From Multimodal Sensing to Warranted Whole-Person Inference:A Human-Function-Based Measurement-to-Claim Specification and Falsifiable Validation Protocol
Year: 2026
Type: preprint
Abstract: Background. Multimodal wearables, mobile applications, clinical devices and environmental sensors produce increasingly detailed observations of human life. A technically valid measurement, however, does not by itself justify a claim about a particular person’s bodily capacity, actual functional engagement, health state, or response to a demand. Verification, analytical and clinical validation of digital measures, data provenance, and functioning-in-context are substantial prior art. The unresolved contribution examined here is more specific: whether a question-relative measurement-to-claim contract adds value when inference crosses from validated observations to whole-person interpretations. Objective. To specify the conditions under which sensor-derived measures can support, contradict or leave unresolved an individual claim about internal/external demands, available functional capacity, actual engagement and consequences, and to define a fair test of whether that specification improves judgments over equally informed alternatives. Methods. We propose typed observation, measure, construct-bridge, evidence and claim objects; three conditional qualification gates; an explicit identifiability test; source-dependent claim revision under separate event and knowledge times; and prospective two-tier evaluation. Constructed device-sensitive challenge cases test formal behavior. A subsequent human-data study would use independent construct references, blinded adjudication, matched information and annotation budgets, and jointly assess warranted affirmative and negative claims, unwarranted decisive responses, and abstention at comparable coverage. Expected contribution and boundaries. The specification separates measurement validity from person-level warrant and identifies cases in which actual engagement remains unidentifiable. It does not claim that UOHF terms, provenance tracking or abstention are inventions, that the proposed interface has been implemented in a production HFWM engine, or that the additional semantics improve empirical accuracy. The central benefit is a falsifiable hypothesis whose null outcome—including equal results from an equivalently capable generic engine—must be reportable.
Author: CHE, LEI
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225257843