Proceedings of the Annual Meeting of the Cognitive Science Society

Much of early childhood learning involves learning how to correctly carry out a series of steps - an algorithm - to reach a goal. It is not known whether the general ability to make inferences about novel algorithms develops in childhood, alongside the ability to execute new algorithms, or if the inferential capacity develops later in life. We tested 4- to 8-year-old children using a set of storybooks depicting animals playing procedural games, each corresponding to an algorithm with a definite set of possible ending states. Critically, the stories depicted only the first few steps of each procedure. Then, participants were asked questions about the properties of those algorithms. We found that children successfully made multiple types of inferences about novel algorithms. Even the youngest children were already sensitive to algorithmic properties. These results provide the first evidence that children understand that algorithms are more than just a series of steps.