At the Inclusion Conference in Shanghai, a demonstrator asked a robot named Yansyn-X2 a strange favour: "Can you shed tears for me?" A few seconds later, a single tear rolled down its silicone cheek. The machine, built by the Ningbo firm Yanxi Technology, does this with a three-layer trick. An AI reads the tone of the conversation, bionic parts arrange the face into sadness, and a microfluidic system pumps a drop of liquid to the corner of the eye. No sorrow is involved. It is plumbing routed through a language model.
The obvious reaction is unease, and unease is worth pausing on, because a very different robot story from the same season provokes the opposite feeling for the same underlying reason. At the World Humanoid Games in Beijing in late August, the machines that went viral were not the ones that broke Usain Bolt's sprint record. They were the ones that faceplanted. Sprinters ran full-speed into walls, boxers swung at empty air and toppled over, and human attendants carried the fallen away on stretchers with deadpan care. Millions watched and laughed, and in laughing, warmed to them.
Both scenes are engineering, and they are engineering the same thing: our willingness to accept the machine. Researchers have a name for the second effect, the pratfall effect, from a 1966 study finding that capable people become more likeable when they make a minor, harmless blunder. In 2017, a team in Salzburg confirmed it holds for robots too; people significantly preferred a robot that made non-dangerous mistakes to one that performed flawlessly. A 2023 study in the journal Frontiers found the mirror image, that robots whose expressions match the emotional tone of a conversation are rated more likeable, trustworthy and human. A tear manufactures warmth. A stumble manufactures relatability. Neither adds an ounce of capability.
What is striking is how openly the people building this understand it. Xenia Bulatov, co-founder of the humanoid league Ultimate Bots, says her events treat failure as the show precisely because "if we show that robots are kind of clumsy and a little stupid right now, it puts people at ease." She claims the stretcher gag, one of the Beijing games' best visual jokes, started with her league, invented because dragging a human-shaped body offstage by the arm simply looked wrong. Trust in a technology shaped like a person, she has written, "does not get built by a product launch. It gets built by years of positive, repeated, emotionally legible exposure." That is a candid description of a campaign, not an accident.
The reason it matters is what these machines are actually for. The humanoids clowning in Beijing and the one weeping in Shanghai are being developed to move into healthcare, service work, retail, manufacturing, and in some cases the military. Getting the hardware to run around a corner is a solved-enough problem; getting the public to welcome it into a hospital ward or a living room is the harder engineering, and it is being done on our emotions rather than the robot's. As one writer at the Beijing games put it, the robots may be the butt of the joke, but we might be their useful idiots. The tear and the pratfall are not opposites. They are two tools reaching for the same soft target, which is the moment we stop keeping our guard up.