An Algorithm is a Camera
Algorithmic Modeling, Big Data, Objectivity

An Algorithm is a Camera


By Ilan Ben-Meir

In 1958, the French film critic André Bazin published an essay titled “The Ontology of the Photographic Image,”¹ in which he sought to define an essential difference between photographs (the representations of the world produced by cameras) and the images produced by “the plastic arts” (other forms of representation, such as painting, sculpture, etc.). Bazin claims that the history of people making images consists of various attempts to ward off death by capturing images of reality in its truth, such that they are able to place what they picture outside the reach of the normal progression of time — at least, until the technological precursors of photography began outsourcing the function of preserving the images of things as they actually were to a techno-mechanical process.

With “the discovery of the first scientific and already, in a sense, mechanical system of reproduction, namely, perspective: the camera obscura of Da Vinci,” Bazin argues, “[…] painting was torn between two ambitions: one, primarily aesthetic, namely the expression of spiritual reality wherein the symbol transcended its model; the other, purely psychological, namely the duplication of the world outside.”² Perspective tempted painters to think that painting could produce true images of the world, both as it seemed and as it actually was, but photography soon banished this illusion.

In Bazin’s account, the invention of photochemical photography in the 19th century “freed the plastic arts from their obsession with likeness. Painting was forced, as it turned out, to offer us illusion and this illusion was reckoned sufficient unto art. Photography and the cinema on the other hand are discoveries that satisfy, once and for all and in its very essence, our obsession with realism.”³ Once the world could be photographed, it no longer needed to be painted in order to be remembered as it really was — and once photographs could give us “the unadulterated facts,” paintings were freed of the responsibility of objectively recording the world, and could instead focus on representing the world as an “artistic” subjectivity perceives or experiences it.

Bazin explains the mechanism of this shift in precise detail: “No matter how skillful the painter, his work was always in fee to an inescapable subjectivity. The fact that a human hand intervened cast a shadow of doubt over the image.” Unlike a painting, Bazin explains, the representation of the world generated by a camera “completely [satisfies] our appetite for illusion by a mechanical reproduction in the making of which man plays no part.” Bazin’s belief in the objectivity of photography follows from his belief that there is no point at which the artist’s subjectivity can influence the image captured on film, once the process of generating a photograph has been initiated. As Bazin puts it:

For the first time, between the originating object and its reproduction there intervenes only the instrumentality of a nonliving agent. For the first time an image of the world is formed automatically, without the creative intervention of man. The personality of the photographer enters into the proceedings only in his selection of the object to be photographed and by way of the purpose he has in mind. Although the final result may reflect something of his personality, this does not play the same role as is played by that of the painter. All the arts are based on the presence of man, only photography derives an advantage from his absence. Photography affects us like a phenomenon in nature […]⁴

These declarations land somewhere between enthusiasm and ecstasy (in all senses, including the religious), but they nevertheless acknowledge that the supposed objectivity of the photograph is bounded on all sides — framed, we might say — by “the personality of the photographer,” as expressed in his choice of subject and the composition of his shot (and to this, we should add, by the way he instructs the camera to behave by selecting the lens aperture and length of exposure — which are themselves best understood as subsets of the more general question of which camera the photographer will choose in order to capture a given shot).

Bazin leaves these qualifications by the wayside, however, as he goes on to assert that “[t]he objective nature of photography confers on it a quality of credibility absent from all other picturemaking. In spite of any objections our critical spirit may offer, we are forced to accept as real the existence of the object reproduced, actually re-presented, set before us, that is to say, in time and space.”⁵ Bazin is explicit that, at least psychologically, a photograph causes a “transference of reality from the thing to its reproduction,”⁶ because “[t]he photograph as such and the object in itself share a common being.” Bazin elaborates on this formulation in a grand pronouncement:

Only a photographic lens can give us the kind of image of the object that is capable of satisfying the deep need man has to substitute for it something more than a mere approximation, a kind of decal or transfer. The photographic image is the object itself, the object freed from the conditions of time and space that govern it. No matter how fuzzy, distorted, or discolored, no matter how lacking in documentary value the image may be, it shares, by virtue of the very process of its becoming, the being of the model of which it is the reproduction; it is the model.

At the heart of this rhetoric is the idea that a photograph takes an impression of reality, and the fact that it does so purely mechanically means that the impression of reality taken by a camera is different in kind from the kind of subjective interpretation that we mean we refer to our “impressions” of someone or something (hence the divergence between photography and Impressionist painting).

Along similar lines, Roland Barthes writes in Camera Lucida, his book of philosophical reflections on photography, that a photograph is a “weightless, transparent envelope”⁷ around its subject, because it is “literally an emanation of the referent.”⁸ He therefore concludes that photography’s “essence is to ratify what it represents,”⁹ and claims that “the Photograph is indifferent to all intermediaries: it does not invent; it is authentication itself; the (rare) artifices it permits are not probative; they are, on the contrary, trick pictures. Photography never lies […].”¹⁰ Thus, Barthes insists that “despite its codes, I cannot read a photograph.”¹¹

The general understanding of the photograph, then, even among post-structuralist thinkers such as Barthes, is that it embodied a certain kind of absolute, material assurance of reality — that by looking at the world as a camera sees it, one could look through the biases distorting one’s own subjective perceptions and gaze directly at The Truth. Photography, therefore, was paradoxically perceived as something of an unmediated medium; the photograph as traditionally construed was a technology of objective, rather than processual, presence — it offered, or seemed to offer, representations devoid of subjective influence.

As deepfakes and other AI-generated images become increasingly commonplace in today’s world, it seems hopelessly naive to claim that “photography never lies” because “the Photograph […] does not invent; it is authentication itself” — but it is easy to explain the shift in how people perceive photography as a direct consequence of the medium’s digitalization. As William J. Mitchell wrote in 1992, two years after the release of the first version of Photoshop:

The camera has commonly been seen as an ideal Cartesian instrument — a device for use by observing subjects to record supremely accurate traces of the objects before them. It is supereye — a perceptual prosthesis that can stop action better than the human eye, resolve finer detail, remorselessly attend to the subtlest distinctions of intensity, and not leave unregistered anything in the field of its gaze … The photographic procedure … seems to provide a guaranteed way of overcoming subjectivity and getting at the real truth.¹²

Following the appearance on the market of consumer-level digital cameras, however, Mitchell argues that “photography was dead — or, more precisely, radically and permanently displaced”¹³ due to a new manipulability of the image.

“When we look at photographs we presume, unless we have some clear indications to the contrary, that they have not been reworked,” Mitchell explains, whereas “the essential characteristic of digital information is that it can be manipulated easily and very rapidly by computer. It is simply a matter of substituting new digits for old. Digital images are, in fact, much more susceptible to alteration than photographs, drawings, paintings, or any other kinds of images.”¹⁴ Bernard Stiegler echoes this claim in his 2002 essay “The Discrete Image,”¹⁵ arguing that when “photons become pixels that are in turn reduced to zeros and ones on which discrete calculations can be performed,”¹⁶ the digital photograph becomes susceptible to manipulation, and can therefore be made to testify to events that never occurred. Stiegler notes that

[t]his possibility, which is essential to the digital photographic image, of not having been, inspires fear, for this image, at the same time that it is infinitely manipulable, remains a photo, it preserves something of the this was within itself, and the possibility of distinguishing the true from the false dwindles in proportion as the possibilities for the digital treatment of photos grow.¹⁷

As a result, Mitchell concludes, society is “faced … with a new uncertainty about the status and interpretation of the visual signifier.”¹⁸ In a post-digital world, there is “no way to determine” whether a given image “is a freshly captured, unmanipulated record or a mutation of a mutation that has passed through many unknown hands.”¹⁹ The newfound susceptibility to manipulation of digital images that appear, for all intents and purposes, to be photographs leads to a general loss of faith in the unimpeachability of “photographic” testimony — which ultimately results in a diminution in the authority accorded to analog photographs, as well.

While Stiegler agrees with Mitchell’s argument that the advent of digital technologies will lead to a shift in the cultural perception of images that appear to be photographs, he takes pains to emphasize that the re-evaluation of photography in light of its digitalization corrects a previous understanding that had always been mistaken. Indeed, Stiegler is explicit about the fact that while the digital image is in its very essence a “discrete” image, “the analog image is always already discrete. Not simply because it is composed of atomic grains, but because it is subject to framing operations and choices about depth of field, because it has its reality effect according to the photographic and literal context in which it is inserted.”²⁰

Mitchell offers a subtler version of the same claim when he asserts that one can view “the emergence of digital imaging as a welcome opportunity to expose the aporias in photography’s construction of the visual world, to deconstruct the very ideas of photographic objectivity and closure.”²¹ In other words, photography was never what the culture made it out to be — the advent of digital photography simply made apparent the fact that photographs are representations like any other, invitations to interpretation rather than testaments to a univocal truth.

Even Barthes allows for the possibility that photography was never quite as objective as he claims, following his declaration that “Photography never lies […]” with a crucial qualification: “[…] or rather, it can lie as to the meaning of the thing, being by nature tendentious, never as to its existence.”²² Even in Barthes’ account, the only “fact” testified to by an analog photograph is the fact that whatever a photograph shows is a result of the procedure of photography (which produces images of “reality” by recording how light interacts with both the external world and the specific film inside the camera). To derive anything “meaningful” from this record requires its interpretation, which is always at least a partially subjective process — and the fact that the film being used, the position of the camera, the length of exposure, and the focal length employed all have an impact on the image that a camera produces means that subjectivity even enters into the process of the creation of a photographic record, itself.

Let us linger for a moment over this point, as it is the crux of this essay’s argument. What makes photographs representations — which is to say interpretations — of reality is not the existence of Photoshop; rather, it is the simple fact that a photograph and the reality that it represents are two different things, such that “reality” is somehow transformed into “a photograph (of reality)” by the procedures of photography. If no such transformation occurs, there would be no way for people to differentiate between a photograph of something and the thing itself — the “map” provided by a photograph would be identical with the reality of the “territory” being mapped. There is more information in reality itself than there is in a photograph thereof; the photograph is produced by subtracting from reality all the kinds of information (smell, temperature, etc.) that a photograph does not record.

It is at precisely this point that Bazin’s understanding of photography goes astray. In order to understand this slippage in Bazin’s thought, however, it is necessary to first untangle a snarl in his language — namely, his decision to refer to a photograph as “the reproduction” of an original something, which he calls “the model.” Let us instead borrow Roland Barthes’ terminology from Camera Lucida, and substitute the term “referent” for Bazin’s “model.” Barthes writes: “Photography’s Referent is not the same as the referent of other systems of representation. I call ‘photographic referent’ not the optionally real thing to which an image or a sign refers, but the necessarily real thing which has been placed before the lens, without which there would be no photograph.”²³

The immediate consequence of this change in terminology is to clarify a fundamental confusion that runs throughout Bazin’s essay: a photograph is not a “reproduction” for which “reality” has served as the “model”; rather, a photograph is itself one kind of “model” of a “referent” that exists in a “reality” that the photograph abstracts away. (Recall Bazin’s claim that “[t]he photographic image is the object itself, the object freed from the conditions of time and space that govern it” [italics added] — in other words, a photographic image represents a referent by modeling that referent in a system that abstracts away the spatiotemporal parameters and constraints that “govern” reality as we experience it.)

Therefore, when Bazin declares that “[o]nly a photographic lens can give us the kind of image of the object that is capable of satisfying the deep need man has to substitute for it something more than a mere approximation,”²⁴ he is fatally and fundamentally mistaken. A photograph is no less an “approximation”/representation/model of reality than a painting; it is simply an approximation of a different and less obvious sort than the more obviously figurative arts. If Bazin were to have replaced a thing in the world — say, a cow — with a photograph of that cow, he would not have “substituted for it something more than a mere approximation.” Rather, he would have substituted for the cow nothing more than a mere approximation, while simultaneously convincing himself that he had done just the opposite — a neat trick, at least until he runs out of milk, for one cannot milk a photograph of a cow.

Bazin’s basic error lies in his erroneous characterization of the relationship between the models/representations of reality generated by photography and the reality that such models take as their referent. In Bazin’s account, photography does not merely produce representations or models of reality; rather, “the photographic” is one register in which reality manifests. Ultimately, this leads Bazin to define photography as “the natural image of a world that we neither know nor can see”²⁵ a techno-mechanical procedure with the “power to lay bare the realities” of the world. “The impassive lens, stripping its object of all those ways of seeing it, those piled-up preconceptions, that spiritual dust and grime with which my eyes have covered it,” Bazin explains, “is able to present it in all its virginal purity to my attention and consequently to my love.”

The lens of the camera, Bazin argues, gives us a way to assess the world objectively — to purge our vision of our personal biases and established habits of mind — and therefore allows us, paradoxically, to see through the surfaces of things, to look beyond superficial appearances and gaze at “the natural image of a world” that was always already there, but which only the camera can make accessible to our vision, and by extension, our knowledge.

If such rhetoric sounds familiar, it is because it is no less pervasive in today’s culture than it was at the dawn of photography — we have just detached it from photography, and attached it instead to algorithmic and computational modeling. As Cathy O’Neil writes in Weapons of Math Destruction, “by 2010 or so, mathematics was asserting itself as never before in human affairs, and the public largely welcomed it.”²⁶ O’Neil explains that in what she calls “the Big Data economy” of today’s society

[a] computer program could speed through thousands of résumés or loan applications in a second or two and sort them into neat lists, with the most promising candidates on top. This not only saved time but also was marketed as fair and objective. After all, it didn’t involve prejudiced humans digging through reams of paper, just machines processing cold numbers. […] Nevertheless, many of these models encoded human prejudice, misunderstanding, and bias into the software systems that increasingly managed our lives. Like gods, these mathematical models were opaque, their workings invisible to all but the highest priests in their domain: mathematicians and computer scientists. Their verdicts, even when wrong or harmful, were beyond dispute or appeal.²⁷

In O’Neil’s account, “the algorithm” has come to replace “the camera” as the technology of objective truth — the machine that allegedly eliminates subjectivity from our encounters with reality, but that actually camouflages that subjectivity behind a mechanical apparatus that makes its presence easy to overlook, or to dismiss.

This understanding is mistaken, of course. “A model,” O’Neil explains, “is nothing more than an abstract representation of some process, be it a baseball game, an oil company’s supply chain, a foreign government’s actions, or a movie theater’s attendance. Whether it’s running in a computer program or in our head, the model takes what we know and uses it to predict responses in various situations.”²⁸ According to O’Neil, however, today’s algorithmic models too often “define their own reality and use it to justify their results,” leading her to argue that “[i]nstead of searching for the truth,” the model “comes to embody it.”²⁹

O’Neill criticizes models in words that could as easily be describing the fundamental error in Bazin’s analysis: his confusion between “model” and “referent,” which leads him to misidentify reality as the model for the photograph, and the model of reality produced by photography as identical with the reality to which it refers. As O’Neil explains, there will “always be mistakes” in a model’s representation of reality “because models are, by their very nature, simplifications. No model can include all of the real world’s complexity or the nuance of human communication. Inevitably, some important information gets left out.”³⁰ The attempt to “minimize […] human bias” in interpreting a situation by putting one’s trust in the supposed objectivity of the model does not result in the minimization of bias, but simply in its naturalization, as an understanding of the sort that can only be arrived at through a (necessarily-subjective) process of interpretation is presented as a univocal “truth.”

Just as Barthes’ claim that a photograph offers inarguable testimony to the existence of its referent ultimately collapses into the much more modest — but also more defensible — claim that a photograph testifies to the fact that something in reality was photographed, the claim that a model can show us the truth of a situation collapses into the humbler claim that the only “unimpeachable” testimony offered by a model is to the fact that something in reality was modeled. Like a photograph, a model can also offer insights into the procedures that produced it, and how those procedures interact with the precise portion of reality being modeled. These insights will always be locally limited, however, due to the fact that both modeling and photographing reality are procedures by which reality, which is continuous, is abstracted and reduced into a representation of reality, which is discrete.

Like a camera, an algorithm can tell us about the world — but what it tells us is not objective knowledge, so much as testimony about how the world appears from the perspective of the particular apparatus that is looking at it. That perspective is necessarily limited, however, by the fact that it excludes whatever the apparatus leaves out in order to transform reality into the form of a representation. What the apparatus “sees” will inevitably be influenced by the decisions made by its designers — that is, on subjective choices susceptible to the influence of personal biases, preferences, and values. Like a camera, algorithms and mathematical models give us a way of looking at the world from a perspective other than our own, but these perspectives are not views from nowhere — nor views of everything, from everywhere, all at once, which is what would be required for a model to align perfectly with reality. Rather, they remain perspectives: partial views of a unified whole that will ultimately spill past the edges of any attempt at its representation.

Citations

[1] Bazin, Andre. “The Ontology of the Photographic Image.” In What is Cinema? Volume 1, edited and translated by Hugh Gray, 9–17. Berkeley: University of California Press, 2005.

[2] Ibid. 11.

[3] Ibid. 12.

[4] Ibid. 13.

[5] Ibid. 13–14.

[6] Ibid. 14.

[7] Barthes, Roland. Camera Lucida: Reflections On Photography, translated by Richard Howard. New York: Hill and Wang, 1981, pg. 4.

[8] Ibid. 80.

[9] Ibid. 89.

[10] Ibid. 87.

[11] Ibid. 90.

[12] Mitchell, William J. The Reconfigured Eye: Visual Truth in the Post-Photographic Era. Boston: MIT Press, 1992, pp. 27–29.

[13] Ibid. 20.

[14] Ibid 7.

[15] Stiegler, Bernard. “The Discrete Image,” in Derrida, Jacques and Stiegler, Bernard, Echographies of Television: Filmed Interviews, translated by Jennifer Bajorek, 145–164. Cambridge: Polity Press, 2002.

[16] Ibid. 153.

[17] Ibid. 150.

[18] Mitchell, Reconfigured Eye, 17.

[19] Ibid. 51–52.

[20] Stiegler, “Discrete Image,” 155–156.

[21] Mitchell, Reconfigured Eye, 8.

[22] Barthes, Camera Lucida, 87.

[23] Ibid. 76.

[24] Bazin, “Ontology,” 14.

[25] Ibid. 15

[26] O’Neil, Cathy. Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy. New York: Crown, 2016, pp. 10.

[27] Ibid. 10–11.

[28] Ibid. 21–22.

[29] Ibid. 86–87.

[30] Ibid. 23.

Special thanks to Michael Zargham for conversation and inspiration, and Heather Ordway and Lila Langsford for editing and publication.


About BlockScience

BlockScience® is a complex systems engineering, R&D, and analytics firm. By integrating ethnography, applied mathematics, and computational science, we analyze and design safe and resilient socio-technical systems. With deep expertise in Market Design, Distributed Systems, and AI, we provide engineering, design, and analytics services to a wide range of clients including for-profit, non-profit, academic, and government organizations.

You've successfully subscribed to BlockScience Blog
You have successfully subscribed to the BlockScience Blog
Welcome back! You've successfully signed in.
Unable to sign you in. Please try again.
Success! Your account is fully activated, you now have access to all content.
Error! Stripe checkout failed.
Success! Your billing info is updated.
Error! Billing info update failed.