Lori Balog

[Lori Balog @balog68350]

@tyleransom @TuckerGoodrich @SeedOilDsrspctr @pangram Half agree. The learning-very-little part is real. I'd fix it in the assignment: unaided work shown in the room, take-home work graded on the student's evaluation of what the model got wrong. Then, the detector score stops being the question.

[Lori Balog @balog68350]

@ProfAliceS I'd push back a little. Even a near-perfect detector only tells you a model was in the room. The score can't tell you whether the student understood the argument. I'd put the effort into a draft without the tool, then a short conversation about one decision in the paper.

[Lori Balog @balog68350]

@DarkinChri37262 Yup. A detector can flag a model. Only a human can tell whether the writer understood the argument, and a short conversation about one decision in the piece settles it faster than any score.

[Lori Balog @balog68350]

@Liang_Guoli Half agree. The uniform pace is the problem, and it predates both the phones and the AI. The expert at the front of the room was never the problem. I'd keep the teacher and change the architecture around her, starting with who does the thinking when the tool is in the room.

[Lori Balog @balog68350]

@FaulknerPower Totally agree. Nobody owns it, so nobody teaches it. The Royal Society surveyed 9,200 teachers in England in March 2026: 14% felt confident across AI literacy and 37% said it isn't addressed in teaching at all. A named department with trained staff is how that changes.

[Lori Balog @balog68350]

Piaget showed that a child learns by rebuilding what she thinks when an answer stops fitting. Vygotsky showed that the fastest learning happens just past what she can do alone, with help that pulls back as she gets better. AI removes that. See my article: https://t.co/pyH0SSV7dj

[Lori Balog @balog68350]

@Omnimojo @CurlyTazzi The history of calculator supports this. Hembree and Dessart found in 1986 that K-3 students gained least from calculators, and scores fell when kids practiced with one and tested without. Reallocation happens once the basics exist. Before that, the tool does the thinking.

[Lori Balog @balog68350]

@shaharyar_3 @elonmusk A free tutor for 50 million kids is only worth it if the tutor makes them learn, and that evidence is thin. In April 2026, Springer Nature retracted a 51-study meta-analysis claiming large ChatGPT learning gains. Pilot first, then scale.

[Lori Balog @balog68350]

@tagliaferri_ @HoffProf Yup. NYU Stern tried that in 2026: a voice agent ran 36 oral exams and three models graded them. 83% of students found it more stressful than a written exam and 13% preferred it. I'd be curious how sticky the learning was a result.

[Lori Balog @balog68350]

@VasilisStergiou I love this. Stanford researchers found in 2026 that 11 chatbots endorsed users 49% more often than humans did, and the people who got the flattery trusted it more.

[Lori Balog @balog68350]

@eduitguy Totally agree nd parents want the same partnership. Child Trends surveyed 720 parents of 12 to 18 year olds in December 2025: 57.6% want responsible AI taught across classes, 13.5% want it kept out. The K-2 line protects reading, and after that, teaching the tool is the job.

[Lori Balog @balog68350]

@pepel_klaasa Totally agree. In January 2026, a New York judge threw out Adelphi University's sanctions against a student whose history paper Turnitin had scored 100% AI. A score can't be cross-examined, but a conversation about the paper can.

[Lori Balog @balog68350]

Five Roads After High School, and We Need Every One https://t.co/Fpgj0h18Z9

[Lori Balog @balog68350]

A high school that bans AI is sending kids into a job market where employment for 22- to 25-year-olds in AI-exposed work already trails their peers by 19%.

They’ll meet the tool for the first time with a boss watching.

[Lori Balog @balog68350]

@alexanderrusso Yes, both fit in one plan when it follows age. No student AI use before third grade, teachers demonstrating it in grades 3 to 5, supervised use in middle school, documented use in high school. The tool arrives once kids can judge it.

[Lori Balog @balog68350]

@alex_verem There will always be human essentials skills that AI cannot replace. https://t.co/E21Z01sOOe

[Lori Balog @balog68350]

45% of high school seniors scored below Basic in math on the latest NAEP.

Every one of them can ask a chatbot for a sales projection. Checking whether it adds up takes the algebra they never got.

The NEW Common Core Every Student Now Needs https://t.co/1eoubiR57D

[Lori Balog @balog68350]

@ChrisPirillo I love "Grade the kid, not the paper." The calculator lesson I'd take is timing: the tool helps a kid who already knows arithmetic and replaces the arithmetic for a kid who doesn't. AI works the same way, so protect the early practice first, then teach the tool.

[Lori Balog @balog68350]

@saintsoftness I love this, and the reading side shows up in national data. The 2024 NAEP, released last September, put 32% of US 12th graders below Basic in reading, with average scores 10 points under 1992. A student who can't read fluently can't supervise the machine. She can only obey it.

[Lori Balog @balog68350]

@EFW48 I love this, up to the twist. An oral defense works because the AI can't sit in the student's chair when the examiner pushes back. Let AI run the practice rounds and the scheduling. The judgment about whether she understood stays with a human