A new exhibition at the MIT Keller Gallery is redefining the conversation around artificial intelligence and aesthetics, challenging the notion that AI-driven creativity is a product of the 21st century. Beyond Data-Driven Aesthetics, curated by Alexandros Haridis, a researcher and MIT Architecture alumnus, explores how computational methods have influenced aesthetic judgment in architecture and applied arts for nearly a century. The show, open through June 30, bridges philosophy, mathematics, and design to transform abstract algorithms into tangible, interactive experiences.
From 1956 to 2024: Why AI’s aesthetic questions aren’t new
Haridis traces the roots of today’s debates about AI and creativity to the Dartmouth Summer Research Project of 1956, a landmark event that identified aesthetic judgment as a core challenge for artificial intelligence. While modern systems like ChatGPT and Stable Diffusion dominate headlines, their underlying questions—such as how machines evaluate beauty or originality—have deep historical roots in scholarly discourse.
During his PhD research at MIT, Haridis observed how rapidly machine learning tools entered mainstream conversations about design and art. Yet his work revealed that many of these discussions echoed earlier debates from the 20th century, particularly in fields like design computation and shape grammars. These approaches emphasize rule-based systems over purely data-driven learning, offering a counterpoint to today’s dominant AI models.
The exhibition also draws from philosophical and literary sources, including the writings of Samuel Taylor Coleridge and Oscar Wilde, to examine how theories of aesthetic value might inform—or limit—contemporary AI systems. The goal isn’t to reject data-driven methods but to contextualize them within a broader history of computational aesthetics.
Turning dense research into immersive experiences
How does one translate a 1930s mathematical theory of beauty into a gallery setting? Haridis’s exhibition takes a design-first approach, using physical installations, digital visualizations, and interactive elements to make abstract concepts accessible. The show is structured around five thematic areas:
- Aesthetic Measure: Explores mathematician George Birkhoff’s attempt to quantify aesthetic value mathematically.
- Aesthetic Guidelines: Focuses on rule-based design systems that balance human insight with computational rules.
- Algorithmic Aesthetics: Investigates how algorithms interpret and generate artistic styles.
- Aesthetic Appropriation: Examines how computational systems borrow and reinterpret existing artistic traditions.
- Aesthetic Novelty: Looks at how machine learning systems like AICAN evaluate generated images based on familiarity and deviation from known styles.
Each theme acts as a "window" into a specific publication or research paper, using design to translate dense academic language into tangible, experiential forms. The result is a space where visitors can engage with computational aesthetics not just intellectually, but sensorially.
What’s next for computational aesthetics?
Haridis sees Beyond Data-Driven Aesthetics as both an exhibition and an ongoing research platform. One key question he plans to explore further is how computational systems can evaluate designs beyond mere functionality—whether for buildings, products, or urban spaces. The exhibition’s case studies suggest that this inquiry isn’t new; it’s part of a century-long conversation about how computation intersects with human experience.
A particularly promising direction is applying these ideas to the built environment. Haridis is interested in how rule-based and data-driven systems can help designers and engineers create spaces that enhance human well-being. The exhibition hints at a future where computation isn’t just a tool for efficiency but a framework for understanding what makes environments aesthetically and emotionally resonant.
The project also underscores the evolving role of design itself. As Haridis puts it, design can serve as a bridge between technical research and human-centered applications. By making abstract computational theories tangible, exhibitions like this one pave the way for more inclusive and thoughtful conversations about AI, creativity, and the spaces we inhabit.
AI summary
MIT’in yeni sergisi, hesaplamanın mimarlık ve tasarımda estetik yargılara nasıl malzeme olduğunu araştırıyor. Alexandros Haridis’in küratörlüğündeki proje, algoritmalar ve makine öğrenimi sistemlerini fiziksel enstalasyonlara dönüştürüyor.
