Qualitative analysis of AI education: School divide

K-12 classrooms are already showing an "AI divide": some districts train teachers and pilot AI tools, while others leave students to learn informally. This post (grounded in NPR's August 19, 2025 reporting on AI4All at Princeton) shows how to run a qualitative analysis of AI education that turns interviews, camp transcripts, and teacher surveys into actionable insight. You'll get a reproducible 7-step workflow to: identify who has access (and who doesn't), code for themes like access, confidence and representation, compare suburban vs rural cohorts, and package findings for policy or product teams. Use Evidano (www.evidano.com) to ingest transcripts, import survey spreadsheets, run thematic and cross-segment analyses, and produce visualizations you can share with stakeholders. Read on to see exact inputs, metrics from the NPR story, and how to map those into an Evidano workflow that researchers and UX teams can deploy in days, not weeks.

Fast take & source

What to know in one paragraph: NPR reported on August 19, 2025 that programs like AI4All at Princeton run three-week summer camps for about 30 low-income high schoolers to teach the math behind generative AI and to broaden participation. The article documents an emerging "AI divide": a Gallup/Walton survey found only 19% of teachers said their school had an AI policy, and suburban, majority-white, low-poverty districts are roughly twice as likely to provide AI training as urban/rural or high-poverty districts. Read the original NPR piece here: www.npr.org/2025/08/19/nx-s1-5503984/ai-summer-camp-schools-education.

Primary keyword: qualitative analysis of AI education, used here to mean methods for coding interviews, comparing cohorts, and extracting stakeholder narratives from mixed text and spreadsheet sources.
Why this matters: disparities in access risk shaping who designs AI and whose problems AI is built to solve, a core concern for policy teams, UX researchers, and qualitative analysts.

Findings snapshot

Metric	Value	Source	Implication
Article published	August 19, 2025	www.npr.org/2025/08/19/nx-s1-5503984/ai-summer-camp-schools-education	Current reporting window for examples and quotes used below
AI4All camp cohort size	≈30 high schoolers	www.npr.org/2025/08/19/nx-s1-5503984/ai-summer-camp-schools-education	Small, purposive sample, treat as demonstrative, not representative
Camp duration	3 weeks (residential)	www.npr.org/2025/08/19/nx-s1-5503984/ai-summer-camp-schools-education	Intensive exposure; good source of in-depth qualitative data (pre/post interviews)
Teachers reporting school AI policy	19%	www.gallup.com/analytics/659819/k-12-teacher-research.aspx	Low institutional guidance, interview protocols should probe teacher self-training
Teacher survey sample	>2, 000 teachers	www.gallup.com/analytics/659819/k-12-teacher-research.aspx	Quant survey complements qualitative interviews for triangulation
Training likelihood (suburban vs others)	Suburban ≈ 2x more likely	www.crpe.org/ai-is-coming-to-u-s-classrooms-but-who-will-benefit/	Segment comparisons are essential (race, poverty, urban/rural)

What happened (nuts & bolts)

NPR followed several AI4All campers and researchers at the Princeton program and found practical, emotional, and equity themes. Students like 16-year-old Esraa Elsharkawy and Anthony Papathanasopoulos reported a shift from skepticism to seeing AI as a problem-solving tool after hands-on lessons with faculty. The camp (run by co-founder Olga Russakovsky) targets low-income high schoolers to diversify who learns foundational AI math and who later participates in building models.

Evidence in the article: direct student quotes, instructor explanations (e.g., drone navigation examples), and contextual surveys (Gallup/Walton).
Research design implications: small purposive samples (camp participants) + large-scale surveys (teachers) create a mixed-data opportunity for qualitative researchers to triangulate narratives and prevalence.

Implications for qualitative analysis of AI education

For UX & product researchers

Compare narratives from students in programs (AI4All) with students who lack access to formal AI instruction. Code for themes such as tool confidence, perceived risks, and use-cases. Use those themes to inform product requirements that are inclusive of underserved contexts.

For policy & education analysts

Qualitative interviews highlight where institutional policy is missing (only 19% schools with an AI policy). Use thematic analysis to surface barriers to teacher training and to craft targeted policy briefs for districts with low AI readiness.

For academic researchers

Treat camps like AI4All as case studies: capture pre/post attitudes, instructor notes, and coded project outputs. Cross-reference with Gallup-style survey measures to estimate how prevalent qualitative findings may be at scale.

Do more, faster with Evidano

Problem: Fragmented inputs (interviews, camp notes, surveys)

Solution: Ingest transcripts, PDFs and survey spreadsheets into Evidano in one workspace. The platform normalizes text and lets you map survey variables to qualitative segments (e.g., urban/rural, income, prior exposure).

Problem: Inconsistent coding across analysts

Solution: Import a codebook or let Evidano propose hierarchical themes. Apply AI-assisted coding to keep double-coding consistent and export inter-coder agreement metrics for validation.

Problem: Need cross-segment comparisons (suburban vs rural)

Solution: Use Evidano's cross-segment analysis to compare theme frequency, co-occurrence networks, and representative quotes between cohorts, ideal for testing the NPR claim that suburban districts are ~2x more likely to train teachers.

Problem: Multilingual or PII-sensitive transcripts

Solution: Evidano offers transcription and translation with custom dictionaries, plus PII redaction and end-to-end encryption. Data is never used to train third‑party models, which supports school/district privacy requirements.

Problem: Stakeholders want visuals and narratives

Solution: One-click visualizations (word clouds, co-occurrence networks, hierarchical code maps) and downloadable reports make it simple to brief teachers, district leaders, or funders about who is being left behind.

7-step workflow: Reproduce NPR-style analysis in Evidano

Follow these steps to turn reporting and raw data into stakeholder-ready findings:

1) Gather inputs: camp interviews/transcripts, instructor notes, student project summaries, and teacher survey CSVs (e.g., Gallup dataset).
2) Import into Evidano: upload transcripts and spreadsheets to a single project workspace; set custom dictionaries for domain terms ("AI4All", "drone navigation").
3) Preprocess & redact: run automatic PII redaction and language normalization; tag metadata (date, cohort, site).
4) Seed a codebook: import an initial codebook (access, confidence, representation, policy-awareness) or ask Evidano to propose themes from the corpus.
5) AI-assisted coding: auto-code at scale, then perform targeted manual review for edge cases and to raise new subcodes.
6) Cross-segment analysis: compare theme frequency and co-occurrence between cohorts (suburban vs rural, camp vs non-camp), export representative quotes.
7) Deliver visuals & memo: generate co-occurrence networks and hierarchical code maps, and export a short policy/product brief for stakeholders.

FAQ: qualitative analysis of AI education

Q: What counts as evidence of an "AI divide" in qualitative data?

A: Look for recurring narrative patterns: lack of curriculum, teacher self-training, student self-directed learning, and differential access to devices or courses. Triangulate with survey prevalence (e.g., the 19% policy stat) to distinguish anecdote from trend.

Q: How do I compare small camp samples (n≈30) to district-level claims?

A: Treat camps as rich case studies. Use them to generate hypotheses and ground truth themes, then test those themes against larger survey data or follow-up interviews across districts.

Q: How do you protect student privacy when using AI tools?

A: Use platforms with PII redaction, role-based access, and explicit non-training guarantees. Evidano, for example, encrypts data and does not use customer data to train third-party models.

Conclusion & next steps

If your next project is to document where AI education is reaching students (and where it isn't) start by combining small qualitative case studies (like AI4All) with teacher surveys and district metadata. That mixed approach surfaces not just prevalence but the why and how behind inequities.

Ready to run this analysis? Create a pilot workspace, upload a sample of transcripts and one teacher survey CSV, and use Evidano to auto-generate themes and cross-segment comparisons in days: www.evidano.com.
For reference and context, revisit the NPR article that inspired this workflow: www.npr.org/2025/08/19/nx-s1-5503984/ai-summer-camp-schools-education.

Strong CTA: Run a 2-week pilot with Evidano to transform interviews and surveys into evidence-backed recommendations for district leaders and product teams. Visit www.evidano.com to get started.

Keep reading

Browse all articles