AI that turns Museums into Conversations: The Digital Twin

Discussion about AI and the arts can be abstract, both on the up- and downsides. I’d like to offer a concrete potential use that could be transformative. This one is for museums. Next week I’ll offer an idea for performing arts.

Arguably, one of the biggest transformations in the arts over the past thirty years has been the shift in relationship between artists, institutions and their audiences. Institutions went from making things and just putting them out there, to focusing more on how to engage audiences.

Also arguably, for the past thirty years, engagement has been the most abused word in the museum lexicon. It is ubiquitously front and center in grant proposals, strategic plans, press releases, and postmortems, invoked as both a diagnosis and cure. When attendance wanes the prescription is almost always more engagement. More digital. More interactivity. More… noise.

But museums have largely mistaken engagement for amplification.

Since the digital revolution, institutions have mostly wrapped technology around art rather than rethinking the relationship between art and visitor. We built apps and social media feeds and touchscreens and audio tours. We created video and expanded wall texts. We treated digital tools as a way to broadcast louder instead of connecting deeper.

Meanwhile out in the world, “engagement” in the digital age was redefined by being almost anti-broadcast. Digital culture moved quickly from read-only to read-write—from transmission to dialogue, comments instead of captions. Feeds algorithmically tuned to behavior. The big social media platforms learned, sometimes uncomfortably, that meaning is negotiated, not delivered.

Museums didn’t, for the most part, learn these lessons.

Whether you are a scholar, a tourist, or a bored teenager, the museum interacts with you in the same way: a 150-word wall label that compresses years of research and meaning into arid paragraphs. Experts and novices are treated identically, as if curiosity is a generic value. Instead of understanding that interpretation is highly dependent on each visitor’s life experience and context, museums fix their points of view in advance, frozen in text, and largely oblivious to how different visitors actually think or learn or experience.

This isn’t a technology issue. It’s an epistemology problem, a belief about how knowledge flows and who is allowed to shape it.

Museums still operate as if interpretation is a one-way stream, produced by experts and consumed by the public. Visitors may respond emotionally, but the system itself does not register that. It doesn’t detect confusion, fascination, resistance or boredom, let alone adapt to any of these reactions. We measure the bodies that come through the door, not the engagement of those in the galleries.

But the gold standard of museum engagement has always been fully interactive and analog. Some of my best experiences in museums have been with curators touring me through a show. Ask a question about a detail or a painting you barely noticed, and it lights up as they explain. Wonder about the motivation for how an exhibition was put together and you get a story that hooks you in to the concept. Ask about technique and the conversation dives into pigments and process. Ask about history and you’re into a discussion about geopolitics, trade routes, patronage, ideology. The curator listens, adapts to what you’re reaching for, and draws on a lifetime of knowledge to meet you where you are.

This isn’t about more information. It’s about responsiveness. About being understood while you look and meeting you where you are.

Museums own enormous stores of knowledge—curatorial notes, conservation imaging, scholarly essays, archival interviews, microscopic analyses, social and historical context—but there hasn’t been a way to scale that knowledge to respond to a specific person at the moment their curiosity awakes. The best we could do was one-size-fits-all synthesis, better than nothing (like the old days when museums merely labeled what an object is and when it was acquired) but not really very satisfying. Even audio tours, loved by many, are limited.

Instead, imagine an exhibition that doesn’t just speak, but listens and responds. An AI that treats interpretation not as a finished product, but as a live exchange. One that allows meaning to be shaped in real time by the questions visitors ask and the interests they express.

I call it a Digital Twin of the exhibition—not a marketing layer or a novelty app, but a conversational AI model built entirely from the museum’s own scholarship. Not a chatbot hallucinating answers, but a closed system grounded in the institution’s research, debates, and uncertainties, and informed by whatever historical, sociological, and scientific context the curators add

You enter the gallery and the system knows where you are standing through proximity beacons. You look at a sculpture and ask, “Why does this feel unfinished?” or “Who paid for this?” or “Explain this to me like I’m a painter.” The response isn’t a lecture. It’s a dialogue. The system adjusts as you adjust. If you ask about materials, it follows you. Shift to politics and it pivots. A visitor can choose the style of conversation: Playful? Scholarly? Enigmatic? Do you want this to be an educational experience or entertainment?

Instead of one authoritative tour, interpretation becomes plural. Curators shift to design lenses rather than scripts. An Indigenous scholar can overlay a counter-narrative that complicates a traditional framing. An artist-in-residence can annotate the show from inside the practice. Invited guest curators could plan different tours with different emphases and points of view. Visitors could generate their own pathways by saying what interests them.

Technology is often framed as democratization of access, but using AI in this way speaks to a deeper shift in relationship and authority. Museums are public-facing institutions, but they have largely remained epistemologically closed. Listening destabilizes control and shifts the power of the experience. It introduces the possibility that visitors care about different things than curators expect. Which is precisely why listening matters. It’s not about giving up power but in building more powerful influence.

Getting to Know You

There’s another compelling reason for considering this kind of AI. Today, museums know remarkably little about their visitors. They track attendance, ticket type, demographics, and where the crowds congregate, in other words, logistical metrics. But they have little insight into what visitors are thinking as they go through a show, where curiosity takes hold, where attention wanes, or which questions repeat themselves from visitor to visitor.

A Digital Twin changes that. Not by surveilling behavior, but by seeing what questions visitors ask, what they’re interested in and what stories and information register. For the first time, an institution could see patterns of thought: what people ask, what they skip, where meaning resonates. The museum shifts from a broadcasting tower to a listening engine.

This has enormous consequences. It challenges how exhibitions are designed, how education is structured, and even how success is defined. It forces institutions to confront the gap between what they value and what visitors want from the experience.

The Afterlife

It also addresses another long-standing problem: the disappearing exhibition. Museums spend years assembling shows, researching, negotiating loans, building rationales and stories, only for the exhibition to disappear into crates and catalogues after their runs. If it survives at all, it’s as static documentation, read by scholars.

A Digital Twin creates an afterlife. The exhibition survives as a living archive of curatorial thinking, revisable, annotatable, expandable. Interpretation becomes cumulative instead of disposable. A museum building its exhibitions in this manor could ultimately possesses a digital twin of the museum itself, which visitors can explore whether in the museum or at home.

None of this replaces curators. It scales them up, extending the reach of a voice that currently touches thirty people on a Tuesday afternoon to anyone willing to engage on their own terms. It makes visible not just conclusions, but processes, disagreements, and doubt. This is art as a process rather than an assemblage of static objects.

Many worry that AI will hollow out expertise, and not without reason. But I think there is an even greater opportunity for human curators with expertise to use these tools in ways that makes curatorship resonate more individually and more broadly. Additionally, as AI seeps in to more and more of our daily experience, the ways we will want to learn and experience things will inevitably adapt. This alone argues for the need for museums to evolve from presenting experiences that assume the generic to those that speak individually to us.

The most powerful museum moments are not informational, they’re relational, experiential, that moment when something on the wall resonates on a visceral level. We already talk to art in our heads. So maybe use that as a starting place. What would it mean to extend that conversation and have the opportunity to maybe shape where it goes?