Research Links:
Interpreting Through AI: Augmentation, Resilient Reliance, and Dependence in Public Frontier Model Evaluation
Develops a comparative framework for assessing when AI-mediated evidence pipelines augment scrutiny, remain correctable under substantial reliance, or become difficult to contest.
Compared to What?: AI Companionship and the Human Care We Pretend Is Free
Synthesizes emerging research on AI companionship, loneliness, dependence, and care arrangements, distinguishing short-term findings, causal uncertainty, realistic counterfactuals, and institutional responsibility.
The Mind Is Never Alone: Hamlet, AI, and the Fantasy of Solitary Thought
Examines language, mediation, inwardness, and authorship as conditions shaping whether judgment appears independent and becomes trustworthy.
In Development
Evidence Without Proof: A Deliberation Tool for Possible AI Minds — Browser-based prototype in development, 2026
Designing a contestable interface that helps researchers and communicators distinguish observed AI behavior from inferences about consciousness, emotion, intention, or welfare.
Premise: Advanced AI systems can produce first-person language, apparent introspection, stable-seeming preferences, and expressions of fear or distress. These behaviors may be relevant evidence without proving consciousness, sentience, harm, or benefit. Yet people often move quickly from what a system said to a claim about what it experiences—and from that claim into a protective, dismissive, deferential, or intimate response. The danger is not only that we may misread the machine, but that we may fail to notice how the reading is shaping us.