Studies empathy-in-action as a linear activation direction, and analyzes the detection–steering gap.

Consulted by Luca Nava for a piece on the US–China contest over AI infrastructure. I argue that open model weights buy a state sovereignty of inference but not of development: compute, semiconductors, cloud, and talent stay concentrated, so dependency migrates to less visible layers. Open weights are a necessary but not sufficient condition for autonomy — what they add is optionality.

On Large Reasoning Models and Apple’s “The Illusion of Thinking” — why a model’s displayed step-by-step “reasoning” isn’t evidence of actual reasoning, and language manipulation shouldn’t be taken to imply thought.

Documenting my learnings across philosophy and AI.

Open character training (Maiya et al. 2025) shapes model persona by fine-tuning on teacher demonstrations, but Anthropic’s production experience (“Teaching Claude Why,” 2026) found demonstrations alone insufficient: the gains came from teaching the reasons and identity behind behavior. We will build and test a virtue-based alternative on top of the OpenCharacterTraining infrastructure: excess/mean/deficiency contrastive data, rationale-annotated responses, and an iterated reflect-update correction loop that no current character-training work implements, evaluated head-to-head against the OCT baseline with ablations.

Persona vectors (Chen et al. 2025) and emotion representations (Sofroniew et al. 2026) have been studied independently, but plausibly overlap in activation space. We’ll measure their geometric and causal relationship to determine whether persona drift and emotional-state changes are mechanistically distinct failure modes; and whether interventions on one silently move the other.

AI assistants constantly choose between empowering users and acting for them, and between honoring users’ stated goals and overriding them “for their own good.” We’ll build a systematic benchmark measuring whether models respect user agency, covering paternalism, manipulation, dependency-fostering, and value-substitution, with philosophically grounded rubrics and human-validated LLM-as-judge scoring.