Sleep-EVAL Aims

Current architecture updated | September 2026
Written by Maurice M. Ohayon, MD, DSc, PhD

Terminology. Sleep-EVAL denotes the domain-specific knowledge base for the evaluation of sleep disorders. It is not a separate inference engine. The unique inferential system is Ad-Infer, which uses Sleep-EVAL knowledge within its integrated clinical reasoning architecture.

WHY SLEEP-EVAL WAS CREATED

The first population surveys required an assessment method that could cover sleep-wake schedules, sleep habits, the principal sleep disorders, relevant psychiatric and medical conditions, treatment exposure and functional consequences while maintaining the logic of clinical diagnosis. Existing questionnaires generally addressed only selected symptoms or disorders and were not designed to conduct a systematic positive and differential diagnostic evaluation.

Sleep-EVAL was therefore developed as a specialized knowledge base containing the clinical concepts, questions, diagnostic criteria, exclusions, temporal relationships and associated medical and psychiatric knowledge required for the evaluation of sleep disorders. This knowledge can be applied consistently in general-population, clinical, family and longitudinal studies.

RELATIONSHIP TO AD-INFER

Ad-Infer is the integrated inferential system. It uses EVAL-KBS functions to process domain knowledge through Type-2 causal reasoning, fuzzy evaluation, learning processes, deep-learning mechanisms and Bayesian belief updating. Sleep-EVAL provides the sleep-disorders knowledge upon which this reasoning operates.

Language-model capabilities are integrated within Ad-Infer rather than operating as an autonomous diagnostic layer. They are stimulated by the current clinical hypotheses and belief state and are used to extend semantic understanding, adapt questioning and examine how new information relates to direct or indirect clinical causality. Information returned by the interaction is re-evaluated by Ad-Infer and may strengthen the positive diagnosis, modify the current belief state, or challenge a hypothesis and become part of the differential diagnosis.

CORE OBJECTIVES

  • Improve the completeness, consistency and clinical relevance of data collected in large population studies.
  • Represent symptoms and clinical states as graded constructs rather than forcing every observation into a binary category.
  • Evaluate positive diagnostic criteria together with alternative explanations, exclusion criteria and comorbid conditions.
  • Study direct and indirect causal relationships among sleep symptoms, psychiatric conditions, medical disorders, treatments and functional outcomes.
  • Permit previous inferences to be revised when longitudinal information, objective findings, treatments or subsequent clinical events become available.
  • Provide standardized clinical phenotypes for epidemiological, familial and longitudinal research.

HISTORICAL CONTINUITY

The Sleep-EVAL knowledge base and the Ad-Infer inferential architecture evolved together from early expert-system work into the present integrated clinical AI system. The underlying principles remain the same: explicit clinical knowledge, causal reasoning, management of uncertainty, reproducible diagnostic logic and the capacity to reconsider a conclusion when new evidence changes the clinical picture.