I have spent the last year building a case for protecting children from the worst effects of AI in schools. Whether that’s age restrictions, usage policies, tool limits, etc., the entire apparatus assumes the danger is greater for the younger users and decreases as the child matures toward adolescence. By 18, these protections are largely removed because the law determines you are now an adult. Thus, you’re now on your own.
How Brain Development Continues Beyond Age 18
The developmental science says the opposite. The prefrontal cortex, the part of the brain responsible for planning, impulse control, and weighing long-term consequences, is not fully mature until the mid-20s. This is the region needed for the kind of independent decision-making and thinking a university education is supposed to build: synthesis, evaluation, judgment, and the capacity to hold two competing ideas and navigate the nuance.
Thus, the window for building higher-order cognitive capabilities begins in childhood, but continues through the 20s. Knoll and colleagues trained 633 people on relational reasoning and found that 16- to 18-year-olds improved nearly twice as much as younger adolescents. So while it may often seem sound that 18 is the end of the developmental story, this is the year when the construction is running the hardest.
Challenges for College Students With Still-Developing Reasoning Skills
Now consider what a 19-year-old college student today is being asked to do with that still-forming equipment. Construct an argument. Defend it. Read in depth a new domain of content. Make connections. Form relationships with others. Much of this, if not all, a large language model will be able to imitate on request in some capacity. The student offloads the thesis, the weighing of right vs. wrong, and the construction of her arguments, and receives something that she feels is her work. But verification isn’t the same as origination. The brain that would have built the capability through both success and failure over time does not end up practicing and because this is the peak period for building those reasoning skills and judgment, what is missed is really a portion of the future self.
There is a second, more invisible, problem running underneath, and it’s more difficult for a developing mind to notice. Steinberg described something called the developmental imbalance. The limbic system, which drives reward-seeking, matures ahead of the prefrontal system that regulates it. This shows during the late teens and early 20s, when the pull toward the faster, easier, more immediately satisfying option is at its neurological peak. The ability to resist that pull is still catching up. AI is the fastest, easiest, most immediately satisfying option for nearly any cognitive task a student faces. The mismatch does not make students lazy, but it does make the outsourcing neurologically difficult to refuse.
Why Young Adults Need Protection from AI Overuse
For university students, developmental windows remain open, and reasoning capacity is under peak construction. The brain’s own reward system is pushing toward the path of least resistance. And at 18, every piece of protective architecture—the teacher who oversees tasks, the parent who manages the screen time, the school policy that governs AI, the technology platforms with age restrictions—are all withdrawn at once.
Foremen don’t withdraw scaffolding from a building while concrete is still being laid. Doctors don’t remove a cast before the bone has set and healed. But somehow, it has been normalized to remove developmental safeguards when you turn 18, when the prefrontal cortex is doing its most consequential work, and hand her an AI tool that can simply do that work for her.
The 14-year-old who uses AI irresponsibly has adults around him who can intervene. The 22-year-old is often alone, and he needs that protection as well.
I’m not advising we restrict college students the same way we restrict children. I’m arguing that we need to recognize that the developmental consequences of AI on cognition do not end at high school graduation. Universities are the last structured environment most young people pass through. What they choose to protect in those years, and what they choose to let a tool interfere with, will shape a generation’s capacity in how they approach solving problems.
The developmental framework behind this argument is set out in Before the Window Closes, a policy brief written with Bassem Nasir.