AI in Education: Why Human Choices Still Matter Most

For students, parents, and teachers trying to make sense of artificial intelligence, the most important question may not be what a chatbot can do. It may be who decides when, why, and how it is used. In an October 2026 interview published by the Harvard Gazette, Isaac Kohane argued that “the bulk of the benefit and the threat is actually humans.” His point offers a useful frame for education: AI can support learning or weaken it, depending on the choices people make around it.

According to the Gazette’s account, Kohane is the inaugural chair of Harvard Medical School’s Department of Biomedical Informatics and founder of the Human Values Project, an international research initiative focused on values embedded in clinical AI. Although the interview centers substantially on medicine, it raises practical issues for classrooms and homes: when AI should help a learner, when it should not do the work for them, and why local values and human relationships cannot simply be delegated to a tool.

Moving beyond AI panic and AI promises

According to Kohane, public discussions can drift toward two extremes: a “doomer” view that treats AI as an uncontrollable threat and a utopian view that assumes it will reliably improve everything. He instead emphasizes human agency. In the interview, he says people choose to apply AI to particular tasks, and people remain responsible for the consequences of those choices.

That distinction matters in education because “AI use” is not one activity. A student asking for a clearer explanation of a difficult algebra step is doing something different from submitting AI-generated work as their own. A teacher using a tool to brainstorm multiple examples is doing something different from using automated output as the final judgment of a student’s understanding. The technology may be similar; the learning purpose, oversight, and effect on the learner are not.

For families, this can make conversations less abstract. Rather than asking whether AI is categorically good or bad, adults can ask: What is the student trying to learn? What part can they already do independently? What support would help without replacing their thinking? And how will they check an answer?

Help when stuck, not a substitute for practice

Kohane told the Gazette that his college-age children find AI useful when they are stuck, but that using it to solve problems can leave them less prepared for exams. According to the interview, he does not claim there is a single “magic recipe” for AI in education; appropriate use may differ by task and specialty.

That is a valuable caution for K-12 settings. Learning often requires productive struggle: retrieving information, attempting a method, noticing an error, and revising. If a tool provides a finished response before a student has attempted the task, it can remove the very practice that helps knowledge stick. On the other hand, a well-timed hint, an alternate explanation, or a few practice questions can help a student move past a genuine obstacle.

A practical routine can preserve that balance:

Attempt first. Students should write what they know, identify the confusing step, or make a first draft before requesting AI support.
Ask for guidance, not completion. Useful prompts request a hint, a worked example on a similar problem, or questions that reveal the next step.
Verify independently. Students can explain the idea aloud, solve a fresh problem without assistance, or compare the response with class materials.
Show the learning process. Teachers and parents can ask students to describe what the tool suggested and what they changed or understood afterward.

Free support can be most constructive when it is structured around these habits. For example, COSMIQ is a free voice-driven AI tutor, free forever for every K-12 student, and can be used as a place to ask for an explanation after an independent attempt rather than as a replacement for that attempt.

Values are part of the design, not an afterthought

In the Gazette interview, Kohane describes work by the Human Values Project involving doctors presented with difficult clinical scenarios. According to Kohane’s account, doctors and AI models differed in how they weighed values such as autonomy, beneficence, and justice; the interview also says doctors’ views varied across regions. His broader argument is that AI systems can reflect consistent ethical orientations, while human communities can hold different, legitimate priorities.

The education parallel is not that schools face the same decisions as hospitals. It is that educational AI also carries assumptions. A tool’s definition of a “good” answer, its feedback style, the information it foregrounds, and the data it asks students to share can all shape an experience. Schools therefore need more than a rule that says “use AI” or “ban AI.” According to Kohane, universities have a role in considering appropriate applications and ethical frameworks; educators and families can bring that same deliberative approach to their own settings.

Useful questions include:

Does this use deepen understanding, or mainly speed up task completion?
Can a student challenge, correct, or explain the output?
Does the activity protect time for discussion with teachers and peers?
Are expectations clear about attribution, privacy, and assessment?
Would this approach work fairly for students with different access needs and learning profiles?

Human connection remains the educational work

Kohane told the Gazette that AI could make people more siloed or help them look beyond disciplinary boundaries; the outcome, he said, depends on how people use it. He also argued that education may need to clarify the value of personalized engagement that students cannot obtain elsewhere.

For schools, that suggests a constructive priority: use AI to create more room for human teaching, feedback, and conversation, not less. A teacher may use an AI-assisted prompt to generate extra practice pathways, but still be the person who recognizes discouragement, notices a misconception, or helps a student connect a topic to a future goal. Parents can support this by treating AI conversations as a starting point for reflection, not a private shortcut that ends the discussion.

Students preparing for a quiz or building a skill can also use guided practice tools such as the free practice hub by grade and subject while keeping a simple record of which questions they could answer alone and which required help. That record can make gaps visible and gives a teacher or caregiver something concrete to discuss.

A shared responsibility, not a technology verdict

The Harvard Gazette interview does not offer a final formula for AI in education. Its strongest contribution is a reminder, attributed to Kohane, that people have agency and should not surrender it to either fear or enthusiasm. For educators, families, and students, the next useful step is modest but meaningful: set a purpose before using AI, keep human judgment in the loop, and make independent thinking the measure of whether the help was worthwhile.