Compare three ways to study with ChatGPT
At a glance
A learner compared direct answers, hints and comprehension questions across three subjects using ChatGPT. This workflow is listed under Learning.
- Agents / tools
- OpenAI
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) Choose three specific questions: a quadratic equation, why 'have lived' is used in an English sentence, and why the Meiji Restoration happened using Perry's arrival, shogunate politics and foreign pressure. 2) For each question, first ask ChatGPT for the answer and method; then try thinking independently while asking it for hints without the answer, requesting another hint if stuck. 3) Try a third approach by asking ChatGPT to act like a teacher and question whether you understand, then explain the ideas in your own words. Rate the subjective feeling of having learned on a ten-point scale. 4) After some time, attempt the same questions again without AI and record what you can do unaided.
How it went
The author's average subjective scores across the three subjects were 2.7 for direct answers, 5.3 for hints and 8.0 for comprehension questions. In the later unaided check, mathematics was mostly completed and English and history were completed.
Worth knowing
This was one person's three-question comparison, with repeated questions and subjective scores. The later retest did not separate the effects of the three approaches; it does not establish that a method improves grades.