Research
How AI systems reason
Katherine Elkins studies how increasingly capable AI systems reason, persuade, interpret, coordinate, and interact with people in contexts where judgment, language, history, culture, and institutions matter.
Her current research spans model behavior, human judgment, agentic and multi-agent systems, narrative, behavioral evaluation, culturally and historically grounded AI, and governance based on empirical research.
Six research areas
Behavior in complex human settings
The common problem is how models reason, persuade, interpret, coordinate, and fail when language, judgment, incentives, culture, and institutions shape the task.
-
Model behavior & human judgment
Reasoning, ethical judgment, decision-making, persuasion, deception, manipulation, calibration, and linguistic failure modes. This work tests how behavior changes under syntactic reframing, ambiguity, social pressure, and shifts in context, and whether a model’s explanation tracks what it actually does.
Model behavior and evaluation research -
Agents & multi-agent systems
Coordination, debate, explanation, adversarial interaction, emergent behavior, and evaluation of agents. Current mentored and collaborative work examines deception in multi-agent negotiation, emotional stability among interacting agents, and the use of agent debate in consequential decisions.
Agentic and multi-agent research -
Narrative, cognition & emotion
Narrative is a cognitive and emotional tool: a structure for memory and interpretation, a mechanism of persuasion, and a factor in how models and people organize judgment. Current work connects literary study, cognitive science, and philosophy of information to model responses, behavioral effects, human-AI interaction, and vulnerabilities to persuasion.
Narrative and emotion research -
Evaluation & robustness
Behavioral evaluation, benchmark design, model comparison, confidence and calibration, red-teaming, ethics-based auditing, and failure-mode analysis. Elkins and Chun conducted the first ethics-based audit of moral reasoning in deployed large language models; current work continues to examine vulnerabilities conventional performance metrics can miss.
AI evaluation and robustness -
Culturally & historically grounded AI
AI systems that work with archives, cultural memory, provenance, retrieval, context, and expert interpretation. The current anchor for this work is Archival Intelligence.
Archival Intelligence spotlight -
AI, institutions & governance
Standards, comparative regulation, open-model governance, public-interest AI, and institutional evaluation grounded in evidence about model behavior. Elkins co-leads the MLA team in the NIST CAISI AI Consortium and coauthored the first systematic EU–China–US regulatory comparison after passage of the EU AI Act.
Governance and public AI
Research, education & curriculum
Students as AI researchers
What should universities teach when students can increasingly delegate traditional academic tasks to machines?
Elkins’s answer moves students upstream: from users of AI systems to researchers who evaluate, build, interpret, and question them. Since 2016, the Human-Centered AI curriculum and lab have joined technical training with domain expertise and original student research.
AI in Higher Education: curriculum, lab, and archiveCurrent research spotlight
Archival Intelligence
Rescuing endangered cultural archives.
Elkins is Co-PI of Archival Intelligence, a Schmidt Sciences Humanities and AI Virtual Institute project building open AI tools for endangered archives, beginning in New Orleans.
The research asks how AI systems can retrieve, represent, and interpret cultural and historical materials without stripping away provenance, context, ambiguity, or expert knowledge. Generic retrieval can find a record while missing why it matters, who is authorized to interpret it, or how place and community shape its meaning. Those are system-design problems.
Related work addresses cultural data governance and public knowledge infrastructure, including participation in UNESCO’s AI, IP & Culture Repository co-design process.
Explore Archival IntelligenceMethodological foundations
Making difficult phenomena testable
How can context, ambiguity, interpretation, and historical meaning remain present in computational research? The program began with literature and philosophy; its methods now inform research on models, agents, and institutions.
The Shapes of Stories
Elkins and Chun introduced the first rigorous methodology for narrative sentiment analysis, establishing a reproducible framework for selecting, comparing, validating, and interpreting sentiment models across narrative texts.
Ensembles and interpretation
Developed by Jon Chun, SentimentArcs is a novel ensemble method for comparing narrative sentiment trajectories across texts. Middle Reading connects those computational comparisons to interpretation at the scale of passages, scenes, and whole works. Jon Chun’s technical work →
What a representation preserves
The program includes the first application of explainable AI to narrative analysis, computational comparison of translations, and the first multimodal method for measuring long-form film sentiment-arc coherence across modalities.
Can GPT-3 Pass a Writer’s Turing Test?
Coauthored by Elkins and Chun, the first writer’s Turing test of a large language model made literary imitation and reasoning experimentally comparable. The same concerns now shape behavioral evaluation, persuasion research, agent analysis, and culturally grounded AI.
Selected work
Publications and projects
- Model behavior & evaluation · 2024Informed AI RegulationEthics-based auditing of moral reasoning across deployed language models.
- Open models & governance · ICML 2024 oralNear to Mid-term Risks and Opportunities of Open-Source Generative AIA benefit–risk framework for open-source generative AI.
- Narrative method · Cambridge, 2022The Shapes of StoriesRigorous methodology and ensemble comparison for narrative sentiment analysis.
- Model writing & reasoning · 2020Can GPT-3 Pass a Writer’s Turing Test?An early behavioral test of literary imitation, authorship, and higher-order reasoning.
- Cultural and historical AI · Current projectArchival IntelligenceOpen AI systems for endangered cultural archives, provenance, retrieval, and interpretation.
Evidence and uptake
Scholarly reception
The citation record now spans 15 fields and 55 subfields, including AI, digital humanities, literary studies, education, medicine, law, and related fields.
Explore scholarly receptionExplore further
Related work
AI, creativity, and authorship → Philosophy of mind → Narrative, translation, and transmission → The Shapes of Stories →