We evaluate models, compare governance systems, reconstruct damaged archives, and study how machines handle language, culture, and human judgment.
Tests of syntactic fragility, negation, prohibition, persuasion, and reported reasoning. This work contributes language-centered methods to the MLA team’s research in NIST CAISI.
More than 90 model and reasoning combinations tested on high-stakes decisions, including judicial recidivism prediction, with support from the Notre Dame–IBM Technology Ethics Lab.
The question is no longer whether a machine can pass for human, but what kind of mind it presents. The Wisdom Test probes views of selfhood, obligation, the sacred, suffering, and fate to reveal which philosophical and religious traditions frontier models reproduce or omit. Four of eight axes have been piloted; planned outputs include an open multilingual archive and reusable evaluation methods.
Systematic comparison of AI regulation in the European Union, China, and the United States, with work on open models, behavioral evaluation, and institutional accountability.
Rights-aware training, provenance infrastructure, and community control of cultural data, developed through the AI, IP & Culture Repository co-design process.
AI methods to preserve, restore, and analyze endangered multilingual and multimodal archives without stripping away provenance, cultural context, or community authority. Supported by Schmidt Sciences HAVI.
Open methods for tracing emotional patterns across full-length texts and film. Explore code
The 2016 curriculum and 2023 research framework joined computational practice to ethics, governance, and interpretation. Read more