A curator is the person who chooses what hangs where and why, and their name is typically displayed on the wall of any modern gallery. It’s a powerful name. It suggests a viewpoint and a series of decisions made fully conscious of what was being omitted. Open Instagram now. There, decisions have also been made by someone or something. The subjects echo each other, the colors are harmonious, and the rhythm of the pictures seems hazily deliberate. However, the wall bears no name. Only code is present.
Researchers at Oxford’s Internet Institute have been discreetly investigating whether algorithms have assumed a curatorial role in visual culture without explicit consent. In its simplicity, their strategy was subtly brilliant. After uploading photos from the open-access collection of the Metropolitan Museum of Art to an Instagram account, they observed which photos, and in what order, the platform’s algorithm selected to display in other users’ feeds. The results were unsettling in a way that’s difficult to describe right away. They were on display at a gallery in London alongside pieces created by artist Fabienne Hess. There were preferences in the algorithm. They simply lacked aesthetic appeal.
Hess had spent three years creating what she called a Dataset of Loss, which consists of pictures collected with patience and curiosity and arranged according to memory and grief—things that are particularly human in their significance. In contrast, the algorithm was optimizing for something completely different: selling advertisements, attracting attention, and facilitating friend connections. These are not a curator’s objectives. These are a marketplace’s objectives. Nevertheless, whether or not this was the intention, the outcome—a ranked, sequenced selection of photos displayed to millions of people—acts exactly like curation.
This may be the fundamental misunderstanding that merits consideration. The Latin curare, which means to take care of, is where the word curator originates. That etymology conveys the idea that a piece of art needs to be cared for, contextualized, and shielded from misinterpretation. In contrast, algorithmic platforms optimize for engagement metrics that support platform objectives. They are not apathetic toward images; rather, they have a keen interest in those that elicit a swift response. Usually, subtlety loses. Immediate impact and vibrant color usually prevail. Even though it’s hard to quantify in a quarterly report, there is a gradually growing cultural cost to that.
Almost covertly, algorithmic systems are now the gatekeepers of creative content on social media. They mine an ever-growing ocean of images, which is made more difficult by AI tools that produce new ones at scale. This volume is unprecedented in gallery history. The algorithm was given the task without a real vote because of the sheer volume, which makes human curation impossible at scale. Every day, more than one billion photos are shared on Instagram. The algorithm intervenes, sorts them, reveals some, and conceals others, creating a visual culture that is shaped by a formula that most artists are unable to see or fully comprehend, rather than by human judgment.
In late 2024, Instagram launched a “Your Algorithm” feature that allowed users to see what subjects the platform believed they were interested in, but the ranking algorithm itself was kept a secret. Only one of those two judges is being truthful about what the system values, leaving artists to create concurrently for both an audience and a system. It’s difficult not to feel as though something has been subtly given up when artists subtly modify their work to appeal to the feed, softening edges, brightening palettes, and posting at the appropriate time.
Perhaps the most obvious thing the Oxford experiment showed was that the question is no longer theoretical. Curatorial algorithms that are unaware of larger socio-cultural contexts run the risk of creating moral dilemmas, such as the silent marginalization of work that just doesn’t work, the overrepresentation of some aesthetics, and the misattribution of cultural meaning. These are serious issues. These are the same issues that influenced discussions about gatekeeping in conventional art institutions, but the gatekeeper is no longer visible, and there is no name on the wall to hold them responsible. Whether there is a clean solution to that issue is still up for debate. However, asking the question aloud is probably the first step.
