Our research spans AI safety evaluation, governance and regulation, computational social science, cultural heritage preservation, and affective computing — always grounding technical methods in humanistic depth.
Syntactic framing vulnerabilities, threat modeling for multi-agent architectures, and evaluation frameworks for autonomous AI systems. Our co-founders' work on how language models process negation, prohibition, and persuasion informs safety evaluation through their role as Principal Investigators representing the Modern Language Association in the consortium of NIST's Center for AI Standards and Innovation (CAISI).
With the Turing test largely behind us, the question is no longer whether machines can pass for human but what kind of minds we have built. A handful of AI systems now answer moral and existential questions for hundreds of millions of people, while the religious and philosophical traditions beneath their answers stay hidden from benchmark scores. The Wisdom Test is our successor to the Turing test: an evaluation that surfaces the wisdom traditions beneath the leading chatbots — which are reproduced and which quietly disappear. It probes the places traditions most divide (the self, obligation, the sacred, suffering, fate, and more) and asks concretely whether different training regimes give American and Chinese frontier models different stances, and whose inheritance each carries. Four of eight axes have been piloted through our cross-cultural ethical audit methodology, the first comprehensive discourse-level ethical audit of frontier LLMs across traditions. Planned deliverables include the test itself, an open multilingual archive, and evaluation methods designed for adoption beyond current benchmarks.
Comparative global AI regulation (EU, China, US). Open-source AI policy analysis. Behavioral prediction and ethical auditing of LLM decision-making systems. Co-authored policy paper with International Public AI. This work also includes cultural data governance, including provenance infrastructure and rights-aware AI training practices developed through the AI, IP & Culture Repository co-design process.
Multi-agent simulation of high-stakes human decisions including judicial recidivism prediction. Over 90 model/reasoning combinations benchmarked. Funded through Notre Dame–IBM Tech Ethics Lab.
Rescuing endangered New Orleans heritage archives using AI. Community-governed data sovereignty for historically marginalized populations. One of 23 teams selected worldwide for the Schmidt Sciences Humanities and AI Virtual Institute (HAVI) program.
Open-source methodology (code) for diachronic sentiment analysis in text and film. Created the first computational methodology for surfacing emotional arc in full-length literary narratives. Student research using the methodology has been downloaded 100,000+ times from institutions in 198 countries.
Our 2023 paper "The Crisis of Artificial Intelligence: A New Digital Humanities Curriculum for Human-Centered AI" (International Journal of Humanities and Arts Computing) established the intellectual framework for integrating computational methods with ethics, governance, and humanistic inquiry — and the evidence base for why this integration matters for AI safety and public benefit. (higher education)