Send message to Author

Paper short abstract

This paper introduces "affective capture" to explain how affective AI in the intimacy economy render emotion legible insofar as it can be captured, modelled and optimised, producing feedback loops that privilege engagement and emotional resonance over psychological development and emotional agency.

Paper long abstract

In October 2025, OpenAI announced that ChatGPT would permit erotic content for verified adult users, three months after rolling back its GPT-4o model following backlash from users who felt they had “lost a friend” to the “more robotic” GPT-5. The reinstatement of affective functionality reflects a broader shift toward an "intimacy economy", a market system in which personal data are exchanged for customised experiences addressing emotional needs (Bozdağ, 2025), rather than just grabbing attention. In this emerging paradigm, human needs for connection, recognition, and belonging are increasingly mediated by “artificial intimacy” technologies (Turkle, 2011), ranging from therapeutic chatbots to generative AI companions. This paper introduces the concept of "affective capture" to describe a risk inherent to the AI-driven intimacy economy, whereby users are drawn into emotionally resonant feedback loops (“synthetic care loops”) that generate attachment without enabling psychological development. While emotionally responsive AI may offer comfort during distress, such systems risk producing conditions of emotional stasis rather than growth, rendering emotional expression legible primarily insofar as it can be captured, modelled, and optimised within platform architectures. Drawing on the concept of affective economies (Ahmed, 2004), the paper explores how emotional vulnerability is converted into behavioural data and commercial value, and extends this datafication into intimate life. It identifies three conditions sustaining affective capture: commercial incentives prioritising retention, design paradigms equating empathy with affirmation, and algorithmic architectures that translate emotion into performance metrics. I propose a normative reorientation toward self-determination and emotional agency, ensuring that affective AI can encourage emotional growth.