Credence before binary knowledge.
Belief is treated as a graded evidential posture, revisable in response to uncertainty, alternatives, and defeaters.
Research program
A connected program spanning epistemology, credencing, metaethics, evidence standards, rational disagreement, and AI-assisted inquiry.
Conceptual framework
The topics share one diagnostic question: does confidence remain answerable to available evidence?
Belief is treated as a graded evidential posture, revisable in response to uncertainty, alternatives, and defeaters.
Claims should state their support, confidence range, live alternatives, and the evidence that would trigger an update.
Moral language compresses affect, preference, vulnerability, social coordination, and proposed regulation without requiring stance-independent moral facts.
AI can widen search and simulate objections, but its output still requires source traceability, calibration, and human update discipline.
Epistemic architecture
Separating the layers avoids treating disagreement, uncertainty, ignorance, rationalization, and bad faith as one blunt failure.
What information, base rates, testimony, observations, and alternatives are actually accessible?
Which parts of that situation does the agent notice, understand, or suppress?
Does the resulting credence track the evidence as the agent understands it?
Does public language preserve uncertainty, or inflate partial belief into flat certainty?
Selected papers
Start with these representative publication records, then continue to the full first-party index.
Confidence, evidence, and the discipline of leaving room for live alternatives.
The public language required to communicate partial belief without fake precision.
What remains of value and human stakes after removing a stance-independent moral realm.
AI as a tool for search, comparison, and objection rather than an epistemic oracle.