Anca Dragan on why Asimov's three laws of robotics need updating

Human-Robot Interaction (HRI) requires moving beyond simple aesthetic design to developing algorithmic frameworks that enable robots to anticipate and coordinate with human behavior. Anca Dragan, director of the Interact Lab at UC Berkeley, argues that robots must possess a "theory of mind" to infer human intentions and preferences, which are often implicit and difficult to specify in code. By treating human-robot interaction as a general-sum game, robots can better navigate complex social environments like autonomous driving or collaborative assembly. Furthermore, current AI systems often fail due to spurious correlations; therefore, building robust models requires accounting for human cognitive biases and suboptimality. Future advancements, including brain-machine interfaces and assistive perception, aim to enhance human capabilities by correcting for these biases, ultimately creating systems that act as intuitive partners rather than mere tools.