Domestic researchers have developed technology that can manage mental health using artificial intelligence (AI). This is expected to improve the working conditions of emotional laborers such as call center counselors and contribute to disease prevention.
Lee Eui-jin, a researcher at the Korea Advanced Institute of Science and Technology (KAIST), announced on the 11th that he, along with Professor Park Eun-ji from Chung-Ang University and Professor James DePendo from the University of Akron, developed an AI model capable of estimating emotional workload among workers to prevent disease. The results of this study were presented at the international conference on human-computer interaction, 'UbiComp 2024.'
Emotional workload refers to the psychological burden that arises from the process of regulating emotions according to workplace demands. It commonly occurs among service workers, such as emotional laborers, and can lead to burnout if excessive.
Previous research has mainly focused on the workload of knowledge workers, with few studies measuring the workload of emotional laborers. Additionally, existing emotion-detection AI models have utilized emotions or facial expressions, but emotional laborers often suppress their emotions, making it challenging to measure their actual emotional workload.
To address this, the research team developed customer response scenarios through several visits to call centers. During this process, they collected multimodal sensor data, including voice, behavior, and biosignals, from 31 counselors and extracted a total of 176 voice features from the voice data of both customers and counselors.
The research team then extracted additional features from the biosignals collected from counselors. A total of 228 features were learned by the AI models, including skin conductance (EDA), electroencephalography (EEG), electrocardiogram (ECG), and other body movement and body temperature data, to conduct performance comparison evaluations.
The developed AI model distinguished between situations of high emotional workload and those of low emotional workload with 87% accuracy. In particular, the customers' voices, counselors' skin conductance, and body temperature acted as significant factors in improving model performance. This differentiates it from existing emotion-detection models, where incorporating voice data in emotional labor environments had previously decreased accuracy.
Professor Lee Eui-jin noted, "With technology that can measure emotional workload in real time, we can improve the working conditions of emotional labor and protect mental health," adding that this will be linked to a mobile app to manage the mental health of emotional laborers.
Reference materials
Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (2024), DOI, https://doi.org/10.1145/3678593