Epistemic Rubber Ducking with Large Language Models

Using large language models for reflective thinking and metacognition, rather than generating facts or trustworthy outputs