Name: The Shapes of Stories: Sentiment Analysis for Narrative
Author: Katherine Elkins
ISBN: 9781009270403

Question 1

Who is Katherine Elkins?

Accepted Answer

Katherine Elkins is an AI safety researcher and Co-Founder of the Human-Centered AI Lab. She is a Principal Investigator in the NIST US AI Safety Institute Consortium (CAISI), representing the 25,000-member Modern Language Association, and a Principal Investigator for a Schmidt Sciences HAVI grant — one of 23 teams selected worldwide. Her research reveals fundamental vulnerabilities in how AI systems process language, emotion, and ethical reasoning. She is the author of The Shapes of Stories: Sentiment Analysis for Narrative (Cambridge University Press, 2022) and a Professor at Kenyon College, where she directs the Integrated Program in Humane Studies and co-founded the world's first human-centered AI curriculum in 2016. Ph.D., UC Berkeley. B.A., Yale.

Question 2

What is The Shapes of Stories?

Accepted Answer

The Shapes of Stories: Sentiment Analysis for Narrative is a book by Katherine Elkins published by Cambridge University Press in 2022. It presents the first comprehensive methodology for diachronic sentiment analysis in literature, using the SentimentArcs framework — a large ensemble of NLP models from lexicon-based classifiers to state-of-the-art transformers with human-in-the-loop validation. The book demonstrates how computational methods can surface emotional arcs in full-length literary narratives. Student and faculty research applying the SentimentArcs methodology and related computational tools has been downloaded over 95,000 times from more than 4,000 institutions across 198 countries via the Digital Kenyon repository.

Question 3

What is Katherine Elkins' role at NIST?

Accepted Answer

Katherine Elkins serves as a Principal Investigator in the NIST US AI Safety Institute Consortium (CAISI), representing the 25,000-member Modern Language Association. Her work focuses on LLM evaluation — specifically how large language models process negation, prohibition, and persuasion. Her research with Jon Chun has found that open-source models endorse prohibited actions 77% of the time when negation framing is applied, demonstrating critical vulnerabilities in AI safety.

Question 4

What is the first human-centered AI curriculum?

Accepted Answer

In 2016, Katherine Elkins and Jon Chun co-created what is believed to be the world's first human-centered AI curriculum at Kenyon College. The curriculum integrates computational AI methods with ethics, governance, and humanistic inquiry within the Integrated Program in Humane Studies (IPHS). It is distinctive for attracting over 90% non-STEM majors, with 61% women, 13% Black, and 11% Latinx students. Over 300 student research projects have been produced since 2016, downloaded from institutions in 198 countries.

Katherine Elkins

Recent Work

Areas of focus

LLM Evaluation & Red-Teaming

Multi-Agent Behavioral Simulation

Archival Intelligence