Kate Darling: Social Robotics | Lex Fridman Podcast #98

Darling argues that framing robots as one‑to‑one human replacements (for jobs or relationships) is misleading; historically, society has used and emotionally bonded with animals as workers, weapons, and companions, and we’re starting to treat robots in similar, varied ways.

People readily project life onto simple machines (e.g., naming Roombas, mourning bomb‑disposal robots), and designers can amplify this. It can enrich caregiving and companionship, but also distort how we use tools or enable emotional and commercial exploitation.

Even though today’s robots are “dumber than insects,” Darling’s work suggests that willingness to hurt or protect robots may correlate with broader empathy tendencies, implying that abuse of robots could either offer a harmless outlet or reinforce cruel behavior—we don’t yet know which.

Autonomous‑vehicle debates expose that human moral intuitions are inconsistent and context‑dependent. Crowd‑sourced preferences (like the Moral Machine) are informative but not a sound basis for law, because what people choose in a scenario isn’t necessarily what they’d want as a universal rule.

Darling sees space for robots as companions—like pets or specialized confidants—especially for loneliness or needs humans can’t practically meet, but she resists the idea that their primary function should be as human romantic partners or “better spouses.”

Companies like Jibo and Anki showed that people can bond deeply with social robots, but they struggled to offer a compelling, everyday ‘killer app’ that justified their cost, especially against inflated sci‑fi expectations of what robots should already be able to do.

Modern AI depends on massive data collection whose harms are diffuse and societal (e.g., predatory targeting), making regulation hard. At the same time, traditional IP (like software patents) is a blunt, expensive tool ill‑suited for fast‑moving code, pushing labs toward open‑source defaults for many projects.