By Deauna Mansfield

At the recent CASE conference, I explored a critical challenge facing today's educators: how to harness AI's educational potential while ensuring all students develop fundamental competencies in science. The presentation centered on a dual approach: creating both AI-collaborative and AI-resistant assignments that serve different but equally important learning goals.

The Equity Issue We Can’t Ignore
Recent research reveals troubling gaps: 88% of students use generative AI in assessed work, yet only 36% report receiving any AI instruction; 58% say they don't feel they have sufficient AI knowledge and skills, and 53% fear academic integrity accusations despite unclear guidelines. Male students, STEM majors, and socioeconomically advantaged students use AI significantly more than their peers (HEPI, 2025). Without intentional instruction, two harmful scenarios emerge: students lacking AI access fall behind in developing essential digital literacy skills, while students overusing AI without boundaries may miss foundational concepts like experimental design, data analysis, and scientific reasoning that are essential for scientific literacy.

A Framework for Balance
The solution lies in explicitly teaching students both when to collaborate with AI and when to work independently.

I shared five AI-collaborative strategies using science-specific examples:

1. AI as Critical Analysis Tool: Students ask ChatGPT to explain cellular respiration, then identify errors and correct them using peer-reviewed sources
2. AI as Research Assistant: Students use AI to brainstorm questions about CRISPR then verify with scholarly articles
3. Comparative Analysis: Students solve stoichiometry problems independently first, then compare their approach with AI's output
4. Prompt Engineering: Students guide AI to generate accurate explanations of the nitrogen cycle, learning that content knowledge drives effective prompting
5. Self-Directed Learning: Students leverage AI for personalized practice and concept review, verifying solutions independently to identify knowledge gaps

AI-resistant strategies that require authentic skill demonstration are equally important. Examples include:

1. Personal Connection: Assignments such as "Explain Newton's Laws using examples from your commute"
2. Design with Rationale: Require students to justify their scientific design choices on visual projects, e.g., explain chart types or provide diagram details
3. Process Documentation: Lab journals documenting hypothesis evolution and experimental adjustments
4. In-Class Performance: Lab practicals, whiteboard problem-solving, or think-alouds while analyzing data

Practical Implementation
The key is transparency; students need clear boundaries on AI use. Co-create a class agreement distinguishing "green light" AI use (brainstorming lab questions, generating practice problems) from "red light" use (completing problem sets, lab reports). Always explain the "why": "This lab quiz ensures you can safely handle equipment" or "Here you'll learn to fact-check scientific claims."

Start small—choose one collaborative and one resistant strategy to pilot. Use AI as a tool to redesign existing assignments. Most importantly, talk with students about what works and adjust accordingly.

The Path Forward
We cannot close our eyes to AI’s presence in education, but neither can we allow it to replace fundamental learning. By thoughtfully implementing both approaches, we equip all students, regardless of background, with both AI literacy and the independent scientific competencies that remain irreplaceable.

Works Cited
HEPI. (2025, August 27). Student generative AI survey 2025. Higher Education Policy Institute.

Tags