TABLE OF CONTENTS

PANDORA’S BLACK BOX: ALGORITHMS OF TASTE

Dan Graham, Likes (A Computer-Astrological Dating-Placement Service), 1967–69, printed material, dimensions variable.

IN 1969, Dan Graham completed Likes, a “Computer-Astrological Dating-Placement Service,” which anticipated online dating services by using a computer algorithm to match potential partners on the basis of astrological signs, physical appearance, and relationship priorities. The program even incorporated user feedback, via newspaper advertisement, on the success or failure of matches already made.¹ In 1972, Joshua Young, a painter living in Los Angeles, circulated a series of questionnaires to area artists, asking with which of their colleagues they would like to exhibit.² An algorithm used their responses to generate a series of group shows—collectively called the Market Street Program—thereby physicalizing these virtual social networks into what a contemporary reviewer was already calling “clusters.”

Set to launch in January 2012—but at the time of writing (early February) still in development and not accessible to the public—the website Art.sy may, as a search engine and filter for works of art, signal an end point uniting these two trajectories. Art.sy does not grant beta access to the site to members of the press, but based on the company’s own PR, other articles and interviews with Art.sy staff, and a brief viewing of the current state of the site at the company’s open office in New York, a general understanding of the core technologies emerges. After an Art.sy user searches for an artist, work, or artistic attribute that he or she “likes,” to use Graham’s prophetic term, a proprietary algorithm generates clusters of suggested images, ranging from canonical museum pieces to contemporary works on sale at affiliated galleries.

Making these images network in the way that the artists networked in the Market Street Program, Art.sy merges murky formal affinities and social linkages into a single code. While the network surrounded an unopened black box of “taste” in the Market Street Program, now the network generates itself fully within this black box, as a social algorithm that is also an algorithmic formalism. Put another way: The network of subjects and artworks in the Market Street Program accreted around a black box of taste, which operated between the clusters of real-world exhibitions and artworks it brought together. Now, the black box itself has absorbed both subject and work, so that the network enforces the appearance of a morphological similarity between itself and the subjects/objects within it. Taste is replaced by a hyperlinked connectivity that seems to derive from the images themselves, as in sets of photos on Facebook that are immanently networked via sorting features both metonymic (the album) and metaphoric (the tag, the link, the like).

Founded by Carter Cleveland, the twenty-five-year-old son of art collector David Cleveland, Art.sy is financially and strategically buttressed by a network of upper-echelon investors and advisers from the technology, media, and art worlds: The celebrity roster includes Dasha Zhukova, Larry Gagosian, Wendi Murdoch, Peter Thiel (of PayPal), Jack Dorsey (of Twitter), and Joe Kennedy (of Pandora). These backers have placed at Art.sy’s disposal resources ranging from advisory roles to token contributions to considerable investments. In line with the site’s aspirations as both a pedagogical resource and a commercial venture, the roles of adviser (such as that played by the prominent curator John Elderfield) and investor are blurred. After obtaining an initial seed grant of $160,000, Art.sy raised $1.25 million in a second round late in 2010 and $6 million in a third round late last year. With each influx of funding, the algorithm used by the site has grown more effective. More money, one imagines, buys more “data points”: the formal, historical, and contextual qualities the site uses to link one work to another. As of a November 2011 article in Wired, the site included more than 550 criteria, against which each work takes a value from 0 to 100. When I saw the site in February, Art.sy was operating with roughly 800 data points organized into fourteen broad categories, as well as tags that identify characteristics including medium, location, influences, content (“chair”), concept (“identity politics”), and texture (“shaggy”).

Art.sy’s masthead ties together major figures from the separate spheres linked by the site—media both traditional and social, museums and power galleries, and finance (from old-style oil-derived oligarch capital to its pixelated infrastructure in PayPal). Even Google has lent its name in the form of ex-CEO Eric Schmidt, an angel investor. Which makes sense, considering Art.sy’s ambitions as a search engine for contemporary art. As Google’s engineers created a window onto the universe of networked information, with a concomitantly hyperconnected weltanschauung, Art.sy claims to be building a new way to experience art and its own internally networked connections. To “experience” art, in all its contemporary senses: If looking-at (the museum experience) has been collapsed into purchasing (the gallery), Art.sy purports to overlay these experiences with those of the link and the suggestion.

Liz Markus, Midnight Sun, 2007, acrylic on unprimed canvas, 48 x 60".

Its aims are therefore double, as are its target audiences. On the one hand, Art.sy proposes itself as an educational tool to help art-world outsiders navigate the esoteric world of contemporary art by linking new work to canonical predecessors; to display the range of connections between movements, artists, and individual pieces in an easily apprehensible manner, providing a tailored gallery experience even to users priced out of the art market, geographically distant from museums and galleries, or socially barred (teen taggers Art.sying Basquiat discover Dubuffet). On the other hand, Art.sy purports to be a next-gen eBay, marketing contemporary works to potential collectors—especially new collectors at a remove from the gallery and auction house, who would otherwise remain unaware of such commodities, or those whose disposable income doesn’t trigger the attention of an art consultant.

Art.sy has compared its technology and business models to those of music-industry innovations such as Spotify and Pandora (whose “Music Genome Project,” a formulation worthy of Richard Dawkins, inspired what Art.sy calls its “Art Genome Project”). Spotify offers an extensive searchable online music database linked to social-networking media, in contrast to Pandora’s feedback-driven, automated “stations” playing licensed content (the difference between making a playlist on demand and listening to a specific radio show). Art.sy may share aspects of a corporate genotype with these companies, especially in its extended negotiations with museums and galleries for content, but its phenotype—its manifestation on the screen—will be somewhat different. With Art.sy, a set of images that share “DNA” with a source image are clustered in a grid, in an interface that recalls forms of targeted marketing found on Amazon, Google, Netflix, and so on. Clicking on one of the images produces a new cluster. Before viewing the site, I had expected that the search algorithm would make use of established artificial intelligence techniques to learn user habits and augment the results accordingly. Unfortunately, this does not seem to be the case, at least at this stage in Art.sy’s development; the search algorithm seems to be designed top-down, irrespective of user feedback or crowd-sourcing data. Art.sy’s rationale for resisting collaborative filtering is tied to its educational mission; i.e., the site doesn’t want to dilute curated expertise with collectively sourced networks. This insistence on top-down pedagogy, however, may also be part of a long-term business strategy in which Art.sy continues to be profitable in the absence of a workable consignment-sales model, through possible funding from foundations or revenue derived from subscriptions sold to schools.

Art.sy revisits models that tried, to no enduring success, to create online art markets during the first dot-com boom of the 1990s, with the hope that changing conditions of consumption (for both art and other luxury goods) will render these models more viable: Younger consumers inured to Internet purchasing no longer view large online expenditures with suspicion. And while the models at issue here initially arose during an economic boom, Art.sy is redeploying them in a recession. It is no secret that at present, art’s value is largely generated by its ability to constitute itself as a precisely calibrated and targeted network, both formal and social. (Damien Hirst’s simultaneous flood of dots into every node of the gallery network and social media is a particularly focused example.) Such a “networked” model, which may have seemed provisional in its tragic phase, could now, in its farcical phase, become final—with implications for the appearance of the image itself.

The status of the artwork as an image on a screen is already changing the way in which art is marketed, and not simply by the proliferation of dedicated iPads at galleries and fairs. A British site named s[edition] has taken this change quite literally, devoting itself to the sale of ultrahigh-resolution, numbered digital editions by artists such as Damien Hirst and Tracey Emin, presumably intended for flat-screen display in the home. Immediately shared through social media, purchases on s[edition] are identified by buyer on the page dedicated to each edition, making resale value inextricable from social capital—a high-end status update.

A less direct, but perhaps more expansive, example can be found in the influential art blog Contemporary Art Daily, which has effectively dematerialized the white cube into the white screen, creating new conditions of circulation and even, I think, magnifying mutations in the art objects and markets that it merely reproduces via gallery press releases and high-quality install shots. Mutations already in evidence include a polarization of colors (the muted stain paintings of Sergej Jensen; the high-key psychedelia of Kerstin Brätsch), which takes advantage of the attention economy of tablet screens by either therapeutically toning down brightness or hyperbolically amplifying it. If we were in the business of trend spotting, we might predict an upswing in the work of Morris Louis (and even Jules Olitski), something that Art.sy is already figuring by including a painting derived from Louis’s “Veils,” Liz Markus’s Midnight Sun, 2007, on its preview slide show.

Drawing by Nick Mauss in response to “Pandora’s Black Box.”

In anticipation of Art.sy’s launch, then, we can ask a few speculative questions, in which “Art.sy” should be understood, at least in part, as a metonym for a broader situation.

First: Similar ventures provide some clues as to the identity and tax bracket of the potential Art.sy user. Paddle8, “a virtual gallery space,” currently lists only twenty-eight works priced above $50,000, topping out around $350,000 (with another eleven listed as “price available upon request”). On the other hand, the site lists dozens of pages of works priced between $1 and $20,000. We expect a similar market position from Art.sy: a steady aim at the lower upper class. Those above the threshold presumably have no need of Art.sy’s services. For those below, Art.sy may function as an educational public-works project (as much of their press presents it) while still building virtual/notional value for contemporary artists and galleries, like a high-end jewelry catalogue stapled to an archaeological inventory of the gilded scarabs in King Tut’s tomb. Which is also how Art.sy aims to sell its listed wares to its target market: It constellates the works for sale in this price window, a likely majority of which might be relatively recent, relatively homogeneous production by post-MFA artists, with museum-held pieces of determined monetary and cultural value, predicating the value of the former on that of the latter via their “genetic” linkage. Previously, collection advisers acted as beards on the ignorance of their rich but naive clientele. Art.sy would be a beard for the masses, putting the contemporary white cube in the brown paper bag. Is Art.sy then an engine to pump value (derivative of canonicity) into recession-era art (derivative of MFA culture) for recession-strained buyers?

The second question depends on the amount of information that Art.sy would disclose about the gallery stocking a work of art. Would galleries start undercutting Art.sy, bypassing its commission structure in order to sell directly to new clients? Or, more interestingly, would galleries start to produce art explicitly for the Art.sy market, available only through the site? Such art would not only be tailored to whatever niche market Art.sy might potentially generate or find but also be specifically adapted to the site’s media conditions. A potential parallel: The main iteration of the online-only VIP Art Fair this past February will be followed by fairs focused on works on paper (in April) and photography (in July)—both lower-cost media often disseminated in large editions, but also media that translate well to the screen (backlit watercolors, for example).

Third, to what extent will Art.sy be able to convert the handshake model of socially generated value (as analyzed by, say, Bourdieu) into social-networking media? Art.sy aims to position itself in the place traditionally occupied by mediators between art and its markets, such as art advisers, curators, and, to some extent, art magazines. These intermediaries are also mediators between art and the museum, with the adviser selling her knowledge to the client by saying that a work on which the client places a bet will be, in the near future, museum-worthy. While advisers have the social means at their disposal to create the conditions of possibility for this bet to pay off, Art.sy seems to assert that, as a medium, it can have this capability built in. It will have the effect, in other words, of assessing future value in the way that all media have the immanent capacity for constructing innovative ways of producing future value. And this is so regardless of whether they “reproduce” current conditions, or even whether they seem plausible under current conditions. Art.sy encodes a speculative temporality—the future it inscribes for contemporary art is the same future it predicts by constellating marginal images around your “like.” To what extent will the click or the swipe of the tablet screen become central to how art conceives its own future?

Finally, from a recent article in Wired: “It seems silly to search for Picassos by color. But it will be far less silly when a search for paintings by color gives us the next Picasso. That might just be the day when the art market becomes an open market.”³ Alongside a familiar modernist claim for the intrinsic value of the image, perversely patterned on classical assertions of artistic autonomy, what we find here is a less familiar modernist coupling of democratization (the “open market”) and media-determinism. Self-generation within the system and self-determination (by the people, for the people) are collapsed. This openness-effect paradoxically depends on the black-boxing of the algorithm: Users who think they are democratically discovering the next Picasso are themselves being predicted by the determined autonomy of the image installed in the autonomous system. When such total openness coincides with such totalizing formalism, the hegemonic culture industry diverted into a million Twitter feeds, can we distinguish between users in search of a Picasso and Picassos in search of a user?

Michael Sanchez is a critic and art historian based in New York.

The author would like to thank Jeff Nagy for his collaboration in the writing of this article.

NOTES

1. On Graham’s Likes, see Eric de Bruyn, “Topological Pathways of Post-Minimalism,” Grey Room 25 (Fall 2006): 58.

2. For more on the Market Street Program, see Michael Leopold, “Computer Mating/Los Angeles,” The Art Gallery (Summer 1972): np; Peter Plagens, “Los Angeles: The Market Street Program,” Artforum, January 1972, 77–78.

3. Shahan Mufti, “Art.sy’s ‘Genome’ Predicts What Paintings You Will Like,” Wired, November 23, 2011.