An exhibition at the Massachusetts Institute of Technology (MIT) is making a compelling case that computing is far more than a tool for data processing or automation — it is a medium for creative production and aesthetic judgment. “Beyond Data-Driven Aesthetics,” curated by MIT Architecture alumnus and researcher Alexandros Haridis and on view at the MIT Keller Gallery through June 30, explores the rich, often overlooked history of how 20th- and 21st-century thinkers have transformed computation into a vehicle for artistic and architectural expression. The exhibition challenges the notion that questions of machine creativity are new, demonstrating that they have been a central, if underappreciated, thread in computer science and design for nearly a century.
What Does “Beyond Data-Driven Aesthetics” Explore?
At its core, the exhibition asks a deceptively simple question: Can a machine participate in aesthetic judgment, and if so, what does that mean for human creativity? Haridis argues that the current public conversation around AI and art — often framed as a sudden disruption — is the latest chapter in a much longer narrative. The 1956 Dartmouth Summer Research Project, widely considered the founding event of artificial intelligence, explicitly identified creation and evaluation processes as one of seven key dimensions of human intelligence that AI research should address. The exhibition uses this historical perspective to ground its investigation, moving beyond the hype surrounding tools like ChatGPT and Stable Diffusion to examine the underlying philosophical and computational frameworks that have always been at stake.
The work draws on philosophy, mathematics, computer science, and design computation. It translates dense algorithmic theories and machine-learning systems into physical installations and interactive visualizations, making abstract concepts tangible for a gallery audience. The exhibition is organized around five thematic areas: Aesthetic Measure, Aesthetic Guidelines, Algorithmic Aesthetics, Aesthetic Appropriation, and Aesthetic Novelty. Each theme serves as a window into a distinct computational approach to beauty and evaluation, drawn from specific academic publications. For instance, “Aesthetic Measure” examines mathematician George Birkhoff’s 1930s attempt to quantify aesthetic value mathematically, while “Aesthetic Novelty” investigates how the machine-learning system AICAN judges generated images based on a cognitive theory that balances familiarity with deviation from known artistic styles.
How to Translate Computational Research Into a Physical Exhibition
Translating abstract research into a compelling spatial experience is no small feat. Haridis’s methodology hinges on a key insight: design itself can function as a method of interpretative translation. The process begins by identifying the most salient idea within a given research paper or book and then using design techniques — software reconstruction, physical fabrication, and data visualization — to render that idea in a visual, spatial, and experiential format. The goal is to make visible and tangible what traditional academic scholarship typically communicates only through text, scientific diagrams, and mathematical formulas.
This approach turns the gallery into an active research platform. Visitors do not simply observe finished artifacts; they engage with processes of computational thinking. The five thematic areas of the exhibition are designed as distinct “windows” into different eras and methodologies of computational aesthetics. In “Aesthetic Guidelines,” for example, rule-based systems from design computation and shape grammars are translated into physical models, demonstrating how human insight can be encoded and explored through computation without relying on large datasets. This contrasts sharply with “Aesthetic Novelty,” which visualizes how a generative adversarial network (GAN) navigates the tension between reproducing known styles and generating genuinely novel outputs. The exhibition makes the “black box” of machine learning more interpretable, allowing visitors to see not just what the system produces, but the logic — or aesthetic theory — driving its judgments.
Why This Matters for the Future of AI and Design
“Beyond Data-Driven Aesthetics” arrives at a critical moment. As generative AI becomes ubiquitous in creative industries, there is a growing urgency to understand not just what these systems can do, but how they reason about quality, value, and taste. The exhibition argues that many of the questions currently being asked about AI and creativity — Can a machine be creative? How do we evaluate AI-generated art? What role does human judgment play? — have been approached through a range of computational and theoretical models since at least the early 20th century. By surfacing this history, Haridis provides a vital corrective to the narrative that AI represents a complete break from the past.
One of the most forward-looking aspects of the work is its focus on computational evaluation beyond purely performative or functional requirements. This is a question that resonates deeply across architecture, engineering, and product design: how do we assess the aesthetic or experiential quality of a space or object, not just its structural integrity or efficiency? The exhibition’s case studies suggest that computation can inform us about what contributes positively to human experience, a finding with profound implications for how we design the built environment.
Haridis is also exploring how these ideas can move into broader applications related to the spaces and objects people inhabit daily. The goal is to equip designers and engineers with a better understanding of how rule-based and data-driven computation can inform a more human-centered approach to design — one that values experience as much as performance. This shifts the conversation from a narrow focus on mechanizing “beauty” or “taste” toward a more nuanced investigation of how computational systems can participate in a dialogue with human creators, enriching the process rather than replacing it.
What This Means for Researchers and Practitioners
The exhibition underscores a powerful methodological point: traditional forms of research scholarship may evolve through spatial, visual, and public-facing formats. Software reconstruction, visualization, and physical making are not just tools for presentation; they are themselves forms of inquiry that can reveal new insights about opaque computational systems. For researchers in AI, architecture, and design, this suggests new pathways for communicating complex ideas to both academic and public audiences. For practitioners, the exhibition offers a tangible demonstration that the history of computational aesthetics is a resource, not a footnote. Familiarity with the theories of Birkhoff, the rule-based systems of the 1970s, or the cognitive principles behind AICAN can provide a richer vocabulary for evaluating and directing the use of AI tools in creative work today.
For the tech-savvy reader, the key takeaway is that the current moment of generative AI is an opportunity for a deeper, more historically informed conversation about creativity and computation. The questions being asked are not new, but the tools are more powerful than ever. Visitors to the Keller Gallery, or those who follow the exhibition’s documentation, can see for themselves how algorithmic ideas become tangible stories in space — and perhaps gain a new perspective on the machines that are increasingly shaping our aesthetic landscape.
