Jeffrey Nordine and David Fortus published a paper in the 2026 Journal of Research in Science Teaching arguing that the Next Generation Science Standards, while a genuine improvement over what came before, are still built around the wrong organizing principle. The NGSS groups standards by disciplinary content ideas, specifically physical science, life science, and earth science, and expects students to demonstrate competence within those categories. Nordine and Fortus argue this approach is out of step with what learning theory actually supports. Using constructivism, motivation research, and situated cognition, they make the case that standards organized around contemporary issues and real-world contexts, rather than disciplinary buckets, are more likely to produce the kind of informed student agency that society and the workforce actually need.
This is not an argument against scientific rigor. It's an argument about sequence and structure. A student who encounters climate systems, energy policy, or disease transmission as the organizing frame, then learns the disciplinary content in service of understanding those contexts, builds knowledge differently than a student marching through content categories. The research is clear that contextualized learning sticks. The standards, by contrast, are still largely organized as if the content itself is the destination.
At the start of class, give students 90 seconds to write one sentence: the most important thing they learned last week, stated as a claim, not a summary. No notes, no devices. Then ask two or three students to read theirs aloud and defend it with one reason. The exercise takes four minutes, reveals immediately who processed the material versus who memorized it, and gives you usable information before you've taught a word of today's lesson. Any subject, any grade.
Stanford's SCALE Initiative published "The Evidence Base on AI in K-12: A 2026 Review," analyzing more than 800 academic papers on AI and education, with a repository now exceeding 1,100 studies. The finding that matters most: only 20 of those studies meet the standard for causal inference, meaning they can actually demonstrate that an AI tool caused an improvement in student outcomes, rather than merely correlating with one. A secondary finding is equally important: some AI tools produce measurable performance improvements while in use, but those gains fade when the tool is removed. The research does not conclude that AI tools are worthless. It concludes that the evidence for what they actually do is far thinner than the adoption decisions being made in their name.
This is the citation you need the next time a vendor, an administrator, or a colleague makes a confident claim about what an AI tool does for student learning. The honest answer, as of 2026, is that we don't have rigorous causal evidence for most of those claims. That doesn't mean the tools are bad; it means the burden of proof hasn't been met. The "gains fade when the tool is removed" finding is more consequential for classroom practice. It suggests that AI tools used as shortcuts, rather than as scaffolds toward independent competence, may be producing performance without producing learning. The distinction matters if your goal is to prepare students who can think without a crutch.
New York City Public Schools released preliminary AI guidance in March 2026 for its 78,000 teachers and 1.1 million students, structured as a "red light, green light" framework. Red-light uses, prohibited outright, include assigning grades, making promotion or disciplinary decisions, developing special education plans or IEPs, and using individual student data to train AI models. Green-light uses, explicitly permitted, include lesson planning, translating communications, organizing information, and drafting family correspondence. The guidance explicitly prohibits AI from making consequential decisions about students. A more comprehensive playbook, including grade-band-specific guidance and criteria for evaluating algorithmic bias and instructional effectiveness, is expected from the DOE this month. Source: GovTech, Gothamist, The 74, March 2026.
NYC's guidance is the most specific binding AI policy in any major American school system. Its categorical prohibitions on AI for grading, discipline, and IEPs are not recommendations; they are rules, and they draw a clean line between AI as an instructional support tool and AI as a decision-maker for student outcomes. That line is the right one. Even if your district has no formal policy, the NYC framework gives you a defensible position: use AI to prepare for instruction, not to assess its results. Teachers who make that distinction clearly, and who can articulate why, are ahead of the policy conversations coming to every district in the next 12 months.
The 2026 World Happiness Report, produced by the UN Sustainable Development Solutions Network, dedicated its central chapter to social media and adolescent well-being. The core finding: social media harm is not merely individual; it is operating at a scale large enough to produce measurable declines in population-level mental health indicators. The decline is sharpest in English-speaking countries: the United States, Canada, Australia, and New Zealand are all showing significant drops in well-being among people under 25. The harm is not equally distributed. Adolescents from lower-socioeconomic-status households are more harmed by problematic social media use than their higher-SES peers, particularly in life evaluation and psychological complaints. Girls are more affected than boys across most outcome measures. The report draws on surveys, longitudinal studies, social media reduction experiments, and corporate documents.
The significance here is not the finding itself; teachers already know their students are struggling. The framing is what matters. Population-level harm means the problem is not a set of struggling individual students who need intervention. It is a structural condition affecting the cohort as a whole. A classroom of high school juniors in the United States in 2026 contains students whose developmental environment has been meaningfully different from any previous generation's, and the consequences are showing up in aggregate data, not just in the students you've identified as needing support. That changes what "normal" means for this group, and it changes the baseline expectations you can reasonably hold for attention, emotional regulation, and academic persistence.