Peter Watson

In this chapter, the focus is on paving the way for a novel approach to exploring embodied emotions without explicit labels. It begins by establishing the groundwork necessary for unsupervised exploration of emotions, emphasising the need for innovative techniques. The chapter then delves into the intricacies of learning algorithms tailored for detecting emotions in embodied contexts, emphasising their adaptability to unsupervised scenarios. Furthermore, it explores the application of unsupervised learning methods to various data modalities, including EEG signals, images, text data, and audio signals, showcasing the versatility of this approach in deciphering emotions without relying on predefined emotional labels.

Emotions rarely unfold in isolation within a single channel of human experience; rather, they dynamically evolve through coordinated changes in voice, facial expression, autonomic physiology, and cognitive appraisals that are continuously shaped by temporal, contextual, and cultural factors. Each modality, be it vocal prosody, micro facial movements, or cardiac variability, offers a distinct yet complementary vantage point into the underlying affective process ( Scherer, 2009 ). The meaning of emotion thus emerges from the integration of these expressive and physiological modalities, where cross-channel synchrony provides more diagnostic value than any single signal alone ( Kragel & LaBar, 2016 ). Through this multimodal lens, emotion becomes a temporally extended, embodied event rather than a transient, discrete reaction, unfolding across milliseconds in neural and physiological shifts and across minutes or hours in expressive and cognitive alignment.

Emotions shape human experiences deeply, influencing decisions, relationships, health, and overall well-being. Among various ways to express emotions, textual data uniquely captures complex emotional differences through words and language structures. With the rapid growth of digital communication platforms, social media, and online interactions, analysing text has become an essential method for emotion recognition, significantly impacting psychology, health informatics, marketing, and artificial intelligence (AI). This integration aligns with the Artificial Psychology (PsAIchology) framing that connects AI methods and psychological science ( Farahani et al., 2024 ). Textual data naturally contain emotional information, which can be explicit, through clearly stated emotional words (e.g., happy , anxious , angry ), or implicit, conveyed through tone, context, metaphor, or sentiment. Emotional analysis of text requires robust computational methods to decode these intricate emotional expressions accurately. The rise of Natural Language Processing (NLP) techniques has greatly improved our ability to interpret and analyse emotional content in textual data.

Human emotions do not exist in isolation; they manifest vividly through interactions between individuals and their environments. Central to these interactions are visual stimuli (faces, body postures, and contextual scenes) which powerfully evoke and shape emotional experiences. The processes of emotion elicitation, particularly through visual and embodied modalities, have profound implications for understanding emotional dynamics. This chapter explores how affective images and deliberate emotion elicitation techniques help uncover the intricate connections between visual perception, bodily responses, and emotional experience. By integrating historical insights, contemporary methodologies, computational advances, and ethical considerations, we establish a comprehensive understanding of how visual and emotional stimuli shape embodied emotional responses.

This chapter delves into the profound insights that can be extracted from medical imaging, with a specific focus on CT and MRI data for understanding and identifying emotions. It commences by discussing the potential emotional insights that can be gleaned from these imaging modalities, underlining their significance in the domain of embodied emotion. Subsequently, the chapter explores the intricate process of neuroimaging data preprocessing, an essential step in preparing CT and MRI data for subsequent emotion analysis. Moving forward, it uncovers the emotional patterns hidden within CT and MRI scans, shedding light on how these patterns can be interpreted. Furthermore, the chapter delves into the development of automated emotion detection algorithms tailored for neuroimaging data, substantiating these concepts with compelling case studies and experiments that underscore the practicality and efficacy of this approach.

This chapter starts by elucidating the significance of EEG in emotion detection, exploring its potential as a powerful tool for decoding emotional states. Moving forward into the intricacies of EEG data acquisition and preprocessing, ensuring the data’s reliability and accuracy. Subsequently, it unveils the techniques for feature extraction from EEG signals, providing insights into the key parameters that signify emotional responses. This chapter then delves into the application of machine learning methodologies for EEG-based emotion recognition, showcasing real-world case studies and experiments that demonstrate the practicality and effectiveness of this approach.

This chapter begins with an analysis of what we have learned about the interfaces between neuroscience and emotions. This topic provides key lessons for the study and understanding of Embodied Emotions, as it considers the pivotal role of neuroscientific research methods for the study of emotions. Then, the relevance of biomarkers for the study of Embodied Emotions is discussed, emphasising the breakthroughs and thresholds implicated in the search for biomarkers. Moreover, it lays a foundational understanding of Python for delving deeper into emotional artificial intelligence and provides the critical steps involved in preparing data for the comprehensive analysis of embodied emotions. Beginning with the transition from raw data to embodied emotion insights, it explores various data acquisition techniques tailored for modelling emotions within AI. The chapter then proceeds to address the essential stages of data cleaning and feature extraction, ensuring the quality and relevance of emotional data. It extends its focus to specific data types, including time series, visual, audio, and text data, each requiring unique preprocessing techniques to uncover emotional insights within. Additionally, the chapter discusses gaze data analysis as a vital component of embodied emotion recognition and explores the power of multimodal data fusion, enabling a holistic understanding of emotions as they manifest across different data modalities.

This chapter explores the groundbreaking intersection of reinforcement learning and emotions, heralding a new frontier in human-computer interaction. This chapter initiates an exploration of emotion-driven reinforcement learning, revealing how emotions can be harnessed to enhance interactions between humans and machines. It then delves into the crucial application of reinforcement learning in monitoring mental health, demonstrating how AI can play a pivotal role in supporting emotional well-being. Additionally, the chapter delves into emotion-driven education, showcasing how reinforcement learning can be tailored to create personalised and emotionally responsive educational experiences. This exploration not only highlights the transformative potential of emotions in AI but also underscores their significance in various facets of our daily lives.

Computational Approaches to Emotion in Artificial Psychology provides readers with a comprehensive introduction to how emotions can be processed by AI systems.