Affective Computing, Anthropomorphism, and the Psychological Risks of Emotionally Responsive AI

Artificial Intelligence (AI) systems are increasingly engineered to detect, interpret, and respond to human emotional expressions, raising fundamental questions about whether machine emotion recognition genuinely detects emotions or merely identifies associated behavioural patterns. Furthermore, these advancements prompt consideration of the psychological implications arising from human interaction with systems perceived as emotionally responsive agents. Through an integrative review of the literature on the key explanatory theories of human emotional processes, the technical and conceptual underpinnings of Human-Computer Interaction (HCI), affective computing, and emotion-recognition technologies, this work elucidates the mechanisms of personalization, feedback loops, anthropomorphism, and the psychological processes underlying emotional attachments to artificial agents. Furthermore, it analyses the associated behavioural risks, including patterns of over-reliance, addictive use, and overconsumption. Drawing these perspectives together, this thesis posits that contemporary AI systems exhibit an increasing sophistication in simulating emotional responsiveness; however, they remain fundamentally distinct from genuine emotional comprehension. This critical distinction carries significant implications for how emotionally responsive technologies should be designed and regulated, particularly as human reliance upon them expands. The thesis therefore contributes to a more critical understanding of artificial emotional intelligence and its effects on human users, particularly regarding social behaviour and psychological well-being.