Katherine Elkins, AI safety researcher and author

NIST CAISI · Schmidt Sciences HAVI · Cambridge University Press · Kenyon AI CoLab

Katherine Elkins

AI safety cannot be solved by engineering alone. Katherine Elkins works on the parts that require culture, narrative, judgment, memory, and institutions.

She co-leads the team representing the 25,000-member Modern Language Association at NIST CAISI, is Co-PI of Archival Intelligence — a Schmidt Sciences HAVI project rescuing endangered cultural archives — and is the author of The Shapes of Stories (Cambridge University Press, 2022). With Jon Chun, she co-founded the world's first human-centered AI curriculum and lab at Kenyon College in 2016.

Google Scholar CV Wikipedia

NIST CAISI

Co-lead, MLA team — LLM evaluation for U.S. AI standards

Schmidt Sciences HAVI

Co-PI, Archival Intelligence — 1 of 23 teams from 600+ applications

Cambridge UP

The Shapes of Stories — sentiment analysis for narrative

Kenyon, 2016

World's first human-centered AI curriculum & lab

Featured uptake

One of three canonical frameworks

1 of 3 FAccT 2025

When FAccT 2025 researchers mapped how the field defines AI “openness,” they named the ICML 2024 position paper Elkins co-authored — an oral presentation, top 2% of submissions — one of three canonical openness frameworks. And in Poetics Today, narratologist James Phelan took up her essay “AI Comes for the Author” in an instructive debate on what language models mean for authorship. The same questions, argued in machine learning and in literary studies. Trace the reception →

Firsts

Work done early

Dates establish the record; the Reception page documents what followed.

2016

Co-founded the world's first human-centered AI curriculum and lab at Kenyon College, with Jon Chun.

The field's first peer-reviewed account followed in 2023.

2020

The first writer's Turing test of a large language model — “Can GPT-3 Pass a Writer's Turing Test?”

Cited by Floridi and Chiriatti in Minds and Machines; now a named framework in philosophy of AI and NLP evaluation.

2022

The Shapes of Stories, Cambridge University Press — among the first methodologies for sentiment analysis of narrative.

Adopted as evaluation infrastructure (NarraBench, EACL 2026) and applied across 25,000+ story retellings (Scientific Reports).

2024

The first ethics-based audit of moral reasoning in deployed LLMs; PMLA on the university after LLMs; ICML oral on open-source generative AI; CAISI standards work begins.

2025

Co-PI, Schmidt Sciences HAVI — Archival Intelligence; Public AI; OpenAI Higher Education Forum.

Current work

NIST CAISI

Co-leads the team representing the Modern Language Association at the NIST AI Safety Institute Consortium — the only humanities-led team in the federal consortium.

Archival Intelligence

Co-PI of a Schmidt Sciences HAVI project building open AI tools to rescue endangered cultural archives, beginning in New Orleans.

Narrative Intelligence

From The Shapes of Stories to current work on what emotional arcs reveal that plot and character analysis miss.

Selected writing

Recent essays and books

“AI Comes for the Author” — Poetics Today 45.2, 2024.

What happens to authorship when the machine writes well — the essay James Phelan took up in an instructive debate.

Read at Duke UP →

“A(I) University in Ruins” — PMLA 139.3, 2024.

What remains of the university in a world with large language models.

Read at Cambridge Core →

The Shapes of Stories — Cambridge University Press, 2022.

Sentiment analysis for narrative — what emotional arcs reveal that plot and character analysis miss.

All books →

In others' words

Cited, engaged, debated

“An ‘objective’ model, according to Katherine Elkins, cannot exist.”

— Cao et al., arXiv (2025), anchoring a central claim on the PMLA essay

“Debating the system of values we wish these tools to align with is the first step.”

— Tanya Klowden and Terence Tao (Fields Medalist), IJHAC (2023), engaging Elkins and Chun's curriculum work

For the empirical claim that “GPT-3 writes better than many people,” Floridi and Chiriatti (Minds and Machines, 2020) cite the writer's Turing test. Full scholarly reception →

Start here

Find your way in

Journalists

FAQ & Media

Definitions, quotable context, and where she has been quoted.

Scholars & program officers

Research & Reception

Six connected research areas, with named citers across AI safety, NLP, and the humanities.

Students

Mentored Research

Original undergraduate research in the AI CoLab, open to every division of the liberal arts.

Event organizers

Speaking

Keynotes and public lectures on AI safety, narrative, creativity, and the university.

Overview

At a glance

25,000

members of the Modern Language Association represented at NIST CAISI

1 of 23

Schmidt Sciences HAVI teams selected worldwide, from 600+ applications

Top 2%

ICML 2024 oral presentation, open-source generative AI

1 of 3

canonical AI-openness frameworks, per FAccT 2025

6

connected research areas, from ethics audits to Proust

2016

the year the first human-centered AI curriculum launched at Kenyon

FAQ

Common questions

What is human-centered AI?

An approach co-founded at Kenyon in 2016 that uses frontier AI and real engineering to make the oldest human questions — meaning, value, memory, story — quantifiable and testable, while keeping culture, judgment, and institutions at the center of AI safety. It is distinct from human-centered UI/UX design and from purely critical AI ethics.

What is the confidence-scoring method for auditing language models?

Introduced in Informed AI Regulation, it measures how firmly a model commits to a moral judgment versus hesitates — a way to compare normative certainty across models. It has since been applied across 1,613 social decision-making scenarios (COLING 2025) and included among 69 foundational works in the AAAI 2026 “Beyond Verdicts” survey.

Where should journalists or program officers start?

Journalists: the FAQ and Media pages, then the contact form for quotes. Program officers and search committees: the Research page for the six-area map, then Scholarly Reception for named citers and detailed uptake evidence.

All frequently asked questions →

Contact page →

Professor, Kenyon College · Co-Founder, Human-Centered AI Lab · Co-lead, MLA team at NIST CAISI · Co-PI, Schmidt Sciences HAVI Archival Intelligence