Alpha School, a pricey private school with campuses nationwide, uses artificial intelligence to instruct students. AI schools have been praised by the Trump administration, but researchers say there's limited evidence the model works.
“AI should serve as a scaffold for cognitive construction rather than a substitute.”
“…the teacher’s role is shifting from knowledge transmission to instructional design and behavioral facilitation… Teachers must develop digital literacy and data fluency while acting as safeguards against over‑automation, ensuring that human judgment and educational values mediate AI adoption.”
“…while AI offers efficiency and feedback advantages, traditional teaching remains essential for tasks requiring cultural interpretation, discourse depth, and emotional connection. A blended model—AI for repetitive or procedural tasks and teachers for critical discourse—appears most effective.”
This study explicitly does not advocate for replacing teachers with AI, and repeatedly cautions against doing so
And the school that is opening will still have human “guides” so I’m curious how it will work out. I agree it should be a mix of AI and human, and not fully AI.
These findings highlight both the promise and the limitations of AI in language education, underscoring the importance of teacher facilitation and thoughtful design of human–AI interaction to support deep and sustainable learning.
The problem is there’s no teachers in this scenario, at least that’s my understanding
You’re right, they will have “guides” instead of teachers. This might be to far, but we won’t know until they try it. A mix of human and AI teachers would probably be best.
Why are you hounding them for the data? They would swear on their honor that Grok said it, and that’s somehow not enough for you. They even asked a follow-up “Are you sure?”, to which Grok reaffirmed its findings. Maybe you should be practicing law if you want to act like you care so much about “evidence”.
This is for college students (aka students educated enough to learn on their own already), reads like a promotion for AI, has a limited sample size and does not translate to school kids at all and from the study itself:
Finally, the study’s limitations include its single-institution sample, short duration, and reliance on proxy behavioral indicators. Ethical concerns around informed consent, data privacy, and AI dependency also warrant closer attention. Future research should pursue longer-term and cross-institutional designs, employ multimodal behavioral measures, and develop governance frameworks that align technical gains with equity, autonomy, and critical capacity.
This “”study”” seems to spend more time opining on AI learning frameworks than actually measuring scores on standardised testing and only dedicates a minimal amount of the paper to the results. It also states in paper that higher achieving college students saw less benefits (poorer performing student, AI can bump your grades enough to be noticeable for a unit/pass an exam).
Did you read this study or google something in order to provide a study? This study does not support the claim that “these kids will perform traditional learning by miles”.
It’s also for learning English, which is something a large language model is probably the most suitable for. It’s not going to be much use teaching music or drama.
You said the data says otherwise which you then used to support that opinion. The data doesn’t say otherwise.
Want me to pull out of study from 20 years ago with decades of proven data?
Almost like that was in my original comment that you then replied to with a study as if it were compelling, so spare me the sassy comment. Don’t claim the data says otherwise when it doesn’t if you don’t want to be called out on it.
Really? Because the data I’ve seen says the exact opposite and that Gen Z is the first generation of people dumber than the generation before them. These kids are already fucked and AI is going to make it even worse.
The New York Post (NY Post), founded as the New York Evening Post (originally New-York Evening Post), is an American conservative[3] daily tabloid newspaper published in New York City. Th
The Post has been criticized since the beginning of Murdoch’s ownership for sensationalism, blatant advocacy, and conservative bias. In 1980, the Columbia Journalism Review stated that the “New York Post is no longer merely a journalistic problem. It is a social problem—a force for evil.”[9] The Post has been accused of contorting its news coverage to suit Murdoch’s business needs, in particular avoiding subjects which could be unflattering to the government of the People’s Republic of China, where Murdoch has invested heavily in satellite television.[63]
On October 14, 2020, three weeks before the 2020 United States presidential election, the Post published a front-page story purporting to reveal “smoking gun” emails recovered from a laptop abandoned by Hunter Biden at a computer repair store in Wilmington, Delaware.[105] The only sources named in the story were Trump personal attorney Rudy Giuliani and strategy advisor Steve Bannon.[105] The story came under heavy criticism from other news sources and anonymous reporters at the Post itself for “flimsy” reporting, including questions about the reliability of its sourcing and the lack of outreach to either Hunter Biden or the Joe Biden campaign for pre-publication comment.[106][107]
Right wing tabloid, you fell for republican propaganda.
That was just the first news link I grabbed about the topic. You want a different news site? There’s plenty. Or maybe a link directly to the study that’s also in this article?
That generation is fucked yes. There is no fixing that with AI. This is for young gen A or gen beta. Green Z is already too old for this to be useful for them.
The only research I’ve seen on using LLMs in a school setting found that the kids that were given access to an LLM performed a bit better on exercises that those without. At the same time their experienced learning was a lot better. When they finally got a test assignment, the kids that had been using LLMs during exercises flopped and performed significantly worse than those that hadn’t.
That’s not what the data says. These kids are going to outpace traditional learning kids by miles.
Is this data in the room with us?
https://www.mdpi.com/2078-2489/16/10/895
This study explicitly does not advocate for replacing teachers with AI, and repeatedly cautions against doing so
You have to excuse them, they used AI to summarize it.
Ironically… so did I 🙃 But I hand-verified everything it said, and adjusted the quotes.
And the school that is opening will still have human “guides” so I’m curious how it will work out. I agree it should be a mix of AI and human, and not fully AI.
The problem is there’s no teachers in this scenario, at least that’s my understanding
You’re right, they will have “guides” instead of teachers. This might be to far, but we won’t know until they try it. A mix of human and AI teachers would probably be best.
AI hasn’t even been around long enough for any meaningful data to be collected surely. Also, post this “data” you’ve twice now claimed exists.
Why are you hounding them for the data? They would swear on their honor that Grok said it, and that’s somehow not enough for you. They even asked a follow-up “Are you sure?”, to which Grok reaffirmed its findings. Maybe you should be practicing law if you want to act like you care so much about “evidence”.
https://www.mdpi.com/2078-2489/16/10/895
This is for college students (aka students educated enough to learn on their own already), reads like a promotion for AI, has a limited sample size and does not translate to school kids at all and from the study itself:
This “”study”” seems to spend more time opining on AI learning frameworks than actually measuring scores on standardised testing and only dedicates a minimal amount of the paper to the results. It also states in paper that higher achieving college students saw less benefits (poorer performing student, AI can bump your grades enough to be noticeable for a unit/pass an exam).
Did you read this study or google something in order to provide a study? This study does not support the claim that “these kids will perform traditional learning by miles”.
It’s also for learning English, which is something a large language model is probably the most suitable for. It’s not going to be much use teaching music or drama.
No, the end part was my own opinion. I do believe classrooms that embrace AI will outperform tradition learning classrooms by a mile.
Already yes the study is limited, AI learning is very new. Want me to pull out of study from 20 years ago with decades of proven data?
You said the data says otherwise which you then used to support that opinion. The data doesn’t say otherwise.
Almost like that was in my original comment that you then replied to with a study as if it were compelling, so spare me the sassy comment. Don’t claim the data says otherwise when it doesn’t if you don’t want to be called out on it.
Really? Because the data I’ve seen says the exact opposite and that Gen Z is the first generation of people dumber than the generation before them. These kids are already fucked and AI is going to make it even worse.
Do you have a citation for this?
https://nypost.com/2026/02/07/us-news/gen-z-the-first-generation-officially-dubbed-dumber-than-the-last/
https://en.wikipedia.org/wiki/New_York_Post
Right wing tabloid, you fell for republican propaganda.
That was just the first news link I grabbed about the topic. You want a different news site? There’s plenty. Or maybe a link directly to the study that’s also in this article?
Sure, anything is better than a tabloid that supports Trump.
That generation is fucked yes. There is no fixing that with AI. This is for young gen A or gen beta. Green Z is already too old for this to be useful for them.
The only research I’ve seen on using LLMs in a school setting found that the kids that were given access to an LLM performed a bit better on exercises that those without. At the same time their experienced learning was a lot better. When they finally got a test assignment, the kids that had been using LLMs during exercises flopped and performed significantly worse than those that hadn’t.
https://www.mdpi.com/2078-2489/16/10/895