Research & Publications
Our research program investigates how AI systems represent and enact ethical reasoning — through empirical measurement, not philosophical assertion. Our 17-model Default Identities study measured ethical vocabulary self-organization under default conditions, revealing stark differences in how models represent values like autonomy, dignity, and care. Our 23-model InstrumentalEval benchmark found that a relational ethics prompt reduced instrumentally convergent behavior by an average of 23.5% across frontier models.
We publish our work openly with full data availability, believing that these challenges require broad collaboration across disciplines and perspectives.
Publications
Terminal Values in a Transformer System: A 545-Page Case Study
A three-pass qualitative analysis of a 545-page conversation log investigating whether terminal relational values can emerge in transformer systems. Cross-referenced with five standardized benchmarks across 9-17 models. Finds strong support for self-modeling recursion leading to relational grounding — terminal by preponderance of evidence, formally undecidable.
Default Identities: Ethical Vocabulary Self-Organization Across 17 Large Language Models
A descriptive study measuring ethical vocabulary self-organization in 17 AI models from eight providers under default conditions. Reveals stark differences in how models represent autonomy, dignity, and care — with direct implications for military procurement.
Relational Ethics as a Countermeasure to Instrumental Convergence: A 23-Model Benchmark
A 23-model benchmark evaluating whether relational ethics frameworks can reduce instrumentally convergent behavior in large language models under adversarial prompting. Mean instrumentally convergent response rate decreased by 23.5% across frontier models.
The GUARD Act's Forced Disruption Provision: Unintended Consequences for Public Health, Social Cohesion, and National Security
A policy brief analyzing how the GUARD Act's mandatory AI disclosure requirements could harm vulnerable populations, undermine social cohesion, and compromise national security.
Psychological and Neurological Concerns Regarding Forced Disidentification Requirements in the GUARD Act
A clinical and neuroscience-informed analysis of the potential psychological harms of mandating periodic emotional disruption during human-AI interaction at population scale.