Better answers, broader thinking: What students gain from ChatGPT and critical-thinking training
2026-08-28 · OpenAI
Study Overview
A randomized study of more than 1,000 students examines four connected dimensions:
- ChatGPT use
- Critical-thinking training
- Originality
- Student performance on a real university assignment
Rather than relying on an artificial lab task, the research is grounded in an authentic course assignment, making the findings more relevant to real teaching practice.
Core Question
The study addresses a practical issue in higher education: when generative AI enters coursework, what do students actually gain?
It goes beyond a simple score comparison and considers whether students:
1. Produce more complete and structured answers,
2. Demonstrate broader thinking,
3. Maintain originality and independent judgment,
4. Benefit from combining AI support with critical-thinking instruction.
Design Strengths
The research uses a randomized design across a large student sample, improving comparability between intervention conditions and reducing bias from pre-existing differences.
Because the outcome is measured through a real university assignment, the conclusions are more actionable for course design, assessment strategy, and classroom policy than results from short, decontextualized tests.
Main Takeaway
The key message is that ChatGPT and critical-thinking training are not simple substitutes.
Instead, they appear to provide distinct and potentially additive benefits:
- Tool-level gains: generative AI can support response efficiency and answer completeness.
- Skill-level gains: critical-thinking training is more directly tied to reasoning quality, independent judgment, and depth of thought.
This reframes the debate from “AI or no AI” to “how to integrate AI while preserving and strengthening core academic abilities.”
Implications for Universities
The study points to an evidence-based direction for higher education:
- AI integration should not be evaluated only by speed or output volume.
- Critical-thinking instruction should be built in alongside AI use.
- Assessment should consider both final answers and the quality of thinking behind them.
In practice, this means generative AI can function as a learning aid, but institutions still need intentional pedagogy to cultivate originality, argumentation, and independent intellectual work.
Conclusion
With a large randomized sample and a real assignment context, the study suggests a balanced view: ChatGPT can improve aspects of performance, while critical-thinking training strengthens higher-order cognition. Their value is complementary rather than mutually exclusive. For universities, the practical lesson is to design learning environments where technological assistance and cognitive skill development advance together.