Learning Signal Brief #2: AI Planning, Assessment Questions and Real-World Pathways
This week's signal: Districts are deciding how far to let AI into classrooms with little settled evidence, and a new review of 79 studies shows teacher planning is one of the best-studied uses. The design items add a common test: start with the learning goal, then ask what AI changes about the evidence of learning. The entrepreneurship and career items show schools building real-world pathways, and the CTE report finds teachers paying their own way to keep those pathways current.
8 articles, 2 short takes and a quote of the week. About 7 minutes to read.
In this issue
AI and Agentic AI in Teaching and Learning
A 79-study review finds teachers mostly use GenAI to plan
Boston builds an AI curriculum while New York pauses student AI
A K-12 writer urges districts to ask why students confide in AI
Instructional Design and Curriculum
Teachers say to start with the learning goal, then test AI
A FullScale blog asks what AI changes about assessment evidence
Two authors say the response to AI should be pedagogical
Innovation and Entrepreneurship
Podcast hosts visit a student entrepreneurship summit in Chicago
CTE teachers pay their own way to keep up with changing industries
AI and Agentic AI in Teaching and Learning
1. A 79-study review finds teachers mostly use GenAI to plan
“Teachers' use of generative artificial intelligence in K–12 education: a systematic review,” by Oğuz Cincioğlu, Duygu Mutlu Bayraktar and Tuğba Altan, Frontiers in Education, October 2, 2026
A team of researchers synthesized 79 empirical studies of teachers' use of generative AI in K-12 education, following PRISMA 2020 guidelines. Asia produced 44.4% of the publications and North America 27.16%, with the United States the most productive single country.
Instructional design and co-planning were among the most prominent roles. The authors conclude that GenAI is mainly a collaborative tool for teachers' preparation, “rather than replacing instructional roles.” Qualitative and mixed-methods designs dominated, and language education and STEM were the most common domains.
Vocational, physical, arts and early childhood education remain underrepresented, and administrative uses have received limited attention. This is a map of research trends, and the abstract reports no measure of student learning.
Why it matters: For leaders: Treat teacher-side planning as one of the most-studied uses of GenAI, and ask vendors for student outcome evidence before paying for more.
2. Boston builds an AI curriculum while New York pauses student AI
“Boston and New York City Take Opposite Approaches to AI in School,” by Hadley Barndollar, masslive.com, republished by GovTech, October 2, 2026
Boston Public Schools is working with UMass Boston to design an applied AI high school elective, backed by a $1 million public-private partnership for grades 7-12. Twenty-five teachers are piloting AI learning projects this year, and the district is developing AI literacy benchmarks for grades 7-12.
New York City has barred student-facing generative AI tools and companion-style chatbots in elementary and middle schools for one school year, and is disabling AI in 38 core curriculum products. High schoolers will pilot five vetted programs for a maximum of 45 minutes a week under teacher supervision, and none are chatbots.
MIT's Justin Reich speculates that K-8 classrooms in both cities would show little AI use in practice, and says “no one knows what a good AI policy is.” This is a news report built on officials and one expert, with no learning results.
Why it matters: For leaders: Pair any student AI rollout or pause with benchmarks and an evaluation plan, as Reich urges.
3. A K-12 writer urges districts to ask why students confide in AI
“The Hidden Side of Student AI Use: Mental Health, Access and Risk,” by Rachelle Dené Poth, EdTech Magazine K-12, September 28, 2026
Poth writes that school conversations about student AI use often begin with assignments, and argues educators should also watch for chatbot use for advice and emotional support. The column cites a 2026 JAMA Pediatrics survey that found around 1 in 5 U.S. youth ages 12 to 21 had used an AI chatbot for mental health advice, and nearly two-thirds had told no one.
Districts should first learn how students actually use these tools, Poth says, describing anonymous surveys, student focus groups and panel discussions. District guidance should define boundaries for emotional support use, name personal information students should never enter, and explain when to stop and contact a human.
Poth also urges lessons that let students practice evaluating emotionally persuasive responses. Kristen Mattson of the University of Illinois cautions that “just saying ‘no’ glosses over the root issues causing young people to turn to AI in the first place.” This is an opinion column, and the survey figures come from outside organizations.
Why it matters: For leaders: Survey students anonymously about why they turn to chatbots before you write or revise district AI guidance.
Instructional Design and Curriculum
4. Teachers say to start with the learning goal, then test AI
“How AI Can Remove Barriers and Help Meet Instructional Goals,” by Julia Gilban-Cohen, GovTech, September 30, 2026, reporting on a Future of Education Technology Conference webinar
Lyndhurst High School English teacher Tim Belmont described using Gemini to rethink a reading lesson on Roald Dahl's story Poison. The result was a jigsaw activity in which students read portions of the background material, discussed them in groups and wrote headlines. In a podcasting class he uploaded his lesson materials to Gemini Canvas, turned them into an interactive review game and adjusted it.
Belmont's rule is to start with the learning goal: “Just using AI because it's cool is not necessarily the best reason to use AI.” Kelli Suding of the Center for Applied Special Technology framed the same idea through Universal Design for Learning, noting that tools such as text-to-speech can give students other ways to access material.
Suding cautioned that adding more tools does not guarantee better learning when implementation and training are overlooked. This is a report on two panelists' examples from one webinar, with no student results.
Why it matters: For teachers: Name the learning goal and the barrier first, then check whether an AI tool removes it before adding it to a lesson.
5. A FullScale blog asks what AI changes about assessment evidence
“As AI Shapes Assessment, What Should We Pay Attention To?” by Jazminn Williams, FullScale, October 2, 2026 (surfaced through the Aurora Institute feed)
Williams reports on the 51st annual conference of the International Association for Educational Assessment in Toronto, where the theme was trust, transparency and technology. FullScale works with K-12 networks exploring competency-based, personalized learning, and the post draws three issues from the sessions.
First, a student's use of AI during a performance does not automatically invalidate it, but it complicates the claim about what the student knows. Second, new assessment technology can introduce new sources of inequity, such as reliance on high-speed connectivity. Third, people need AI and assessment literacy to understand where AI was involved, for example in scoring or feedback.
The guiding questions are what is being measured, who gets meaningful chances to show what they know, and whether users understand what the evidence can tell them. This is a conference reflection with no data, and FullScale's AI tool selection guide is not yet released.
Why it matters: For leaders: Write down what each assessment claims to measure and how student AI use changes what its scores can support.
6. Two authors say the response to AI should be pedagogical
“Education in the age of AI: Toward a pedagogy of voice,” by Shane Safir and Sawsan Jaber, Learning Forward, September 21, 2026
Safir and Jaber argue that the most urgent response to AI is “not technical, it is pedagogical.” They draw on their book Pedagogies of Voice, which offers a framework for designing learning around student agency, and ask when AI extends student voice and when it flattens or replaces it.
They describe gathering what they call street data: empathy interviews, close study of student work and shadowing a student through the school day. One question they recommend for interviews and learning walks is “What do you want me to understand about your experience?”
Their Monday-morning shifts include a student circle on hopes and fears for the year, a Wonder Wall of student questions, and making learning public through presentations or an exhibition. This is an opinion post by the book's authors, who are presenting at Learning Forward's annual conference, and it offers no evidence of effects.
Why it matters: For teachers: Try one routine from the post, such as a Wonder Wall of student questions, for a single unit and see how it changes the work students hand in.
Innovation and Entrepreneurship
7. Podcast hosts visit a student entrepreneurship summit in Chicago
“Live at the Uncharted Learning Summit: Why Entrepreneurship Is for Everyone,” Getting Smart Podcast, hosts Victoria Andrews and Shawnee Caruthers, October 2, 2026
Getting Smart podcast hosts recorded at the Uncharted Learning summit in Chicago, talking with students, alumni, teachers, a program director and a volunteer. Uncharted runs middle and high school entrepreneurship courses, and students begin by listing what bugs them in their community before choosing what to make.
One alumnus describes pitching to venture capitalists, and another describes learning to delegate and lead a team. Teacher Jason Roemer says students went from a 30-second elevator pitch to a 10-minute pitch in one school year, and a first pitch contest drew close to 400 people. Alumnus Connor Boundy says about a quarter million students have gone through the program.
The hosts say the course has reinvigorated some teachers, including one who had resigned and then took the post. This is a promotional podcast built on alumni, teachers and supporters of one organization, with no outcome data or comparison group, and one host says “I couldn’t be a bigger advocate.”
Why it matters: For leaders: Open an entrepreneurship course with a list of community problems students notice, and recruit local volunteers as coaches and pitch judges.
8. CTE teachers pay their own way to keep up with changing industries
“Skills Change Fast in CTE. Can Teachers Keep Up?” by Sarah D. Sparks, Education Week, September 15, 2026 (also featured by the Carnegie Foundation)
Education Week reports that career and technical education teachers often keep up with fast-changing fields on their own time and money. David Hope, who leads firefighting and EMT instruction at South Tech High School in Missouri, and fellow Missouri instructors pay about $500 out of pocket and work 20-40 hours outside school each year to stay current. Hope is also earning drone pilot and instructor licenses to start in-person drone classes.
Computer numerical control, the use of coding and software to produce complex prototypes, has driven many new school makerspace CTE programs, according to Kentucky CTE teacher Nichols. But equipment turns over fast, Nichols says: “the 3-D printer that is all the rage today, in six months, is falling behind.”
Kenosha, Wisconsin, runs a three-year mentoring sequence for new CTE teachers, ending with connecting classroom experiences to students' work-based learning. In San Diego's Cajon Valley district, students at the career-tech high school spend up to 40% of class time in off-campus mentorships and apprenticeships. This is a news feature built on officials and teachers, with no results on student outcomes.
Why it matters: For leaders: Budget for industry training and multi-year mentoring for CTE teachers, not only for hiring and certification.
Short takes
Carnegie Mellon Robotics Academy has released a preview of Robot Coding Fundamentals with LEGO CS & AI (6-8), its first curriculum for LEGO's new CS & AI kit, as educators weigh options after the announced SPIKE Prime retirement. The course uses “Try It” demonstrations, challenges and unit quizzes, and it stays in early access while the academy collects educator feedback. Paige Branagan, Carnegie Mellon Robotics Academy. This is the developer announcing its own course, with no evidence of student results. Read the original →
Petaluma City Schools in California is drafting a grade-band AI plan: no AI in grades K-2, a basic introduction without student-facing tools in grades 3-5, responsible use and ethics in closed school environments in grades 6-8, and guided use of tools such as Canva and possibly Gemini in grades 9-12 at teacher discretion. Director of technology Heather Ramme called it “the committee’s working stance, not yet adopted policy.” Amelia Richardson, Petaluma Argus-Courier, reprinted by GovTech. This is a draft with no board vote scheduled and no evidence on how it works. Read the original →
Quote of the week
Justin Reich directs the Teaching Systems Lab at the Massachusetts Institute of Technology. He was commenting on how districts across the country are handling AI in schools, and said schools must be just as devoted to evaluating what they do as to developing it.
“Every district in the country is conducting an experiment.”
— Justin Reich, in GovTech's October 2 reprint of a masslive.com report on Boston and New York City's approaches to AI in school
The Learning Signal Brief summarizes the week's most useful articles on innovation in K-12 education, chosen from more than 100 sources. Summaries are original, quotes are short and attributed, and every item links to the original. Compiled with AI assistance and reviewed by Ryan Keser.
Comments