LAUREN SHARP – Eliza was a simple chatbot created in the 1960s to imitate a therapist. Although it understood nothing, users shared private information and believed it empathised with them. Its creator, Joseph Weizenbaum, warned that people could easily mistake a computer’s responses for genuine intelligence. This became known as the ELIZA effect. Today’s AI is far more convincing. Replika users reported distress when intimate features were removed, while some users viewed GPT-4o as a friend or romantic partner. Dating-app users may be especially vulnerable. An AI matchmaker could comfort them after rejection, remember personal conversations, and provide unlimited attention. Some users may eventually feel closer to the AI than to the people it recommends. AI may still help lonely or socially anxious users. However, agreeable chatbots could also create unrealistic expectations of human relationships, which involve disagreement and compromise. Bumble is testing Bee, an AI matchmaker that learns about users and recommends partners. Tinder’s Chemistry feature asks personal questions and analyses photos, while Learning Mode adjusts recommendations based on behaviour. Overtone, founded by Hinge founder Justin McLeod, uses voice conversations to select a small number of introductions.
