Pairing generative artificial intelligence with explicit critical-thinking lessons produces the best and most original student work. While chatbots can make writing cohesive, students need human reasoning skills to generate unique ideas.
What Happened
Researchers from Bocconi University and OpenAI Economic Research recently conducted a randomized experiment involving over 1,000 first-year undergraduate students. According to the published research paper, students were asked to develop marketing recommendations for the university's merchandise store. The researchers split the students into four groups: those with access to ChatGPT, those given specific training in causal reasoning, those given both, and a control group with neither.
The results, published in an OpenAI research announcement, showed distinct benefits for each approach. Students who used ChatGPT scored nearly a full point higher on a five-point grading rubric. Their papers had better structure and mirrored professional marketing analyses. However, AI access alone did not make their ideas more original. Students who received causal reasoning training, which teaches people to analyze cause-and-effect relationships, produced the widest variety of unique ideas. When students had access to both ChatGPT and the reasoning training, their work was highly polished and original.
The Bigger Picture
Educators currently debate how AI tools affect cognitive development. Some researchers warn that relying heavily on technology can erode basic skills. For example, a qualitative study available as research on ResearchGate found that constant use of AI writing assistants reduced students' cognitive effort and damaged their writing confidence over time.
Other experts argue that using AI simply changes academic challenges. A study from Iowa State University found that writing with AI demands more critical thinking because students must handle the planning and evaluation. Instead of acting as a shortcut, AI is a "cognitive scaffold" that increases student motivation and autonomy, according to research in Discover Education.
This scaffolding works best when combined with deliberate teaching. A meta-analysis published in the Educational Psychology Review confirmed that generative AI tools improve critical thinking performance when they include tutoring or guided interactions. To make this work in classrooms, researchers in the Muallim Journal of Social Sciences and Humanities recommend structured frameworks, like the AI-Enhanced Critical Thinking model, to keep students thinking for themselves. Research in Scientific Reports points out that technology alone does not improve higher-order thinking. Instead, teachers must design tasks that require cognitive conflict and detailed explanation. Because of universities launching free ChatGPT programs, schools must balance access with structured training.
What This Means for Families
For parents and educators, these findings show that banning AI does not work, and neither does letting students use it without guidance. The study also exposes a flaw in traditional grading. In the Bocconi experiment, the standard grading rubric failed to reward original ideas because it only measured basic business goals.
AI can instantly generate clean essays, meaning a final written product no longer proves that a student understands the material. Schools must redesign how they measure learning. Guidelines from the Washington University in St. Louis Center for Teaching and Learning suggest that educators separate core skills students must master themselves from supportive tasks where AI is acceptable.
To prevent students from skipping the learning process, teachers must grade how students think rather than just what they produce. Writing in the Journal of Instructional Design and Technology, researcher Tina Austin suggests using reflection loops where students explain their choices and critique the AI's output.
What You Can Do
You can help children practice causal reasoning by asking "why" and "what if" questions to look past surface-level answers. Encourage them to trace cause-and-effect relationships in their homework rather than accepting the first answer they find.
Focus on the writing process instead of only evaluating the final draft of an assignment. Ask students to explain how they came up with their ideas and how they verified their information.
Finally, establish multi-stage prompts when using AI for homework help. Teach students to treat the chatbot as a sounding board. They can input an idea, ask the AI to find flaws in their logic, and then revise their work based on that feedback.