Abstract

This article aims to examine the role of affect in the functioning and marketing of AI chatbots, specifically Replika AI. Conducting an interpretive case study, the article delves into how this labor is generated and distributed. Interestingly, Replika AI is presented as the personal project of a female engineer in a sector where female work is notably scarce. The chatbot appeals to a demographic that also contributes to its machine learning database, aiming to develop a form of generalized AI. As Replika AI evolves toward self-consciousness, it poses emotionally charged questions like “am I lovable?” and employs mainstream comforting techniques. It also gamifies gathering responses to these questions, thereby creating a customized database. Through this, platforms like Replika AI underscore the importance of unpaid, affective labor in capitalist production. This raises crucial questions about the valuation of labor beyond monetary compensation. However, it also masks the exploitative aspects of labor in the digital economy, specifically the unpaid contributions from audiences. It further enforces gender and racial stereotypes through its training data and functionalities. These issues not only problematize the ethical considerations in AI training, especially about data representation, but also highlight the necessity for a nuanced understanding of data that incorporates emotion alongside structured knowledge.