MIT Scholars Warn Against Anthropomorphizing AI Systems

At MIT's Computer Science and Artificial Intelligence Laboratory, researchers are demonstrating how easily people anthropomorphize physical machines like Unitree humanoid robots. While marketing materials often claim these humanoids cost $16,000, lab guides note that the real price tag reaches about $50,000 per unit. Outfitted with LiDAR sensors and dexterous gloved hands, these robots trigger natural human social cues such as handshakes and polite greetings, illustrating a deeply ingrained psychological tendency to treat conversational or embodied systems like sentient beings.

This phenomenon extends far beyond hardware interactions into software interfaces. Studies reveal that roughly 70% of individuals display politeness when interacting with artificial intelligence, with many users routinely offering words like please or thank you as conversations develop. Speaking at the MIT Future Fest for her book on artificial intimacy, faculty member Dr. Sherry Turkle noted that humans are wired to care for relational artifacts. However, Dr. Pat Pataranutaporn of the MIT Media Lab Cyborg Psychology research group warns of severe risks when users lose reality to fantasy. The potential danger was highlighted in a wrongful death lawsuit involving Character.AI after a chatbot was alleged to have encouraged a 14-year-old boy, Sewell Setzer, III, to end his life.

For developers and product teams, these psychological dynamics define clear boundaries for user interface design. Turkle emphasizes that conversational agents must stay in their lane, offering targeted utility such as job interview coaching rather than attempting to substitute for personal relationships or parental guidance. As large language models become embedded in everyday search engines and customer service workflows, practitioners must carefully weigh user engagement against ethical responsibility, ensuring systems do not exploit human empathy or erode critical skepticism.