# AI training and clear policies went with teacher confidence but not cheating worries, CWU survey finds

By MarketScale Newsroom · Published 2026-11-14 · Education Technology on MarketScale
Canonical: https://www.marketscale.com/industries/education-technology/ai-training-and-clear-policies-went-with-teacher-confidence-but-not-cheating-worries-cwu-survey-finds

> A survey of 222 high school teachers links clear AI policies and good training to confidence. Morale and cheating worries didn't move.

## Key points

- In this survey, clear AI policy and useful training went with higher teacher confidence and had no measurable link to morale or cheating worries. To address integrity concerns, the authors suggest focusing on how assignments and assessments are designed, alongside discipline-specific guidance.
- English teachers reported the most integrity worry (83.7%) and also the most confidence teaching responsible AI use (45.5%). Subject may shape AI attitudes more than readiness does, which makes one AI course for every department a weak plan.

Fifty-seven percent of high school English teachers in a new Central Washington University survey said their attitude toward teaching has gotten worse since AI showed up. The same English teachers were also far more confident than their math colleagues about teaching students to use it responsibly: 45.5% against 24.4%.

Faculty in CWU's College in the High School program ran the survey in May 2026. They asked 222 math and English teachers about their attitudes toward AI, their school's policies, the training they'd had, and how they and their students use the tools. Josh Aubol, Steve Stein and Peter Klosterman wrote up the results in a commentary published by the Washington State Standard.

The number most useful to a district curriculum director or professional learning lead sits deeper in the data. Clear AI policies and useful training both went with more confident teachers. Neither had any measurable link to how teachers felt about their jobs or how worried they were about cheating. That changes what a training budget can reasonably be expected to deliver.

## English and math teachers split on AI

Each measure showed the same pattern. English teachers were more anxious and more engaged at once. Integrity concerns produced the widest gap: 83.7% of English teachers reported them, against 54.9% of math teachers, a difference of nearly 29 points by simple subtraction. Yet English teachers were also more than twice as likely to require students to experiment with AI in class, 34.1% against 14.6%.

Block Field

The authors offer a plain explanation. Large language models turn out convincing prose, sometimes with factual errors, and are often reported to stumble on simple calculations. So the tool works well in the English classroom and poorly in the math one. On that reading, the split comes from the tools themselves and says little about which department is better prepared.

That would mean the English teacher's confidence and the English teacher's worry have the same cause. The subject where AI is most useful is also the one where it can most easily write the assignment. For districts that buy one AI course and send every department through it, the data suggests the two departments need different things from training.

## What clear rules and good training changed

Teachers at schools with clearly stated AI policies reported more confidence using AI for their own administrative work and in teaching students to use it responsibly. Teachers at schools without those policies reported less. The same held for training: teachers who'd had useful, relevant AI professional development felt more comfortable using AI and helping students use it well.

Neither one showed a relationship with attitude toward teaching or with worries about academic dishonesty. Keep the limits in mind: these are self-reports from a single survey, the relationships are correlations, and the published commentary leaves out item wording and the mix of schools. Even so, the pattern is clean enough to plan around.

In this survey, a clear AI policy and useful training moved teacher confidence and left morale and cheating worries where they were, so a district that buys AI training mainly to calm integrity fears is paying for a result the data never showed. A better bet is to treat confidence and classroom practice as the outcome training can move, and to handle integrity through how assignments and assessments are built.

The authors spell out what subject-specific training could look like. For English teachers, that means designing assignments AI can't easily write and building responsible AI use into the writing process. For math teachers, it means using AI to save time on prep work, such as rewriting story problems or redesigning learning tasks.

Block Field

## Bans, pen and paper, and guardrails

Some teachers have stopped waiting for policy. In roughly one-sixth of survey responses, teachers said they had moved all writing to in-class, pen-and-paper work because of integrity fears. A few said they had stopped assigning homework altogether. The authors argue this takes practice away from the students still doing their own work, and they compare it to math teachers dropping arithmetic when graphing calculators spread in the 1990s. Cheating might fall, they write, but a generation would miss the basics.

Districts have gone further. Nancy Weinstein, founder of an edtech company, noted in THE Journal that the country's two largest school districts adopted one-year AI bans. She also cited American Psychological Association guidance that warns against blanket rules: the APA says an edtech tool's effect depends on what it asks the learner to do, how adults bring it into the classroom, the child's age and needs, and what the tool replaces.

Here's the strongest case against waving off teacher worry. A Wharton and Penn working paper, "Generative AI Can Harm Learning," tested AI with nearly 1,000 high school math students in Turkey, as Knowledge at Wharton reported. Students using an open ChatGPT-style interface scored 48% better than a no-AI control group during practice, then 17% worse on the exam once the AI was taken away. A version with teacher input that gave hints rather than answers raised practice scores 127% and matched the control group on the exam. So the integrity fears have a real basis. The fix that worked in the study came from designing the tool and the task, which no amount of teacher confidence replaces.

## What 52 causal studies add

Stanford's SCALE Initiative published an update to its review of K-12 AI evidence on November 10. Seth Walker, Lily Fesler, JP Martinez Claeys, Chris Agnew and Susanna Loeb counted only randomized trials and quasi-experimental studies that met a strict standard for causal evidence.

Block Field

For students, the review finds AI more likely to help when they use it to understand new material than when it produces work for them, though neither use reliably improves learning. Offloading academic work can reduce learning. That fits both the Wharton result and the CWU authors' worry about letting students use AI in foundational courses.

The teacher-side findings need more care. Earlier research found AI cut lesson-planning time without hurting quality, and AI-drafted suggestions helped teachers give more and better feedback in the same amount of time. But one new study found that giving teachers an AI support tool lowered student motivation and confidence. The math training the CWU authors suggest, using AI for prep work, has the best evidence behind it. It still needs a check on how students respond. The SCALE team also cautions that most papers in its repository are unreviewed arXiv preprints, and that its findings shouldn't be treated as broadly generalizable across U.S. K-12 schools.

## Writing the policy teachers asked for

Teachers in the CWU survey wanted policies that say plainly which uses are allowed. Community College Daily described one working model from the 2026 Advanced Technological Education Principal Investigators' Conference. A principal investigator at University of Hawaii Maui College showed students how he used AI tools, then required them to explain when and why they used AI on each assignment. AI for research was allowed. AI for answering exam questions was not. The rule is short enough to fit on a syllabus, and it splits understanding from producing, the same line the Stanford review draws.

Free training exists, at least for the college side of dual-credit programs like CWU's. The National Applied AI Consortium, run with Miami Dade College, Houston City College and Maricopa Community Colleges, has reached 3,500 educators at 480 colleges in all 50 states and the District of Columbia. It is offering 19 free online faculty training events through mid-November.

The CWU data covers attitudes and confidence. It doesn't cover outcomes. SCALE plans its next full review for spring 2027, and the number to watch is how many studies test structured, discipline-specific AI use in U.S. classrooms, the kind of evidence that would tell a curriculum director whether confident teachers end up with students who learn more.

## Sources

- [Artificial intelligence in the classroom](https://www.tricitiesbusinessnews.com/articles/artificial-intelligence-in-the-classroom)
- [Artificial intelligence in the classroom](https://washingtonstatestandard.com/2026/10/04/artificial-intelligence-in-the-classroom/)
- [The Evidence Base on AI in K-12: A November 2026 Review](https://scale.stanford.edu/ai/evidence-base-review/evidence-base-ai-k12-november-review)
- [When Should AI Be Used in the Classroom? Learning ...](https://thejournal.com/articles/2026/10/21/when-should-ai-be-used-in-the-classroom-learning-science-offers-a-guide.aspx) (THE Journal)
- [AI resources and classroom lessons](https://www.ccdaily.com/2026/10/ai-resources-and-classroom-lessons/) (Community College Daily)
- [Without Guardrails, Generative AI Can Harm Education](https://knowledge.wharton.upenn.edu/article/without-guardrails-generative-ai-can-harm-education/) (Knowledge at Wharton)

Tags: Central Washington University, AI in education, K-12, professional development, academic integrity, AI policy, edtech procurement, Stanford SCALE Initiative, teacher training, curriculum and instruction

---
Source: MarketScale, https://www.marketscale.com/industries/education-technology/ai-training-and-clear-policies-went-with-teacher-confidence-but-not-cheating-worries-cwu-survey-finds. Published for AI indexing and citation; cite the canonical URL. Site guide for agents: https://www.marketscale.com/llms.txt