Isaiah
Type
Web Application
Role
Interaction Design · Programming
Tool
Open AI · Dataset
Preview
Concept
Curation is the act of researching, selecting, organizing, and presenting art to create meaning beyond the individual pieces. In the art world, curators aim to create a synergy where the sum of the works offers sociopolitical or art historical interpretation, education, or inspiration for the audience.
Despite its importance, curation often comes with limitations. Decisions are shaped by the personal values, biases, and experiences of curators, which means the narrative of an exhibition is never entirely objective. This subjectivity, while central to the creative process, can also limit inclusivity or overlook certain perspectives. Additionally, curation has an exclusive nature that creates a gap between the general public and art appreciation, often making exhibitions less accessible or relatable to broader audiences.
Isaiah addresses these challenges by exploring the potential of AI to enhance curation. By analyzing extensive datasets of art history, criticism, and cultural context, the project aims to uncover overlooked works, foster inclusivity, and personalize the art experience. Isaiah is not about replacing curators but augmenting their expertise, bridging the gap between curatorial practice and public accessibility through a collaborative model of human creativity and technological capacity.
Research
The idea for Isaiah stemmed from the evolving role of curators and the integration of AI into creative practices. Studies such as the Duke Nasher Museum's AI-curated exhibition and IBM's Pinacoteca de São Paulo project showcased the possibilities and limitations of using AI in art curation. These experiments demonstrated AI's potential to identify patterns and streamline the curatorial process but also highlighted its inability to match the emotional and cultural depth of human curators.
Building on this foundation, I was particularly inspired by Laura Herman's research, The Algorithmic Pedestal, where she used Instagram's algorithm to curate an exhibition. Her study compared AI-curated results with those of a human curator, revealing that the human curator created more cohesive and meaningful narratives. However, Herman's research relied on the same limited dataset for both approaches, which constrained the AI's potential. This led me to explore a different direction: leveraging AI's capacity to process and analyze vast datasets. While human curators are naturally limited by memory and the scope of their research, AI can uncover hidden thematic connections and recall overlooked works from expansive collections.
Isaiah builds on these insights, aiming to bridge the gap between human curation and AI's capabilities. Isaiah explores how AI can complement curatorial expertise by employing methodologies like embeddings and multimodal models, balancing data-driven efficiency with the emotional and cultural nuance necessary to craft compelling and inclusive exhibitions.
Laura Herman, The Algorithmic Pedestal: A Practice-Based Study of Algorithmic and Artistic Curation, Oxford Internet Institute
Prototyping
I began by exploring available art datasets, including the MoMA API, the Metropolitan Museum of Art Collection API, and Google Arts & Culture. I quickly realized that while these datasets were rich in metadata, they lacked the nuanced context needed for meaningful curation. For example, titles and descriptions often didn't capture thematic connections or emotional resonance between works, which made me rethink how to incorporate AI effectively.
To address this, I built multiple prototype versions, mainly using OpenAI's API to process user input and generate keywords for thematic searches. Early versions were overly reliant on simple tag matching, which didn't capture the complexity of curatorial thinking. I refined the workflow by incorporating clustering methods and embeddings, allowing the model to compare the contextual and visual dataset with user input to better understand the connections between artworks and user input. In later iterations, I incorporated curatorial questions to engage users and refine the curation process. These questions, generated with OpenAI's API, helped Isaiah align its selections more closely with the user's intent. I also experimented with a semantic search approach using a richer dataset, Artpedia, which provided detailed descriptions and better contextual information. This enabled Isaiah to create more cohesive thematic exhibitions while offering explanations for its selections.
Prototype early versions workflow development
I went through many iterations to refine curation logic, dataset limitations, and thematic coherence
Outcome
I interviewed curators from respected institutions such as the MET and the Brooklyn Museum, as well as independent curators, to refine Isaiah's approach to curatorial decision-making. Their feedback emphasized the importance of comprehensive datasets in building meaningful connections between artworks, leading to selecting Artpedia and MET resources with detailed descriptions for deeper analysis.
The final iteration of Isaiah utilizes OpenAI API and Transformer.js embeddings to analyze both textual and visual data. User inputs are transformed into thematic keywords, which are used to query enriched datasets and select artworks based on criteria such as relevance, contextualization, narrative arc, diversity and representation, and complementarity. These criteria ensure the exhibition establishes a compelling theme, provides historical and cultural context, creates a cohesive narrative journey, includes diverse perspectives, and avoids redundancy in its selections.
Isaiah's workflow integrates user interaction with advanced dataset analysis, creating a hybrid approach that addresses traditional curation's limitations. The final workflow diagram illustrates how Isaiah leverages comprehensive datasets and AI-driven analysis to present cohesive, personalized exhibitions.
Final version workflow
User interface frame design for the final version exhibition
Exhibition spatial layout design
I remade the final version to show version and exhibited in the IMA gallery. I used projection mapping for displaying the artworks.
Next Steps
I plan to integrate larger and richer datasets to offer more variety and improve the quality of curated exhibitions. Next, I aim to compare Isaiah's results with human curation through user testing to refine its approach. Lastly, I want to add collaborative features, allowing multiple users to co-curate exhibitions in real-time.
Credits
Software / Code / Design
Mickey Oh
Advisors
David Rios
Daniel Shiffman
J. H. Moon
Gottfried Haider