When it comes to the rapid integration of artificial intelligence into K-12 classrooms, educators are being pulled in two very different directions.

One prevailing media narrative stokes such profound fears about the emerging strengths of artificial intelligence that it could lead one to believe it will soon be “game over” for everything we know about good teaching. At the same time, a sweeping executive order from the White House and tech-forward education policymakers paint AI as “game on” for designing the educational system of the future.

I work closely with educators across the country, and as I’ve discussed AI with many of them this spring and summer, I’ve sensed a classic “approach-avoidance” dilemma — an emotional stalemate in which they’re encouraged to run toward AI’s exciting new capabilities while also made very aware of its risks.

Even as educators are optimistic about AI’s potential, they are cautious and sometimes resistant to it. These conflicting urges to approach and avoid can be paralyzing.

Related: A lot goes on in classrooms from kindergarten to high school. Keep up with our free weekly newsletter on K-12 education.

What should responsible educators do? As a learning scientist who has been involved in AI since the 1980s and who conducts nationally funded research on issues related to reading, math and science, I have some ideas.

First, it is essential to keep teaching students core subject matter — and to do that well. Research tells us that students cannot learn critical thinking or deep reasoning in the abstract. They have to reason and critique on the basis of deep understanding of meaningful, important content. Don’t be fooled, for example, by the notion that because AI can do math, we shouldn’t teach math anymore.

We teach students mathematics, reading, science, literature and all the core subjects not only so that they will be well equipped to get a job, but because these are among the greatest, most general and most enduring human accomplishments.

You should use AI when it deepens learning of the instructional core, but you should also ignore AI when it’s a distraction from that core.

Second, don’t limit your view of AI to a focus on either teacher productivity or student answer-getting.

Instead, focus on your school’s “portrait of a graduate” — highlighting skills like collaboration, communication and self-awareness as key attributes that we want to cultivate in students.

Much of what we know in the learning sciences can be brought to life when educators focus on those attributes, and AI holds tremendous potential to enrich those essential skills. Imagine using AI not to deliver ready-made answers, but to help students ask better, more meaningful questions — ones that are both intellectually rigorous and personally relevant.

AI can also support student teams by deepening their collaborative efforts — encouraging the active, social dimensions of learning. And rather than replacing human insight, AI can offer targeted feedback that fuels deeper problem-solving and reflection.

When used thoughtfully, AI becomes a catalyst — not a crutch — for developing the kinds of skills that matter most in today’s world.

In short, keep your focus on great teaching and learning. Ask yourself: How can AI help my students think more deeply, work together more effectively and stay more engaged in their learning?

Related: PROOF POINTS: Teens are looking to AI for information and answers, two surveys show

Third, seek out AI tools and applications that are not just incremental improvements, but let you create teaching and learning opportunities that were impossible to deliver before. And at the same time, look for education technologies that are committed to managing risks around student privacy, inappropriate or wrong content and data security.

Such opportunities for a “responsible breakthrough” will be a bit harder to find in the chaotic marketplace of AI in education, but they are there and worth pursuing. Here’s a hint: They don’t look like popular chatbots, and they may arise not from the largest commercial vendors but from research projects and small startups.

For instance, some educators are exploring screen-free AI tools designed to support early readers in real-time as they work through physical books of their choice. One such tool uses a hand-held pointer with a camera, a tiny computer and an audio speaker — not to provide answers, but to guide students as they sound out words, build comprehension and engage more deeply with the text.

I am reminded: Strong content remains central to learning, and AI, when thoughtfully applied, can enhance — not replace — the interactions between young readers and meaningful texts without introducing new safety concerns.

Thus, thoughtful educators should continue to prioritize core proficiencies like reading, math, science and writing — and using AI only when it helps to develop the skills and abilities prioritized in their desired portrait of a graduate. By adopting ed-tech tools that are focused on novel learning experiences and committed to student safety, educators will lead us to a responsible future for AI in education.

Jeremy Roschelle is the executive director of Digital Promise, a global nonprofit working to expand opportunity for every learner.

Contact the opinion editor at opinion@hechingerreport.org.

This story about AI in the classroom was produced by The Hechinger Report, a nonprofit, independent news organization focused on inequality and innovation in education. Sign up for Hechinger’s weekly newsletter.