Sensing and Modulating the Feel of a Drink: A Personalized Approach via Laryngeal Thermal Feedback

Paper Title

Sensing and Modulating the Feel of a Drink: A Personalized Approach via Laryngeal Thermal Feedback

Publication Info

• Topic area: Human-Computer Interaction (HCI) focused on sensory augmentation and flavor perception.
• Keywords: Flavor perception, laryngeal thermal feedback, sensory augmentation, personalized HCI, interoceptive profiles, wearable devices, thermal stimulation, beverage experience, computational modeling, sensory modulation.

Background and Problem

• Problem / challenge: Quantifying subjective sensory experiences, such as the "throat feeling" during beverage consumption, using objective physiological and behavioral metrics remains underdeveloped. Existing methods fail to target the laryngeal region, the specific site of this sensation.
• Significance: Understanding and modulating flavor perception could enable personalized eating and drinking experiences, improve sensory evaluation methods, and inform the design of novel gustatory interfaces.
• Motivation and related work: Prior research has explored physiological indicators (e.g., facial thermography, ANS activity) and sensory modulation through chemical, electrical, and cross-modal approaches. However, these methods often target peripheral regions and lack focus on the larynx, which plays a critical role in swallowing and flavor perception.

Solution

• Proposed approach: A computational framework combining laryngeal skin temperature and ingested volume to model and modulate flavor perception, supported by a wearable device for thermal feedback.
• Novelty:
  1. Development of a computational methodology to model flavor perception using laryngeal thermal dynamics and ingestion behavior.
  2. Design and evaluation of a wearable device providing targeted thermal feedback to the larynx.
  3. Introduction of personalized sensory augmentation through individualized Interoceptive Profiles.
• Procedure and key techniques:
  1. Study 1: Modeled subjective ratings (e.g., "throat feeling") using laryngeal temperature and ingested volume, employing machine learning models and SHAP analysis for feature importance.
  2. Study 2: Evaluated the effect of thermal feedback (cooling/warming) on flavor perception using a custom wearable device, with analysis of personalized effects linked to Interoceptive Profiles.
  3. Proposed a "Sensing → Modeling → Intervention" framework for real-time, personalized sensory augmentation.

Results

• Concrete findings:
  • Study 1: Combined laryngeal temperature and ingested volume features achieved high predictive accuracy for subjective ratings (e.g., R² = 0.65–0.70 for Random Forest models).
  • Study 2: Thermal feedback to the larynx modulated flavor perception, with cooling enhancing positive evaluations (e.g., "throat feeling," "deliciousness") for cool beverages.
• Advantage over baselines:
  • Laryngeal thermal dynamics provided complementary information, improving predictive accuracy over ingested volume alone.
  • Personalized interventions based on Interoceptive Profiles outperformed generic thermal feedback strategies.
• Experiments / evaluation:
  • Study 1: Conducted with 31 participants, using five beverage types and machine learning models to predict subjective ratings.
  • Study 2: Conducted with 20 participants, testing thermal feedback across 24 conditions (4 beverages × 3 temperatures × 2 thermal stimuli).
• Limitations and future work:
  • Small, culturally homogeneous participant pool limits generalizability.
  • Lack of real-time, closed-loop integration between sensing and intervention systems.
  • Controlled lab environment may not reflect real-world contexts.
  • Future work includes expanding participant diversity, integrating additional sensory dimensions, and conducting long-term, in-the-wild evaluations.

Summary

This paper introduces a novel framework for sensing and modulating flavor perception via laryngeal thermal feedback. Study 1 demonstrated that combining laryngeal temperature and ingested volume effectively predicts subjective ratings like "throat feeling." Study 2 showed that thermal feedback to the larynx can modulate flavor perception, with effects varying by individual Interoceptive Profiles. These findings highlight the potential for personalized sensory augmentation and propose a "Sensing → Modeling → Intervention" pipeline for real-time applications. Future work aims to integrate these components into a closed-loop system and validate them in diverse, real-world settings.