5 Questions Science Educators Should be Asking About AI

As AI tools become a routine part of how students study, science educators are rethinking assessment, understanding, and course design. These five questions highlight what needs to change in science education in the AI era.

AI is already part of how students learn. Recent surveys show just how widespread this is. A global study from the Digital Education Council found that 86% of students are using AI in their studies, with more than half using it weekly. 

They’re using it to study, to check answers, to generate explanations, and sometimes to skip steps altogether. That shift is happening faster than most courses are adapting to it. A Harvard study found that nearly 90% of students use AI, and about one in four use it as a substitute for things like office hours or assigned readings

This is the context behind the questions we’re addressing at the upcoming Reimagining Science Education Summit in San Francisco. Here are some of the many we'll be addressing.

1. What should students be able to do without AI?

AI can solve a chemistry equation, walk through a physics problem, and explain a biology concept in plain language. That can genuinely help students learn. The problem is that it can also produce the right answer without producing any understanding.

So where does independence still matter? Is it foundational knowledge, problem setup, Interpretation of results, or something even harder to name

If everything is AI-assisted, it becomes harder to tell what a student actually knows versus what they can access.

2. Am I teaching students how to question AI, or just use it?

Using AI is easy. Evaluating what it produces is a different skill entirely.

A student who gets a convincing but slightly off explanation may never realise it. And once a misunderstanding takes root, it's stubborn. Critical evaluation of AI output isn't happening automatically; it has to be taught.

If a student asks AI to explain a concept and gets a convincing but slightly wrong answer, they may not realize it. And once that misunderstanding sticks, it’s hard to correct.

3. What does it mean to “understand” science now?

Dr. Megan W. Taylor, Chief Teaching + Learning Officer at the Exploratorium frames the shift this way: “What if we made classrooms a fundamentally different place? Instead of assessing how quickly you can solve a quadratic formula, which any AI tool can now do in seconds, what if we asked why you would need the quadratic formula in a given situation? Or even better, why you would model something with a parabolic equation instead of a linear one to make an accurate prediction about the world.”

“Those are the kinds of questions people actually ask as they make sense of the world, modeling it with science and mathematics, and then applying that understanding in real situations.” 

The line between knowing something and being able to access it is getting blurry. How we assess whether or not a student understands needs to catch up.

4. What parts of my course are still measuring what they're designed to?

A lot of assignments were designed for a world where getting to an answer took time and effort. Now a take-home problem set or a standard lab write-up can now be generated in seconds.

Detection tools are evolving, but institutions are starting to acknowledge that enforcement alone doesn’t solve the problem. It’s worth asking what still holds value, creating the thinking, time and effort that they did in a pre AI world.

5. How do I design learning experiences that actually require students to think?

AI can support deeper thinking, or it can shortcut it. The difference often comes down to how the work is structured.

If the task is just to produce an answer, AI will do that quickly. If the task requires interpretation, judgment or decision-making, AI becomes a tool rather than a replacement. Especially if it means explaining why an answer makes sense, comparing approaches, applying a concept to a real and specific scenario.

These aren’t easy questions, and they don’t have one right answer. But they’re worth sitting with, because AI is changing both the tools students use and the role of the classroom itself.

We’ll be digging into these questions and more at the “Skills for the Scientist of 2030” panel during the Reimagining Science Education Summit on May 15.

Further reading:
How the AI Tutor Drives Student Success
Cheaters Never Win, But Are They Really Cheating? Rethinking Integrity in the AI Classroom

FAQs

How is AI changing science assessments?

AI is pushing educators toward "process-based" grading. Instead of only checking for the correct final answer, instructors are increasingly grading the steps taken, the justification of methods, and the critical analysis of AI-generated content.

Can AI help with lab reports?

Yes, AI can generate standard lab report structures and even interpret data. To counter this, many instructors are moving toward in-class data analysis or requiring students to reflect on specific, unique variables encountered during their physical lab session.

How does Achieve help with academic integrity?

Achieve uses a variety of assessment types and LearningCurve adaptive quizzing that provides unique paths for each student. This makes it much harder to use a one-size-fits-all AI prompt to bypass the learning process.