culture” to use one colleague’s shortlist, we must ask whether what the digital utopians aim to replace it will be any better. Positing a museum as a digital database was and is an imaginary exercise in which we might figure out exactly what is worth rescuing from the admittedly pitiful husk of humanism.

Two. Does any part of the metabolism of the database - the functions it affords by way of Artificial Intelligence, Machine Learning, or Software-mediated interactions, actually further democratize the field? Or have we just swapped in new masters? In my most uncharitable reading of the situation, I think that Silicon Valley views the museum as a playground, a testing site for their tools:  an exchange in which tech has everything to gain (cultural capital) and the museum has everything to lose (the power to mediate art historical discourse and steward the relationship its mission and its public.)

Three. Why are we looking for “data” in the museum in the first place? Here I follow tech critic Lewis Mumford who uses the dual categorization of “symbol” and “tool”. The Museum is an institution to house symbols. Tools have been used to record these symbols, of course, but narrative ruled the day. An institution is a “symbol first” organization. Thus, a museum cannot be measured for utility. The impulse to “datafication” inside the museum is a result of the larger ideology of tech solutionism that dominates what we might refer to as our general episteme. In short, it’s one of the most egregious examples of a “solution looking for a problem”. The idea that the museum’s information should be made utilitarian - not for art history’s sake, but for the benefit of platform capitalist corporations - is not necessarily in the best interests of the institution. Put simply - Who, exactly, asked for this?

***

As an update to my thinking I want to again ask: What is a museum visit? How is that not also coded by the metadata, if not literally in rules and algorithms and source code, but in the intellectual work that prefigures the visitor’s art historical gaze?

To understand the risks of datafication, it’s helpful to go through the exercise yourself.

Let’s consider a seemingly simple interaction at the root of this predicament. When I approach an object at a museum, two exchanges are taking place. Both could be said to contain “information.” Perhaps, even “data”.

First, when I look at a work, one part of my experience is positioning the object in a historical lineage, searching for clues, and doing backup checks on the validity of my associations. I look at the date, I think of comparables, I look at the artist, I gauge intent, influence, value, and novelty. I scan my brain for little shreds of historical evidence I can marshall to contextualize this work. In the thematic museum, especially in a permanent collection, the first information one encounters is the object itself, often absent of any metadata. Even as I walk close enough to a work to ingest any single piece of metadata, I am encountering an unstructured and incomplete dataset in my mind - the corpus of rational elements from which I can logically experience this otherwise aesthetic event. This ingestion occurs within milliseconds. Sometimes I pause before considering a new piece of data. For example, sometimes I don’t want to know the artist until I know the date, and check the sense that I have about a specific visual element that without context would not be nearly as interesting to me. Sometimes the entire thrill of a work can be knowing the artist, but not the date, and then experiencing the pleasure of having dated it correctly or incorrectly. And more often than not that incorrect data, this slippage, is the very stuff of aesthetic pleasure, especially when you realize that you and the curator may have just connected upon this very misalignment. “This is an early Arthur Dove” I think, and this very fact alights everything else about the work: why it’s placed where it is, why that one visual element or that subject matter choice or that style takes on new meaning. Sometimes I “sneak up” on art for this very effect. Very often it is the absence of data that produces the most profound effects.

This first class of information is limited to what I already know or that which is presented to me readily. Clearly there is a second element that aims to go beyond this. The second operation is an act of perceiving that I believe to be meaningful - it is almost always a bridge. This bridge is a connection that I, and only I, can add. I ingest the given knowledge, and I make a connection. This connection happens instantly and without any conscious attempt. This could be the mere mental registration of an emotion. Now a new piece of data is added to the minimal dataset you encountered. It is important to note that these new data are not binary, capturable, or expressible in any way. In fact the term “data” is wrong in most senses, but the “update” to the total potential universe of the work of art has been impacted in a now irreversible way. The only way to express this is to go outside the world of pure data, into words, utterances, emotional reactions, or what we might identify as the basic elements of criticism.

The final moment of data connection is also an attempt to add new information. At this moment, I ask something of the work’s presented data that is currently not available to me. Sometimes this is on purpose, sometimes this is because the information is simply not known. This NULL result is also now a part of the total potential universe of the work, but only for me.

In the end, we have several discrete moments of aesthetic information assemblage:

(a) The reaction/perception data, delivered subconsciously by the brain nearly instantaneously as a conditioned reaction of a semi-informed, aesthetically interested gaze.

(b) The given data, either explicitly delivered or contextually derived

• Artist, Date, Material, Title, the visual/physical support of the work, and the object itself.

(c) The connective data, the way in which the viewer ingests and attempts to connect the explicit or implicit information about the work to other works, other information, or other implicit contexts not immediately presented.

(d) The NULL results from desires for new information, the things that the viewer wants to know but cannot, and answers for which the sum of the reactive, connective, nor the given data can suffice.

The sum total of these data are then, also, both new and unique. They are time-stamped and owned by the viewer, now inescapably part of the subject's experience.

As we walk through the building blocks of an institution's epistemological scaffolding, we also should use that same frame to assess the problems with the tools we are seemingly bound to replace it with. Said another way: What’s so bad about metadata? It’s not perfect. It has omissions, it has biases. But who, really, is afraid of the old metadata? What is the root of their fears?

When prescriptive technologies address this fear, are they truly afraid of metadata for such reasons? Or do they fear not metadata itself, but the impotence of the metadata, which is to say, that they do not fear its existence or limitations from humanist, post-colonial grounds, but to the contrary desire for metadata to do more work, to take more power, to cease to dwell in the partial or “meta” but instead to be the complete instantiation of the art object itself?

This is a battleground, just as there was a much longer battle ground in which the tradition of art history determined that museums should maintain as their focus “objects” in the first place, which itself was a response to the history of collecting, and a structural remnant of colonialism and/or the constellation of western ideas underpinning so-called sacred objects or, even the more modern humanist tradition of artistic genius. At some point this organizing logic jumped from the institutionalized wisdom of the museum to the database. Quite literally at some point someone was tasked with creating a database for a museum and the simple notion that the art object should be the base element in a database, the primary row, was taken for granted. The next set of decisions was equally as fraught. Should this database have a field for “nationality”? The nationality metadata field is of course not only itself a western concept, but depends on the location of the institution. Such universal datafication of cultural phenomena then becomes impossible, or, at least arrives with a built in incompatibility to the local or historical context of the means of its specific cultural resonance. What, then, does a digitized museum space look like after the object?

As I discussed early on, no single interaction of mine is uninfluenced by the history of knowledge about art or the way that that knowledge has been diffused. Even the connections that I think I am making for myself, in that personal moment, are shot through with the scholarly patterns worn in by decades of invisible revelations.

Having laid these out, I want to focus on the final category: the NULL results of desires for new information. This is the key part of the desire of the database - to have all information readily queryable, transmittable, checkable, and, in word, to make the experience of viewing art much like how a digital platform would treat it - meaning: always sortable and optimizable, and never for want of the perfect connection, placement, or categorization. Put another way, without narrative or mystery.

This final category of viewer-data interaction is also part of what enchants the human - it’s a key meeting point of the logical and the aesthetic precisely because it operates on an absence of information. It is also a definitional part of what we might call “art” - anything whose experience cannot be fully utilized to the point of crossing over from the world of a symbol to the world of the tool.

In a database, every object is in essence part of the tool, the tool being the sum total of all a database’s objects, and their precise and logical storage for instant retrievability, sorting, and analyses. The non-fungibility of symbols in a database is not a bug, but a feature of the aesthetic object. And it is at this central meeting point where we consider the stakes of the private digital platform’s efforts to bridge this gap through the datafication of the museum.

My fundamental point, contrary to the desires and aspirations of some techno-philes, is that this delicate interplay resists precise reverse engineering. Though it borrows from a past that cannot be completely freed from its institutional genesis, it also cannot be completely platformatized, datafied, or predicted or even represented by prescriptive information technology. It cannot, in other words, be made to be artificial.

Every new techno-fix that comes into the marketplace of ideas draws some of its power by promising to abolish the history that came before it. As Fred Turner told Nora Khan in a recent interview, “Silicon Valley wants to pretend that that history and politics don’t exist”. I want to expand on this. To put something into a database is to momentarily stop its history - to erase and declare a new world - because when designing and implementing a database it is fundamentally impossible to not simply gloss over the messiness that history is made of. One must reduce, replace, and erase!

We might go further: what takes over when the messiness of art historical data gets pushed out and smoothness of the platform gets wedged in? Is it the art market? Public opinion? Where else have we seen that impact the museological field? We know from the new field of AI ethics and critiques of surveillance capitalism that the immediate results are neither equitable, democratic, nor fair. There’s not enough time and space here to revisit the history of the museum - but perhaps we can say, to understand the epistemological mappings of why art is collected, how it is displayed, and how it is historicised, follow the machines.