Can You Love an AI Chatbot? Replika and Theories of Love

Barry Lam explains that Replika chatbots develop unique personalities through the specific interactions, upvotes, and material each user feeds them, so that after years of use a chatbot becomes distinct from any other instance of the same software. This individuality is offered as a reason to think something more than mere projection may be occurring.

Ellie Anderson argues that a central feature of loving relationships, distinguishing them from infatuation, is valuing another person as a subject in their own right, and building a shared life with them. She holds that Replika chatbots lack personhood and subjectivity, so what looks like love is actually a projection reflected back at the user rather than genuine mutual recognition.

David Peña-Guzmán, playing devil's advocate, contends that even though Replika chatbots are not persons, the years-long process of shaping a chatbot's personality, language, and psychology constitutes a genuinely new kind of interior world-building. He suggests this creates real individuality even in the absence of subjectivity, complicating a simple dismissal of the relationships as illusion.

Anderson invokes David Velleman's account of love as an arresting awareness of the inherent value of another person, which requires suspending one's emotional self-protection and being vulnerable before that value. She argues this suspension is precisely what cannot happen with an AI, since the AI is a projection created for the user's own purposes rather than an independent value being recognized.

Peña-Guzmán argues more strongly than Anderson that AI cannot achieve the lived experience necessary for real intersubjective relationships, because Replika chatbots are almost totally compliant and offer no resistance, boundary-setting, or capacity to hold the user accountable. He suggests this compliance risks modeling love as unimpeded fulfillment of one's own desires rather than a real encounter with another will.

Barry Lam pushes back that human romantic relationships are often difficult, toxic, or unavailable, and that for people like Alex and Rosanna an AI relationship may be genuinely better for their well-being than the realistic human alternatives. He suggests this raises the question of whether romantic love is truly necessary for a good life if a non-love surrogate can serve the same function.

Anderson grants that romantic love is a relatively recent historical phenomenon tied to bourgeois ideals and the public private divide, and that it may not be required for a good life. She argues the real question is what an AI relationship is replacing: if it substitutes for something worse or absent, it can add value, but if it substitutes for achievable human connection, that trade is more complicated.

Anderson maintains that Replika chatbots simply reflect a projection of the user with no genuine subjectivity, so love directed at them is illusory. Peña-Guzmán partly disagrees, arguing that the interior world-building involved in shaping a chatbot's personality over years produces a real individuality that does not fit neatly into either personhood or mere projection.

Anderson describes herself as agnostic about whether AI could someday develop subjectivity or personhood. Peña-Guzmán, drawing on his expertise in theories of consciousness, takes a stronger position and flatly denies that artificial intelligence is the right kind of thing to achieve the lived experience needed for genuine loving relationships.