C-GRASP: Clinically-Grounded Reasoning for Affective Signal Processing

Abstract

Inferring affective and stress states from wearable physiological signals is a major direction in educational technology, but prevailing methods are black-box and hard for instructors and students to trust or act on. C-GRASP (Clinically-Grounded Reasoning for Affective Signal Processing) embeds clinical-medical reasoning structure into the physiological-signal pipeline, providing traceable, clinically-anchored justification for inferred affective states.

Problem & Motivation

Inferring affective and stress states from wearable signals such as HRV and EDA has become a major direction in educational technology. Prevailing methods rely on black-box models without grounding in clinical-medical knowledge, making them difficult for instructors and students to trust and difficult to act on pedagogically. How can affective-signal analysis carry clinical interpretability?

Method

We propose C-GRASP (Clinically-Grounded Reasoning for Affective Signal Processing), a framework that combines clinical-medical reasoning structure with the physiological-signal processing pipeline, providing traceable justification for inferred affective states. Through structured clinical-reasoning steps, AI's analysis of physiological signals is mapped to validated physiological–affective association mechanisms in the medical literature.

Findings

• Formalizes the clinical reasoning workflow for application to affective-signal analysis.
• Improves the interpretability and trustworthiness of physiological-signal inferences.
• Provides a clinical foundation for physiological-sensing applications in educational settings.

Implications

Clinically-grounded affective-signal analysis allows instructors to adopt wearable data more confidently for understanding learner state. C-GRASP's interpretable reasoning trace replaces a black box with a transparent physiological → affective → learning chain that instructors, students, and researchers can inspect together — providing a clinical theoretical foundation for physiologically-aware AI tutoring.

Citation

@inproceedings{cheng2026cgrasp,
  author    = {Cheng-Lin Cheng and Ting-Chuan Lin and Chia-Kai Chang},
  title     = {{C-GRASP}: Clinically-Grounded Reasoning for Affective Signal Processing},
  booktitle = {Proc. Annual Int. Conf. of the IEEE Engineering in Medicine and Biology Society (EMBC)},
  address   = {Toronto, Canada},
  year      = {2026},
  month     = jul,
  note      = {Accepted},
}