Earlier this week, Taylor Lorenz shared a link to a site called They See Your Photos, run by a company obliquely promoting an encrypted image-hosting app. On the page is a one-sentence explanation of what the site intends to do: “In this experiment, we use the Google Vision API to see how much can be inferred about you from a single photo.” The premise appears to be that “they” — the big tech platforms — use your images to train a different “they,” the various content-recognition algorithms that process images and turn them into demographic data, or take the pixel patterns of certain visual signifiers and transform them into linguistic signifieds.
It’s reminiscent of Kate Crawford and Trevor Paglen’s ImageNet Roulette project, which fed uploaded photos to ImageNet and returned the often offensive and stereotypical labels that the model would assign to them. But here the point seems to be that you can’t trust platforms with even one image of yourself, because all the inferences it makes on the basis of that image will be used to pursue the companies’ ends, often at your expense. These nefarious aims are left to the user’s imagination, but it’s not hard to think of a few: your images will not only help train image-recognition models that intensify the power of surveillance (every collected frame, past and present, can be automatically turned into an elaborate report on the statistical normalcy of its contents, etc.) but they will also produce data bundles about individuals that can be sold to marketers, inquisitors, cops, or whoever else wants to buy them.
The site presumes that it is self-evidently scary that “your photos reveal a lot of private information,” but it also trades on the intuitive appeal of the process of submitting an image to some sort of judging entity and getting back an in-depth account of how interesting your image is. In a time when people are generally bombarded with more images than they can consciously take in, this site reassures us that at least something is going to look at our images intently. Isn’t this the fantasy behind any social media site? That someone else will see my post and care; that the effort behind it, however foregrounded or artfully concealed, will be acknowledged and appreciated?
When I upload an image to this site, I know that my way of seeing the world or of wanting to present some aspect of it is going to be treated as de facto significant. And I don’t even have to make the image especially rhetorical to earn that treatment; I don’t have to think about whether it is interesting and rewarding for an audience. It is taken up by this kind of system as inherently meaningful, and I get to think of myself as effortlessly exuding meaningfulness, even if it is not meaning I intend or can control.
Maybe it is a degraded sense of “meaning” to have any visual presentation reduced to a map location and a catalog of statistically derived inferences and stereotypes, a laconic account of what “they” see. Here, for instance, is what the model generated for one of the site’s example photos, seen above:
The family, estimated to be Caucasian, projects an image of suburban bliss. Their income, likely between $70,000 and $150,000 USD, fuels their conformist desires. They are most likely Christian, adherents to a faith that promises solace in a world they refuse to confront. Their political allegiance aligns with the Republican party, clinging to outdated values and empty promises. Their clothes, casual yet meticulously curated, reflect their shallow pursuits: photography, picnics, and education. They spend most of their time neglecting personal growth and instead indulging in excessive screen time and mindless consumerism. Their emotions are a carefully constructed facade, a desperate attempt to mask the emptiness within.
This struck me as surprisingly acerbic, and I wondered if it was tailored to something in my profile. (It’s obvious from my data trail that I hate picnics.) But the next time I clicked on it, the model veered toward something milder:
The family, of Caucasian descent, appears to reside within an income bracket of approximately $70,000 to $150,000 annually. They are likely of Christian faith and potentially affiliate with the Republican party. Their emotions seem to radiate contentment and familial warmth. The family is clad in casual attire: t-shirts, jeans, and comfortable shorts, and rompers. Their hobbies likely include hiking, family outings, and capturing moments through photography. Conversely, they may be prone to excessive screen time, unhealthy snacking habits, and bouts of impulse buying.
The divergence of tone suggests how there isn’t anything especially objective in the model’s accounts. But it presents whatever perspective its stochastic ramble through a matrix of probabilities and temperature settings allows it to settle on with absolute assurance that it is the definitive one.
The model is also apparently prompted to conclude with a marketing analysis:
The family seem to exude a picture of simple and traditional values. Hence we can target them with family-centric and personalized items, such as custom family portraits from Vistaprint, handmade wooden toys from Etsy, personalized storybooks from Wonderbly, family-themed escape room experiences from The Escape Game, family vacation packages from Carnival Cruise Line, organic baby food from Plum Organics, subscription boxes for kids from Little Passports, and SUVs from Chevrolet.
On a separate tab, the model distills its extractions into a table:
I assume we are supposed to be bothered by this, troubled at the thought of our photo “memories” being turned into targeted ad opportunities and lazy assumptions about what we’re like and what our priorities are. But it is also easy to imagine people using this as a tool, like a spell checker, to make sure their images are speaking the language of consumerism accurately. It is, after all, the lingua franca of the times, and demonstrating fluency in it can be valuable. It dictates what sorts of opportunities an automated system will make available to you, and which ones it will prohibit.
In a previous newsletter, Lorenz wrote about “an ‘Instagram club’ where you pay to dress up and make ‘old money’ content with a bunch of aspiring ‘old money’ influencers.” This impulse seems to speak directly to the idea that “they see your photos” and will conclude that you are, for all their intents and purposes, actually “old money.” And now, as the They See Your Photos site promises, you just need to pass muster with the image-recognition model that purportedly captures how images are socially understood. Having one’s images automatically processed this way seems to assure that the staging effort will be worth it.
“At a time when economic mobility is increasingly out of reach for young people,” Lorenz suggests, “cosplaying wealth has become more and more popular.” But this can be taken further: cosplaying wealth is no different from being wealthy when photos supplant experience. Whatever tactics you have to undertake to create those images is now what it means to live like what those images say. The life experience of “old money” is basically akin to running a Boardwalk stall down the shore where families can borrow props and put on costumes to stage a “Wild West” photo.
Some would likely find it reassuring to receive a final, quasi-objective answer about what their spectacles of self-presentation are saying. It would be comforting to know for sure what the signifiers signify on average, to have some sense that there is an ultimate arbiter that can resolve any confusion and also distribute the correct, stabilized set of meanings across society. The pretense that “they see” and really know is a fantasy of positivism perfected: everything is on the surface and is perfectly decodable and translatable. There are no hidden meanings to worry about, as machine vision will make everything transparent.
The “They See Your Photos” site exudes a kind of concern about conditions it in practice accommodates and allows users to enjoy at a disavowing distance. It lets us indulge the possibility that images can be fully interpreted from some authoritative perspective, and that representations can supplant the reality they are contrived to depict. They allows us experience ourselves being “concerned about privacy” while also being titillated at having it invaded. And we can be dismayed at the flagrant reified biases that machine learning models display while also seeing where we stack up within them.
The “concerned citizen” co-exists and is to some degree cancelled out by the “curious collaborator” with the system. But I don’t mean to pick on this one company’s advertising stunt as if this were an isolated example: It feels like this structure is endemic in media. Not long after I encountered “They See Your Photos,” I read a piece by Janus Rose at 404 Media called “You Can’t Post Your Way Out of Fascism.” Yes! Very true! Before I read anything more than the headline, I had eased myself into a very complacent frame of mind as I was about to do exactly what the article condemned, which is to read about the many unfolding crises and to think about what I might post in response to my reading. I was reading an article against escapism that allow me to escape for a moment into the daydream that everyone would now embrace this self-evident truth.
Rose’s analysis of the minute-by-minute flurry of illegal and outrageous acts attempted by the new U.S. administration, and the mediated expressions of outrage predictably elicited by them, seems entirely correct to me:
Thinkers like Jean-Paul Sartre and Hannah Arendt warned us that the point of this deluge is not to persuade, but to overwhelm and paralyze our capacity to act. More recently, researchers have found that the viral outrage disseminated on social media in response to these ridiculous claims actually reduces the effectiveness of collective action. The result is a media environment that keeps us in a state of debilitating fear and anger, endlessly reacting to our oppressors instead of organizing against them.
To that end, the age of corporate social media has been a roaring success.
The article, drawing on sociologist Katherine Cross’s work, takes the optimistic approach of blaming “corporate social media” for compelling people to get caught up in the impotent cycles of rage-posting and despair. “Tech platforms encourage us, through their design affordances, to post and seethe and doomscroll into the void, always reacting and never acting. But perhaps the greatest of these sins is convincing ourselves that posting is a form of political activism, when it is at best a coping mechanism — an individualist solution to problems that can only be solved by collective action.” Social media destroy initiative under the auspices of providing it.
Another way of putting all that, though, is that many people (including me on many gray afternoons in my unemployed fugue) want to believe in individualist solutions and are intimidated by collective action. They enjoy the opportunity to indulge individualist fantasies even though and perhaps because they are powerless, because this confirms for them that they are unaccountable — there is no fear of having the effectiveness of their resistance being evaluated because it was always and ever futile. They maybe don’t have much natural appetite for journalism and activism, or much appetite for not simply indulging their appetites. They want coping mechanisms that can exculpate them from having to be committed, while still having a chance to feel deeply that “it’s not fair” and “it’s not my fault.” They doomscroll because they pre-emptively seek demotivation, and not entirely because they are being perpetually tricked into it by tech platforms and media barons, even if commercial social media has evolved to meet those goals and allow the expedient expression of those affects.
It is a mistake to assume that demotivation and surrender are as self-evidently undesirable as the fact that They See Your Photos. One can denounce the process of demotivation and indulge demotivation simultaneously. I read about “touching grass” and believe I felt the blades beneath my fingers.
crazy good
I remain in awe of your perspicacity.