LLM Agents Display Human Biases but Exhibit Distinct Learning Patterns