Whenever a new Boston Dynamics robot video drops, my first reaction is always panic. I’m not trying to be an anti-technology alarmist; I just genuinely feel a sense of dread when I see an extremely human-like robot doing a backflip, or picking up a box, or any number of the impressive robo-tricks they can now do. My terror at seeing the Boston Dynamics door-opening dog is the new high watermark. It immediately conjured images of the Jurassic Park raptor opening the door and the Black Mirror “Metalhead” robot dog grabbing the knife. But no matter how scared I get for the end of humanity, I immediately feel bad for the robots during the robustness tests. As it turns out, I’m not alone in that emotional turnaround.

According to Kiley Hamlin, a psychologist at the University of British Columbia, even 6-month-old infants can tell when something is helping or hindering. In an experiment in 2007, she found that even a rudimentary puppet show was enough for the infants to infer which shape was exhibiting prosocial behavior, an action intended to help others, and which shape was being a real jerk. She showed one shape traveling up an incline and then a second shape that helped with the journey. Finally, a third shape hindered the journey. Hamlin and her team switched the shapes at random and blindfolded the parents so they couldn’t influence their infants, and the results found that an overwhelming majority of the infants could distinguish each behavior.

This might end up killing humanity through our own misplaced sympathy, but it could also be the thing that stays our execution at the hands of our inevitable robot overlords.