Robots Can Now Guess What You Want by Watching Your Choices

Yale and university researchers have developed a way for robots to understand what people want without being told. The secret? Robots subtly guide you into situations where your choices give them the information they need. They call these "critical decision points" — moments where the path you pick clearly shows your goal.

For example, imagine you're cooking with a robot. If it needs to know whether you're making soup or a salad, it might hand you the knife and watch what you reach for. Your choice reveals the recipe. In tests, this approach let robots guess human goals more accurately and earlier than older methods — during both cooperative tasks like cooking and competitive ones like hide-and-seek.

This isn't about mind-reading. It's about smart questioning through action. The robot balances two things: making progress on the task and learning more about what you intend. If it's unsure, it creates a small situation that forces your hand. If it's confident, it just helps.

The team tested their method in simulations and with real robots, in both a fully visible cooking scenario and a partially hidden hide-and-seek game. In both cases, the robot's guesses improved. For everyday life, this means robots could become better teammates — anticipating what you need in the kitchen, workshop, or hospital. It also hints at future AI opponents in games that predict your strategy faster. The catch: robots will sometimes nudge you in ways you might not even notice.