A glass box, not a black box.
Claridian is a research and educational prototype that estimates psychosis risk from clinical indicators such as suspiciousness, unusual thoughts, perceptual changes, social withdrawal, thinking and attention difficulties, sleep patterns, daily functioning, family history, and substance use. Information can be collected through a guided questionnaire, a written note, or a conversation with an AI therapist.
Unlike many AI systems that rely on opaque neural networks, Claridian uses transparent fuzzy-logic reasoning. Each clinical signal is converted into interpretable levels such as low, moderate, or high and evaluated through a visible set of expert-defined rules designed to reflect clinical reasoning.
Every risk score is fully traceable. Users can see which factors contributed to the assessment, how the relevant rules were applied, and how the final 0–100 risk score and risk band were produced. This transparency makes the system easier to understand, evaluate, and trust, turning risk estimation into a process that can be inspected rather than simply accepted.
What the score measuresRisk of developing psychosis, not a diagnosis that you have it.
Claridian estimates the likelihood that a person is in the prodromal phase, the at-risk mental state that can precede a psychotic disorder such as schizophrenia, and how strongly their current signals point toward it. It does not detect or diagnose psychosis, and it does not say a person will certainly develop it.
A high score means several early warning signs cluster together in a pattern that research associates with raised risk during this phase. It is a probability, not a destiny: most people who receive early support never go on to develop a psychotic disorder. The score is designed to prompt a conversation with a qualified clinician, never to replace one.
