A Personal Thermal Comfort Model Based on Causal Artificial Intelligence: A Physiological Sensor-Enabled Causal...

Area of Science:

• Artificial Intelligence
• Environmental Science
• Human-Computer Interaction

Background:

• Personal thermal comfort significantly impacts health, well-being, and productivity.
• Subjectivity and dynamic environmental factors make understanding thermal comfort challenging.
• Existing models often lack predictive and explanatory power for individual comfort.

Purpose of the Study:

• To develop a personal thermal comfort model using causal artificial intelligence.
• To encode personal thermal comfort satisfaction by connecting human factors and environmental influences.
• To provide a human-intelligent framework for predicting and explaining thermal comfort.

Main Methods:

• Utilized a causal-and-effect framework with random variables and a structural causal model.
• Employed do-calculus (d-separated, d-connected) for causal structure representation and common-sense reasoning.
• Quantified model parameters using a directed acyclic graph and variable elimination with observational data.

Main Results:

• The proposed model effectively encodes physiological and physical factors for predicting thermal comfort satisfaction.
• Demonstrated robust prediction and explanation capabilities, aligning with human-like interpretations.
• Verified causal relationships using causal odd ratio, sensitivity, and impact analyses.

Conclusions:

• The causal AI model offers a novel approach to understanding personal thermal comfort.
• The framework enables intelligent systems to predict and explain individual thermal comfort dynamics.
• This research contributes to improved occupant well-being and productivity through personalized environmental control.