Abstract
On the technical side, both traditional, shallow feature-based methods (e.g., Support Vector Machines and Random Forest) and deep neural network models (e.g., LSTM and CNN types) have been utilized. A key focus has been the development of multimodal systems that integrate diverse data sources, such as facial expressions, voice tone, dialogue sentiment, and physiological signals. Multiple signals enable the aggregation of complementary information, leveraging the strengths of various measurement techniques to enhance robustness and accuracy. Technology is only effective with users; thus, beyond developing methods, fostering adoption among professionals and citizens is essential. Alongside technical advancements, researchers have explored psychological and social factors influencing the acceptance and trust of AI systems, particularly in healthcare and education. Trust has proven pivotal for AI adoption, with studies investigating how to build it and what system attributes ensure trustworthiness. Key trust factors include explainability, transparency in processes and data usage, and the credibility of the institutions behind AI development. This Research Topic explores novel theories, methodologies, and applications of Emotion AI in healthcare and education, addressing technical aspects, models, and AI acceptance and literacy.
Introduction: The "Last Mile" of Affective Computing
Emotion AI, or Affective Computing, is transitioning from controlled laboratory environments into the complex, "in-the-wild" realities of healthcare and education. While the field has existed for decades, it currently faces a significant "last mile" problem. Technically, models that perform exceptionally on curated datasets often struggle in real-world settings due to environmental factors like poor lighting and physical occlusions. Socially, the adoption of these tools depends heavily on the trust and acceptance of clinicians, educators, and patients.
This editorial synthesizes recent research addressing these dual challenges: enhancing technical robustness through multimodal sensing and data-efficient learning, while simultaneously investigating the psychological factors—such as transparency and AI literacy—that underpin successful human-AI interaction.
Technical Advancements: Robustness and Data Efficiency
A primary focus of recent research is the development of multimodal systems that integrate facial expressions, voice tone, dialogue sentiment, and physiological signals (e.g., EEG). By aggregating complementary information, these systems aim to enhance accuracy in noisy environments.
Bridging the Performance Gap
Studies on parent-infant interactions highlight the difficulty of real-world application. Automated facial coding software currently detects only about 25% of expressions compared to manual human coding in naturalistic settings. This gap underscores the need for models capable of handling rapid movement and varied lighting conditions, ensuring that Emotion AI can function reliably outside of the lab.
Learning with Limited Data
In clinical settings, expert-labeled data is often scarce and expensive. Innovative approaches in semi-supervised and self-supervised learning are proving effective. For instance, semi-supervised models for stress monitoring have achieved 77% accuracy using only 17% labeled data, matching the performance of fully supervised approaches. Additionally, self-supervised techniques like Contrastive Predictive Coding (CPC) have shown promise in extracting meaningful features from EEG data without requiring manual labels, offering a practical path for continuous monitoring.
Human-Centric Design: Trust and Literacy
Technology is only as effective as its adoption. Research into the psychological and social factors of AI acceptance reveals that trust is a pivotal driver, particularly in sensitive fields like healthcare.
The Transparency Premium
In high-stakes environments, such as AI-powered pharmacy services, patients naturally exhibit lower trust in AI compared to human providers. However, providing informed consent and transparent communication significantly bridges this gap. Explainability is not merely a technical feature but a functional requirement for user trust; when patients understand how a system works, their willingness to use it increases.
AI Literacy and Psychological Needs
In education, fostering AI literacy is crucial for student engagement. Research indicates that satisfying students' basic psychological needs—specifically autonomy and competence—is more effective at improving AI literacy than providing direct technical support. This suggests that AI-enhanced learning environments must be designed to empower the user rather than just provide instruction.
Diverse User Attitudes
Understanding user diversity is essential for designing emotionally intelligent assistants. Research among older adults identified four distinct personas: Cautious Optimists, Skeptical Traditionalists, Positive Optimists, and Technophiles. Tailoring AI interactions to these specific attitudes can significantly improve acceptance and utility across different demographic segments.
Future Directions and Ethical Considerations
The emergence of Generative AI (GenAI) and foundation models like GPT-4 presents a new frontier for Emotion AI. These models offer the versatility to recognize emotional indicators across diverse contexts and may help mitigate the data scarcity issues currently hindering specialized model development. Furthermore, there is growing interest in the emotional analysis of groups and crowds, broadening the field's scope.
However, significant hurdles remain. Clinical applications are often limited by small sample sizes and a lack of longitudinal studies. Furthermore, the use of Emotion AI raises critical ethical concerns regarding data privacy, surveillance, and the potential misuse of sensitive personal information. Future research must prioritize cross-cultural validation, long-term real-world evaluation, and robust ethical frameworks to ensure that Emotion AI serves as a beneficial and trusted tool in next-generation healthcare and education.
Summary updated