Disclaimer: I do not have children, and when I was a child/teenager, I was an extreme misfit, so take what follows with a truckload of salt, and keep in mind that this is meant as kind of a wistful “would be nice if this were the case” rather than “rawr, my way or the high way, there oughtta be a law.”
Broadly speaking, I personally don’t think minors should be on any form of the internet, including multiplayer online games, until they’re old enough to get their driver’s license, and as with the driver’s license, there should be some kind of initial training process under adult supervision, ideally by the minor’s guardian. I don’t think laws banning children from the internet are enforceable, so I don’t know if legislating this is the answer. I do believe that the minor’s legal guardian should be potentially liable for any criminal activities the minor commits, online or offline, and that the guardian should have to prove in a courtroom beyond a reasonable doubt that they were not in a position to prevent the minor’s criminal activities. Maybe that is already the case, I don’t know.
Education should ideally be offline, dealing with memorization-oriented reading, writing, math (and the guardian’s religion, if relevant) until the learners show that a different approach is necessary, meaning that they’ve either matured enough for a more analytical approach to work for them or have been shown to be neurodivergent in a way that requires a different style. I am in favor of learning block printing and cursive at a young age, and (short) assignments requiring longhand writing, because there are studies showing that this helps with remembering stuff, and it’s also useful in situations where your smartphone battery is dead, your phone coverage is garbage, etc. I am not in favor of the “three crossouts and you have to start the handwritten assignment over again” rule I was trained up on, and I am also not in favor of the way certain teachers used to hype the students with pretty penmanship. You should praise a child who does well at handwriting exactly as much as a child who does well at subtraction, neither more nor less. Legibility by the teacher is the short-term goal, but in the long-term nobody but the actual individual and their loved ones and a few banking or government officials needs to be able to read their handwriting, usually.
The more advanced subjects that should be taught once children are mature enough to grasp them include physics, civics, history, home economics (including basic cooking, cleaning and cloth mending, managing their personal finances) and however much in the way of advanced math and foreign languages they can handle. There is a case to be made for starting foreign languages very young, when rote memorization is easier, but this is something that should be based on the child’s interests and abilities, rather than being an automatic thing. As a possible starting point, I personally favor Latin, which has a strong influence on the grammar of German and the vocabulary of the Romance languages (Spanish French etc), plus a weaker influence on English through French and the Greco-Latin mashup vocabulary used in science and engineering. Ancient Greek or Ancient Hebrew are potentially workable for the kind of child/teenager who likes secret codes like the one in Sherlock Holmes and the Dancing Men, and offer insights into ancient Greek philosophy and literature on the one hand, and the Bible on the other.
I am unconvinced that literature should be taught as a separate thing. Encourage children and teenagers to read steadily at their own pace, about things that interest them and that their guardians are okay with. In middle school, I was obsessed with the Sherlock Holmes short stories, with O. Henry, and with a particular translation of the Grimm Fairy Tales, which must have been pretty unexpurgated given the gory content and stern portrayal of certain heavenly figures in it. I don’t necessarily endorse any of those for other children. I’m just saying that the motivation to read stuff can take many forms. If the child would rather read about how cars are made, or how dresses are made, and doesn’t share their guardians’ love for, say, Harry Potter or Middle Earth or Jane Austen, the guardians need to be fine with that as well.
Maybe screens can serve a limited purpose in preteen education, like showing the kids book adaptations or interesting documentaries about how things are made. Actual keyboarding and computer operation should be more of a teenaged thing, and should be taught offline, on air-gapped machines, with internet operation, including a broad understanding of the risks, being taught at around age fifteen-sixteen at the young end. Also around this age is when the average kid will gain enough autonomy to warrant a dumbphone. Ok, fine, if you were a very responsible, autonomous ten-year-old working on the family farm, and you hold your own kids to the same standard, go ahead and give them dumbphones at an early age, or even smartphones if you think they can handle it. But you and your kids are outliers, and this isn’t aimed at you.
A smartphone should be something young people should buy when they’re old enough to legally buy beer. Yes, there are young people who will find a workaround, but smartphones are a lot more expensive than beer, which means that it’s more inconvenient for the young people to pull off. If you want to argue the libertarian, spherical-chickens-in-vacuum position that we shouldn’t be regulating any of this, fine. It should be culturally appropriate for young people to own smartphones at the same age when it is culturally appropriate for them to vote, drive, and consume alcohol.
The other things young people need to learn in their teens is logic and critical thinking. I don’t know how long LLMs (Large Learning Models, the form of software popularly called AI at present) will be a thing, because they are, in engineering terms, an inelegant, brute-force approach to the problems they purport to solve. But right now, what I see as being the main end-user problem with LLMs is that people don’t understand, or don’t want to understand, that they need to second-guess the output handed to the users by the LLMs. You need to know math, because the LLMs don’t. You need to know art and history and literature, because the LLMs don’t. All the LLMs are doing when you prompt them is hand you back a response based on what their training tells them is probably acceptable to the User. You need to do at least a gut check if you’re asking them for anything with more serious implications than basic fanfic or fanart, and for that to work, you need to have developed your gut instinct through a well-rounded education that includes analyzing what you’re told for gaps in logic and causality.
And that’s why I see both the internet in general and the AI chatbots in particular as something young people should not be encouraged to seek out below a certain level of maturity and education. There may be cases where there’s a computer programmer in the household, who has trained up his own pet LLM and runs it through Ollama or LM Studio or something, and feels that it can be trusted to play nicely with the children, but like the two generations of super-autonomous farm kids, this situation is an outlier, and the middle class cultural norms “ought” (for certain values of “ought”) to lie elsewhere. And yes, Virginia, there will also be young people who defy the cultural norms. That’s not a reason to do away with the cultural norms altogether. The testing or transgression of norms can be useful to some extent in some cases, either by generating innovation or expanding people’s horizons, but that transgression is essentially parasitical on those norms. Encourage it too heavily and it kills the host norms. Take away the host norms, and transgression dies altogether. No one benefits from either of those scenarios. Hence, norms.
