Do Language Models Exhibit the Same Cognitive Biases in Problem Solving as Human Learners?