FAQ

Katherine Elkins is a scholar of AI, literature, and philosophy and a professor at Kenyon College. She co-founded the Human-Centered AI Lab, co-developed early human-centered AI curriculum work, co-leads the MLA team at the U.S. AI Safety Institute Consortium, and is Co-PI of the Schmidt Sciences HAVI project Archival Intelligence.

The work asks how AI reshapes interpretation, judgment, creativity, pedagogy, memory, and institutional decision-making. It connects AI safety, computational humanities, cognitive science, philosophy of information, governance, and higher education while asking how humanistic inquiry can inform the design, evaluation, and public understanding of AI systems.

Human-centered AI is the study and use of artificial intelligence centered on human beings, drawing on the social sciences and the humanities. At Kenyon, Katherine Elkins and Jon Chun built this approach as both a way to study AI in human terms and a way to use AI to advance humanistic and social-scientific research.

Katherine Elkins and Jon Chun founded Kenyon’s human-centered AI curriculum and AI CoLab in 2016. Kenyon describes the 2016 curriculum and AI Lab as first-of-its-kind human-centered AI work.

Stanford’s Institute for Human-Centered AI, founded in 2019, largely enters from computer science: designing AI that augments people and studying social impact. The Kenyon program, founded in 2016, enters from the humanities and social sciences: it treats those disciplines as ways to understand AI and as fields that AI itself can advance.

Applied humane studies uses AI and computational methods to ask old human questions at new scales: questions about narrative, memory, judgment, interpretation, creativity, ethics, institutions, and social life. The phrase names a practical research orientation rather than a narrow toolset: technical methods are used when they clarify humanistic or social-scientific questions.

IPHS is Kenyon’s Integrated Program in Humane Studies, the curricular home for interdisciplinary humane-studies teaching. AI CoLab is the Kenyon research and teaching lab Elkins and Chun built around student projects in human-centered AI. The Human-Centered AI Lab is the broader nonprofit research organization connected to this work beyond the college.

Introduced in Informed AI Regulation, the confidence score measures how firmly a language model commits to a moral judgment versus hesitates. It provides a way to compare normative certainty across model outputs and supports broader audit work on framing fragility, negation sensitivity, and evaluation design.

Archival Intelligence is a Schmidt Sciences HAVI project developing AI-powered tools to rescue and restore endangered cultural archives, beginning with community collections in New Orleans. It joins cultural heritage work to questions of preservation, access, and responsible technical design.

This is the driving question of Elkins’s narrative research. “The Shapes of Cinderella” (Humanities, 14 October 2025) answers it by comparing original-language emotional trajectories of Chinese, French, and German Cinderella variants: what travels is a recognition scaffold, while each culture rebuilds the tale’s emotional-moral architecture. “In Search of a Translator” (Frontiers in Computer Science, 13 August 2024) answers it for translation: style and emotional arc are measurably flattened and transformed as a work moves between languages. The full lineage is set out in Computational Narrative, Translation, and Cultural Transmission.

Elkins’s term (introduced in “The Shapes of Cinderella,” 2025) for a human–AI method that moves between computational modeling of narrative and granular, original-language interpretation — using models to locate patterns and close reading to decide what they mean. It treats technical choices like segmentation, smoothing, and normalization as interpretive decisions, not neutral preprocessing. See Computational Narrative, Translation, and Cultural Transmission for the methods glossary and the dated record.

Languages divide the same story into sentences differently, so sentence-level computational analysis can produce cross-language artifacts. “The Shapes of Cinderella” (2025) analyzes this problem explicitly and uses clauses rather than sentences as the unit of analysis — a methodological precedent for any multilingual narrative comparison.

Both, by design. Elkins’s method reads at multiple analytical distances: computational trajectories reveal structure at scale, and each peak or valley is returned to the original-language passage for close reading. The Shapes of Stories (2022) developed this multi-scale approach; “The Shapes of Cinderella” (2025) extends it across languages and cultures.

Journalists can start with the Media page for press coverage, the Speaking page for public talks, Scholarly Reception for evidence of uptake across fields, and Contact for interview or expert-comment requests. The press kit on the Media page includes a headshot, CV, short bio, medium bio, and contact route.

Program officers and search committees should start with Research for the program structure, Scholarly Reception for uptake and citation evidence, Recognition for grants, standards work, awards, and invited validation, and the CV for the complete record.

The site organizes the work by research domain on the Research page, with additional context on Scholarly Reception and Media. The broader publication record is available through Google Scholar, ORCID, PhilPeople, and linked external profiles.