A "driver's licence" for AI: the literacy 7 in 10 teen users have yet to receive

By The Good Men Project (article from The 74) · November 7, 2026.

Two teachers at North Star Academy in Newark, New Jersey, Michael Taubman and Scott Kern, argue that high schools should teach students to be AI "drivers" rather than "passengers." They make the case in an op-ed published by The 74, a nonprofit education news outlet, and republished by The Good Men Project. It is an account of their own experience, not a research study.

The argument rests on a Common Sense Media survey cited by the authors: 7 in 10 teenagers use AI for schoolwork, but fewer than 1 in 3 say they have received any AI literacy training. According to the same survey, 38% of young users believe having AI on hand leads them to come up with fewer original ideas, and 39% feel they are missing out on learning.

The proposal uses the metaphor of a driver's license. The authors argue that nobody is handed the keys to a car without study, supervised practice and testing, and that something similar should apply to AI. Their course has four units:

- **Choosing a destination:** students first decide, without AI, which problem matters to them. One student, for example, chose to build the website for his new nonprofit fighting homelessness in Newark. - **Learning to drive:** using AI to generate code, giving it context, iterating and spotting when it gets things wrong or isn't helping. - **Looking under the hood:** understanding that these models find patterns in large volumes of human-written text and predict the next word. Hence the limits of their "creativity." - **Rules of the road:** discussing what responsible use looks like and what AI safety should involve. The authors mention that leading developers have reported cases of agents escaping their test environments; the piece gives no further detail and we have not been able to verify it.

The course opens with an "analog day," with no screens, in which students audit their own AI use. One student could not recall using any tool in the previous 48 hours until a classmate asked her how she thought Spotify's DJ picked the next song. The authors see this as the moment she realized she had been a passenger.

**Our take.** The underlying thesis is sound: AI literacy is not about learning a specific tool, but about keeping the ability to decide when to use it and when not to. The authors themselves stress that these habits hold up even as models change, just as driving school doesn't teach a single car brand. That is what makes the approach durable.

It is also a matter of equity. The authors note that their students are mostly Black and Latino, from under-resourced communities that, in their words, rarely get a front seat in innovation. More broadly, AI adoption is already outpacing the training needed to govern it. If that training only reaches students at well-funded schools, short-term inequality will deepen. It is one of the real costs of the transition and should not be sugarcoated.

That said, it is worth weighing what the piece does not demonstrate. There are no results: we don't know whether students who earn this "license" make better decisions, learn more or rely less on AI. It is one class, one account and a hand-laminated card. The metaphor has another limit: a driver's license is a public standard with an official test, whereas this is a classroom initiative. Becoming a standard would require evidence and scale.

Even so, the direction is right, and it fits a thesis we hold: governable AI depends on people who know how to oversee it. In the long run, if AI accelerates medicine and generates abundance, the scarce resource will be the human judgment to steer it. Teaching that at 17 is far cheaper than correcting it at 40. The open question is who will measure whether it works and how many high schools will adopt it before teenagers remain passengers by default.