Schmidt Sciences HAVI · 2025 Cohort
Rescuing New Orleans' Endangered Cultural Legacy
Archival Intelligence is building an open, end-to-end AI pipeline for endangered cultural archives. The project connects the capture and recovery of fragile sources with recognition, description, search, analysis, and responsible access.
01 The project
Our initial research focuses on the cultural record of New Orleans and the emergence of jazz: newspapers, correspondence, photographs, sheet music, and recorded sound that pose different problems for existing models. This multilingual and multimodal archive provides a demanding test case for methods intended to serve small and under-resourced institutions more broadly.
Louisiana French is part of that end-to-end research. In collaboration with Nous Foundation, we are examining how existing models perform on written and recorded materials shaped by regional vocabulary, historical spelling, language variation, and code-switching—and how that performance might be improved without treating meaningful linguistic variation as error.
Led by Co-Principal Investigators Katherine Elkins and Jon Chun, Archival Intelligence is supported by Schmidt Sciences’ Humanities and AI Virtual Institute. The research brings together collaborators at Kenyon College, Berklee College of Music, Louisiana State University, and other institutions, with the Human-Centered AI Lab serving as its broader research home.
Uneven printing and aging paper affect recovery; dense layouts and historical vocabulary complicate recognition; and provenance, language, and relationships shape description and interpretation. The research concerns how these problems interact across the archival process.
End-to-end archival pipeline · Multimodal AI · Louisiana French · Responsible access
02 The pipeline
Archival Intelligence is building an open AI pipeline that connects the recovery of fragile sources with recognition, description, search, analysis, and responsible access. Its initial research focuses on the multilingual and multimodal cultural record of New Orleans.
Archival technologies often address digitization, transcription, cataloging, and search as separate tasks. Archival Intelligence investigates how they can operate as a connected process across different media, languages, and states of preservation.
Newspapers, sheet music, correspondence, photographs, and recordings carry different forms of damage and different kinds of evidence. The project evaluates and improves existing models while keeping their outputs connected to provenance, language, rights, historical context, and human interpretation.
The project studies how tasks often addressed in isolation can operate as a connected process, from the recovery of a source through content-level discovery and responsible access.
The project examines where current systems succeed, where their performance deteriorates, and what kinds of improvement are required for archival use. Evaluation includes materials whose physical condition, layout, medium, or language falls outside the assumptions of conventional digitization.
03 Stewardship
Improved recognition and search can make archival materials substantially more accessible, but access is neither uniform nor automatic. Ownership, attribution, cultural sensitivity, and intended use differ across collections. The pipeline therefore carries provenance, permissions, community knowledge, and human review alongside computational outputs. Archival Intelligence is a signatory to the Open Heritage Statement.
Research with Nous Foundation begins with Louisiana French in written and recorded archival materials. Linguistic, historical, and community expertise informs questions of recognition, representation, attribution, access, and appropriate use across the pipeline.
Archival Intelligence has signed onto the Open Heritage framework, aligning the project with principles of access, cultural sensitivity, community participation, stewardship, and data sovereignty.
Language affects every stage of archival processing, from recognition and description to search and interpretation. Louisiana French materials may contain regional vocabulary, historical spelling, variation among speakers and communities, and code-switching across written and recorded sources.
In collaboration with Nous Foundation in New Orleans, Archival Intelligence is beginning research into how existing models perform on these materials and how that performance might be improved. Linguistic, historical, and community expertise is necessary to distinguish model error from meaningful variation.
Louisiana French therefore serves as an early test of the pipeline as a whole. Linguistic features affect recognition, archival description, connections among materials, search results, interpretation, attribution, and access.
Research use
date · creator · place · language · collection · provenance · rights · historical context · relationships · uncertainty
Archival Intelligence is an independent research initiative supported by Schmidt Sciences’ Humanities and AI Virtual Institute, with research conducted in collaboration with Kenyon College, Berklee College of Music, Louisiana State University, and other partners.
This approach grows out of broader work at the Human-Centered AI Lab on how AI systems interact with language, culture, judgment, and human values.
The stages below describe the research problem at a functional level. They do not assume that every source follows an identical path: the requirements of a newspaper, manuscript, photograph, musical score, or recording differ.
documents · images · audio
legibility · layout · signal quality
print · handwriting · speech · music
metadata · provenance · language · rights
people · places · works · collections
content-level discovery · cultural research
human judgment · permissions · appropriate use
04 People
Archival Intelligence brings together work in artificial intelligence, archival studies, cultural history, music, language, museum systems, and human-centered design.
Leadership
Co-Principal Investigator
Katherine Elkins is a professor at Kenyon College and co-founder of the Human-Centered AI Lab. Her research examines artificial intelligence, narrative, cultural preservation, judgment, and human values.
Co-Principal Investigator
Jon Chun is co-founder and director of the Human-Centered AI Lab. His research focuses on AI systems, computational methods, technical infrastructure, and research design.
Co-Investigator · Louisiana State University
Supratik Mukhopadhyay contributes research in multimodal machine learning, pattern recognition, and scalable systems for heterogeneous archival materials.
Co-Investigator · Berklee College of Music
Lori Landay contributes research on digital media, museum interfaces, and the representation of musical and cultural materials.
Consultant · Columbia University
Brent Hayes Edwards advises the project on jazz history, African American cultural production, translation, and archival theory.
Current research
Researches archival collections, their historical contexts and provenance, and the relationships among datasets and metadata.
Evaluates collection-management systems, including metadata and taxonomy structures, permissions, audit trails, provenance, search, export, and interface workflows.
Project history
Surveyed candidate collections and researched historical context, provenance, metadata, and possible relationships among datasets.
Developed benchmarking research for degraded documents and newspapers, including difficult layouts, scan quality, typography, calibration, sample selection, performance, and cost.
Investigated language preservation and orthographic variation in oral-history materials from under-resourced languages.
Contributed to OCR benchmarking on difficult historical documents and newspapers.
Project contributors
Coordinates project infrastructure, research meetings, team connections, and institutional and community relationships, including work with Nous Foundation.
Contributes expertise in digital collections, metadata, provenance, digitization, archival infrastructure, and data stewardship.
Contributes research on museum workflows, digital collections, metadata, interfaces, interpretation, discovery, and public access.
Contributes expertise in museum studies, collections, material culture, digitization, interpretation, and user experience.
Institutional relationships
Researchers at these institutions contribute work in the humanities, computing, media, and jazz studies.
The project’s New Orleans relationships connect technical research with language, cultural knowledge, collections, stewardship, and access.
The project is supported through the Humanities and AI Virtual Institute and is a signatory to the Open Heritage Statement.
05 Chronology
Project updates, collaborations, events, publications, media coverage, and related research.
A public conversation at Kenyon College on preservation choices, the limits of AI-assisted reconstruction, and the governance of archival access and interpretation.
Team member Abigail Foster will volunteer at Trust in the Loop, the 2026 conference of Artificial Intelligence for Libraries, Archives & Museums. The Washington, DC conference is co-hosted by the Library of Congress, National Gallery of Art, and Smithsonian Institution.
Archival Intelligence and Nous Foundation Inc., doing business as the New Orleans Foundation for Francophone Cultures, signed a non-binding memorandum of understanding establishing Nous as a Community and Cultural Partner.
The initial research focus is Louisiana French in written and recorded archival materials. The collaboration brings linguistic, historical, and community expertise to the evaluation and improvement of existing models, with attention to regional vocabulary, historical spelling, language variation, code-switching, attribution, access, and appropriate use.
Materials of interest include newspapers, correspondence, sheet music, photographs, recordings, and oral histories connected to Louisiana French language and culture.
The summer research program examined candidate archival collections, historical context and provenance, OCR performance on difficult newspapers and documents, museum and archive systems, metadata and institutional workflows, language variation, and interface requirements.
Jessica Daugherty and Ty Kushi conducted dataset research. Ayesha Aslam and Ashwin Krishnan contributed to OCR benchmarking. Adrian Mangine reviewed collection-management systems. Abigail Foster, Cat Ashley Fairchild Perloff, and Jenna Nolt contributed research on museum practice, digitization, metadata, stewardship, and adoption. Nic Pechon-Elkins investigated language preservation and orthographic variation, while Lori Landay developed interface research and early design studies. Anne-Duncan Enright coordinated the program and initiated discussions with Nous Foundation.
The work identified limitations in existing recognition, search, interoperability, export, and archival-data systems and informed the next phase of pipeline research. Further information appears on the Team page.
Archival Intelligence became a signatory to the Open Heritage Statement, an international initiative developed through the Creative Commons-convened Open Heritage Coalition.
The Statement connects public access with stewardship, community participation, cultural sensitivity, and data sovereignty. These principles inform the project's treatment of provenance, attribution, permissions, and reuse.
Katherine Elkins contributes a chapter to the forthcoming UNESCO volume Artificial Intelligence and Culture: A Global Perspective from Cairo. It examines generative AI and cultural production and includes early Archival Intelligence research on poorly digitized New Orleans materials.
Katherine Elkins and Jon Chun co-lead the Modern Language Association team participating in NIST’s Center for AI Standards and Innovation. Their work applies humanistic methods to model evaluation, with particular attention to cases in which linguistic framing changes model behavior.
Related public work includes When Prohibitions Become Permissions, a benchmark study of how language models respond to negated instructions. This research is distinct from Archival Intelligence but informs its approach to language variation, model reliability, and human review.
A Kenyon Alumni Magazine feature on the College's human-centered AI program, including the origins and aims of Archival Intelligence.
In this position paper, Katherine Elkins and seven co-authors argue that open-source models require complementary public infrastructure and institutions if they are to be broadly accessible, maintained, and governed in the public interest. The argument provides a wider policy context for Archival Intelligence’s commitment to open tools.
The Ohio Newsroom reports on the project’s plans for an open archival AI system, its New Orleans research materials, and the role of human review in archival interpretation.
Katherine Elkins discusses AI safety research and the role of the humanities in AI governance.
Kenyon College’s announcement of the $330,000 Schmidt Sciences award supporting development of an open AI system for endangered archives in underrepresented communities.
Local coverage of the Schmidt Sciences award and the project’s intended application beyond its initial New Orleans research setting.
A feature on Kenyon College’s human-centered AI program and the interdisciplinary research context from which Archival Intelligence developed.
Katherine Elkins presented computational-humanities research related to the Archival Intelligence methodology at OpenAI’s Education Guild Higher Education Forum.
Schmidt Sciences selected Archival Intelligence as one of 23 Humanities and AI Virtual Institute teams. The 18-month project received a $330,000 award.
The New Orleans Jazz Museum provides the project’s initial institutional and cultural setting for research on heterogeneous archives related to the emergence of jazz.
06 Join us
Talk with us about preservation challenges.
We welcome research and institutional conversations.
Learn about student and collaborative research opportunities.
Photo: Louisiana State Museum.