EduHK AI in Educational Study

2021

Building Emotional Intelligence in AI: Validating Physiological Data for Educational Tools

No. of Participants:	239 Subjects (Divided into two stages)
​Nature of Event:	Academic Research Study
Products Used:	Upmood Bands, Upmood Mass
Metrics Collected:	HRV, Mood (Happiness/Sadness)

The Quest for Objective Emotion Recognition

This case study focuses on an academic research initiative aiming to advance the development of Artificial Intelligence (AI) tools in the education sector. The research team, led by Professor Joanne Wai Yee Chung, sought to move beyond subjective self-reporting and establish an objective method for assessing specific emotions—namely happiness and sadness.

The primary goal was to validate whether physiological data, specifically Heart Rate Variability (HRV), could serve as a reliable ground truth for training AI models. By partnering with Upmood, the researchers utilized wearable technology to bridge the gap between human feelings and machine understanding.

Experimental Procedure: Multi-Stage Data Collection

The study was conducted in 2021 and involved a cohort of 239 subjects. To ensure the data was robust, the engagement was split into two distinct stages:

• Stage 1 (Calibration): Half of the subjects were asked to categorize various video clips sourced from YouTube. This step established a baseline for the emotional content of the stimuli.

  
  

• Stage 2 (Validation): The video clips were presented randomly to the remaining subjects. During this stage, participants wore Upmood bands to capture real-time physiological data while simultaneously inputting their subjective emotion ratings.

This dual-layer approach enabled the team to correlate participants' self-reported feelings with objective biological signals captured by Upmood.

Why Upmood?

Upmood’s technology was critical to this study because it provided a scalable, non-invasive way to capture Heart Rate Variability (HRV)—a key biomarker for emotional state. Upmood’s PPG sensors offered Objective Accuracy by capturing precise HRV data, providing the "hard numbers" needed to validate the prediction of happiness and sadness. Furthermore, the Upmood Mass system demonstrated Scalability by allowing the researchers to efficiently monitor a large number of subjects (239) without compromising data quality. Finally, the Ease of Integration provided by the unobtrusive wearable bands ensured that the subjects' natural emotional responses to the videos were not hindered by bulky equipment.

Key Outcomes and Results

The study successfully utilized a Partial Least Squares Discriminant Analysis (PLS-DA) model to analyze the data, yielding significant findings for the future of AI in education:

☺️Happiness Detection: The model achieved a sensitivity of 70.7% in correctly identifying a subject’s happiness.

😞 Sadness Detection: The model achieved a specificity of 58.4% in correctly identifying sadness.

💓 Validation of HRV: The results supported the hypothesis that the prediction of happiness and sadness using HRV measures is viable.

🧾 Proof of Concept: The study confirmed that HRV measures provide a legitimate, objective method to assess emotions, laying the groundwork for more empathetic and responsive educational AI tools.

Conclusion

This research demonstrates the vital role of physiological monitoring in the development of academic technologies. By leveraging Upmood’s ability to capture objective emotional data, the research team successfully validated a model for recognizing happiness and sadness. This partnership highlights how Upmood can empower the academic community to build data-driven, emotionally intelligent solutions for the classroom of the future.

Thank you!

We extend our sincere gratitude to Professor Joanne Wai Yee Chung, Henry Chi Fuk So, Marcy Ming Tak Choi, Vincent Chun Man Yan, and Thomas Kwok Shing Wong, for their dedication to advancing the science of emotion monitoring and its application in education. Their rigorous work has validated the reliability of physiological data in developing the AI tools of tomorrow.

If you're interested in conducting your own academic study on emotion validation or trying out Upmood products, please send us an email at support@upmood.com