A Connecticut immunologist recently used an artificial intelligence model to solve a three-year-old biological mystery about how immune cells fight cancer. This discovery shows a shift in how science is conducted, moving from manual trial-and-error toward AI-supported exploration. For parents and educators, the breakthrough demonstrates why science education must focus on critical thinking rather than memorizing facts.
What Happened
In 2022, Dr. Derya Unutmaz, a professor at The Jackson Laboratory and the University of Connecticut, ran an experiment on how glucose affects disease-fighting T cells. He found that T cells exposed to deoxyglucose, a molecule similar to glucose, became highly inflammatory. Cells in a low-glucose environment did not change. Unutmaz and his team could not explain the result, so they shelved the project.
According to the OpenAI report, Unutmaz revisited the data in late 2025 using GPT-5 Pro. The AI model quickly suggested that the molecule interfered with the production of IL-2, a protein that normally prevents cells from becoming inflammatory. GPT-5 Pro also simulated a separate, unpublished laboratory experiment and predicted how certain T cells destroy lymphoma. Unutmaz said the AI acted as a collaborator, narrowing down possible hypotheses and saving months of laboratory work.
The Bigger Picture
This approach is also changing classrooms. Instead of forcing students to follow step-by-step instructions in traditional labs, some schools use AI tools to help students grasp abstract scientific concepts. A study in Scientific Reports found that undergraduate students using AI-supported virtual laboratories reduced their scientific misconceptions by 71.6%. Students in traditional labs only reduced misconceptions by 36.4%. These virtual labs also lowered educational costs by 80.9%, which helps underfunded schools access advanced experiments.
Research in Frontiers in Education shows that AI learning analytics and virtual tools improve understanding of STEM topics by increasing student engagement.
Educators warn that AI cannot replace core knowledge. A study on physics students in the Jurnal Penelitian Pendidikan Sains (JPPS) revealed that many students accept AI answers too quickly without questioning their accuracy. Students who verified AI claims showed stronger critical thinking skills. An analysis in PLOS Biology similarly warns that outsourcing writing and data analysis to large language models risks "decoupling writing from thinking" and can harm reasoning skills.
What This Means for Families
As AI models become common lab tools, the goal of science education is shifting. Students no longer need to spend hours memorizing chemical formulas or running repetitive procedures that an AI can simulate in seconds. Instead, they must learn enough about the subject to recognize when an AI's output is flawed or correct.
School systems are drafting policies to guide this transition. For example, the Appleton Area School District K-12 AI Guidelines state that AI should act as a partner for critical thinking, requiring students to cite AI tools and banning plagiarism. The Minnesota Department of Education and Scottish school guidelines argue that humans must remain the central decision-makers in the classroom. As we previously reported, using these tools requires school systems to monitor privacy and safety to prevent the misuse of student data.
What You Can Do
- Parents can teach skeptical inquiry at home by encouraging children to treat AI-generated answers as hypotheses to verify, not absolute facts.
- When helping with homework, focus on the "why" instead of the "how." Ask children to explain the science concept rather than just looking up the answer.
- Parents can also review their school's AI policy and talk with teachers to understand how tools like GPT-5 Pro are used in class and what citation rules exist.