Elena Leib
Relational thinking: The overlooked component of executive functioning

Relational thinking, the ability to represent abstract, generalizable relations, is a core component of reasoning and human cognition. There are open questions about the nature of the relationship among relational thinking and executive functions (EFs) – the set of effortful processes that enable goal‑directed behavior, such as working memory and inhibitory control. Relational thinking and EFs have been shown to support learning and reasoning in the domains of math and science and predict academic achievement. However, relational thinking is not often examined in the context of EFs, and therefore, it has been difficult to determine the extent to which relational thinking is a separable capacity from EFs. We hypothesize that relational thinking should be considered an EF in its own right, as one of the core mid-level cognitive abilities that support cognition and goal-directed behavior. One way to test this hypothesis is by examining how EFs and relational thinking come together to predict performance on tasks in the domains of math and science. I will present findings that support this hypothesis from a study that investigates performance on fraction comparison in the math domain. I will also describe future research directions in the domain of science involving scientific reasoning.

Eliza Kosoy
Title: Are kids a-maze-ing at exploration? Children’s intrinsic exploration of mazes compared to exploration algorithms

Abstract: Children are naturally curious, and now even reinforcement learning models in machine learning are channeling this child-like curiosity in order to improve exploration in novel environments. For example the Intrinsic Curiosity Module (ICM) in which prediction error serves as an intrinsic reward that enables an agent to better explore its environment. Another basic strategy for exploration is Depth First Search (DFS). That is, continuous exploration along a new path until a dead-end is reached, followed by backtracking until the closest previous branching point is reached. Using DeepMind Lab, a first-person maze environment, we are able to directly compare the exploration behavior of children vs. exploration algorithms. We examine the results of the children and show some comparisons to algorithms.