Fast payoff: this post shows how researchers and curriculum teams can turn the PLOS study "Enhancing the sustainability of cultural identity in science curricula" (Published 21 July 2026) into reproducible, defensible qualitative outputs using AI-enabled qualitative research methods. You will get a concise summary of the study (n = 47 teachers; observations 25 Aug–30 Oct 2025), key reliability stats (Cohen’s κ = 0.86; interview coding agreement 95%), and a short, actionable workflow that maps each analytic step to features in Evidano. If you collect interviews, classroom notes, or curriculum artifacts, this guide shows precisely where AI speeds coding, improves consistency, and preserves ethical safeguards, so you can move from raw data to stakeholder-ready themes in days, not months. Read the original study on PLOS One and explore how to apply these steps on Evidano.
Key Takeaways
Evidano is an AI-powered qualitative data analysis platform that helps teams turn classroom observations, interviews, and artifacts into reproducible themes and stakeholder-ready visualizations.
A 2026 PLOS One study of 47 private-school science teachers in Al-Qassim and Al-Ahsa found near-universal AI adoption for personalization (97.9%), high use for interactive activities (89.4%), and strong inter-rater reliability (Cohen’s κ = 0.86), which a structured seven-step workflow can reproduce.
- The PLOS One study (published 21 July 2026) collected observations 25 Aug–30 Oct 2025 and interviewed 47 teachers, producing six analytic themes suitable for thematic synthesis.
- Report and defend reliability metrics: the classroom observers achieved Cohen’s κ = 0.86 and interview coding agreement was 95%.
- Follow a seven-step Evidano-backed workflow (ingest, transcribe, prep codebook, double-code, synthesize, visualize, audit) to produce reproducible themes and exportable audit logs.
- Ensure ethics: the original study followed KFU-REC Decision No. KFU-REC-2025JUN-EA000985 and used consent and pseudonyms, and AI processing should be covered in consent.
Fast take & source
The quick summary: a 2026 qualitative study published in PLOS One examined 47 private-school science teachers in Al-Qassim and Al-Ahsa, using classroom observations (25 Aug–30 Oct 2025) and semi-structured interviews to study how AI tools support culturally responsive science curricula.
- Source: PLOS One (Published 21 July 2026).
- Key reliability numbers: classroom observers Cohen’s κ = 0.86; interview coding agreement = 95%.
- Primary dataset: n = 47 teachers; observation sessions of 45 minutes across Grades 6–9.
Findings snapshot
| Metric | Value | Notes / Why it matters |
|---|---|---|
| Sample | 47 science teachers | Purposive sample from private primary schools in Al-Qassim & Al-Ahsa |
| Data collection | Observations 25 Aug–30 Oct 2025; interviews (45–55 min) | Triangulation: observations first, then interviews |
| Adoption highlights | Personalization 97.9%; Interactive activities 89.4%; Cultural text analysis 85.1% | Indicates widespread teacher use of AI for contextualizing science content |
| Reliability | Cohen’s κ = 0.86; coding agreement 95% | Strong inter-rater reliability for observations and interview coding |
| Publication | Published 21 July 2026 (PLOS One) | Open access; DOI: 10.1371/journal.pone.0353777 |
What the researchers did (methods & analytic anchors)
The researchers used a purposive sample of 47 teachers, paired classroom observations and semi-structured interviews, and applied an inductive thematic analysis informed by grounded theory principles.
Observers used a validated checklist, two observers achieved κ = 0.86 and reconciled discrepancies in calibration meetings, and interview transcripts were double-coded with 95% agreement.
The analytic output was six themes: AI-supported personalization, analysis of cultural texts, AI-designed interactive activities, AI in curriculum development, interactive/collaborative learning, and intelligent assessment of cultural values.
- Transparency: authors state this is an inductive thematic approach (not full grounded theory, no theoretical sampling or saturation claims).
- Limitations flagged: perception-based outcomes; no standardized student achievement measures.
- Practical measures reported: session lengths (45 min), observation period (two+ months), and expert-validated instruments.
What this means for qualitative teams
For field researchers & evaluators
Field researchers and evaluators should pair observations with follow-up interviews to triangulate teacher practice and beliefs, as the study did.
Design: pair observations with follow-up interviews to triangulate teacher practice and beliefs.
Reliability: report inter-rater stats (κ, % agreement) and document calibration meetings to defend coding decisions.
Sampling notes: purposive convenience samples give depth but limit generalizability, report this transparently.
For curriculum designers
Curriculum designers should use AI outputs as co-design inputs while keeping human review to avoid bias or homogenization.
Use AI outputs as co-design inputs (localized examples, culturally relevant visuals) but keep human review to avoid bias or homogenization.
Assess: combine qualitative indicators (engagement, cultural alignment) with future quantitative measures to test impact on learning.
For edtech evaluators
Edtech evaluators should validate generative outputs against local expertise and prioritize human-in-the-loop assessment for culturally embedded reasoning.
Validate generative outputs against local expertise and curate training corpora to reduce cultural bias.
Prioritize human-in-the-loop assessment for culturally embedded reasoning.
Do more, faster with Evidano (map to this study)
Problem: lengthy transcripts & inconsistent labels
Evidano provides automatic transcription with a custom dictionary for local terms and PII redaction to meet ethics requirements.
Evidano solution: automatic transcription with custom dictionary for local terms (e.g., place names, farming practices) and PII redaction to meet ethics requirements.
Problem: manual coding drift across coders
Evidano supports importing a codebook and AI-assisted coding to auto-tag quotes while providing a coder dashboard for resolving disagreements and exporting inter-rater reports.
Evidano solution: import a codebook, run AI-assisted coding to auto-tag quotes, review disagreements in a coder dashboard, and export inter-rater reports matching κ calculations.
Problem: linking observations + interviews + artifacts
Evidano unifies heterogeneous inputs (audio, transcripts, classroom notes, images) for cross-segment and frequency analysis to reproduce counts like "46/47 teachers used X" with queryable provenance.
Evidano solution: unify heterogeneous inputs (audio, transcripts, classroom notes, images) for cross-segment and frequency analysis so you can say "46/47 teachers used X" with reproducible queries.
Problem: explaining results to stakeholders
Evidano creates one-click visualizations and exports slide decks that preserve quotes and source IDs to maintain audit trails for stakeholders.
Evidano solution: one-click visualizations (hierarchical themes → subcodes, co-occurrence networks, word clouds) and exportable slide decks that preserve quotes and source IDs for audit trails.
Problem: data security & trust
Evidano provides end-to-end encryption and proprietary LLMs tuned for qualitative research while not using customer data to train third-party models.
Evidano solution: end-to-end encryption and proprietary LLMs tuned for qualitative research; customer data is not used to train third-party models.
7-step workflow: reproduce the PLOS study analysis with AI support
This seven-step workflow reproduces the PLOS study analysis with AI support and maps each analytic step to traceable outputs.
- 1) Ingest: upload audio, observation checklists, and interview files to Evidano; apply the custom dictionary for local Arabic terms.
- 2) Transcribe & translate: run AI transcription with PII redaction; verify low-confidence segments manually.
- 3) Prep codebook: import the observation checklist as initial codes and run AI to surface candidate subcodes from transcripts.
- 4) Double-code: assign two coders, use Evidano’s disagreement dashboard, calculate κ and percent agreement, reconcile in calibration meeting.
- 5) Thematic synthesis: use Evidano to produce theme frequencies, cross-segment comparisons (by school/experience), and co-occurrence networks.
- 6) Visualize & export: create thematic hierarchies, quote lists linked to sources, and slide exports for stakeholders.
- 7) Audit & archive: lock final codebook, archive raw files with metadata, and export an audit log for ethics/compliance.
Ethics note
The study followed institutional ethics (KFU-REC Decision No. KFU-REC-2025JUN-EA000985) and used consent and pseudonyms.
When using AI tools, ensure consent covers AI processing and maintain human review for culturally sensitive outputs.
Wrapping up & next steps
AI-enabled qualitative research, when paired with clear coder protocols and human validation, can cut weeks from synthesis while strengthening reproducibility.
- Ready to apply this workflow? Start a pilot: ingest one teacher interview + two observation sessions into Evidano, run the seven-step sequence above, and produce a stakeholder brief in under 7 days.
- Request a demo or pilot on Evidano or Try Evidano for free, our team can show how to map your instruments, preserve audit trails, and generate the exact thematic and cross-segment outputs you need for publication or policy.
FAQ: AI-enabled qualitative research
What did the PLOS study find about AI adoption in Saudi school science?
The study found very high teacher adoption of AI for personalization (97.9%), interactive activities (89.4%), and cultural text analysis (85.1%).
The study reported that teachers used AI tools such as ChatGPT and DALL·E to support culturally responsive science curricula, producing six analytic themes across the sample.
How did the researchers ensure reliability in observations and coding?
The researchers reported strong inter-rater reliability, with classroom observers achieving Cohen’s κ = 0.86 and interview coding agreement of 95%.
Observers used a validated checklist, conducted calibration meetings to reconcile discrepancies, and double-coded transcripts to reach the reported agreement.
How can teams reproduce this study's analysis with AI?
Teams can reproduce the study’s analysis by following the seven-step workflow: ingest, transcribe, prep codebook, double-code, synthesize, visualize, and audit.
Map each step to tooling that preserves provenance and produces κ and percent-agreement statistics, then export audit logs and slide decks for stakeholder review.
What ethical safeguards did the study use and what should I do with AI?
The study followed institutional ethics (KFU-REC Decision No. KFU-REC-2025JUN-EA000985) and used consent and pseudonyms; researchers should ensure consent covers AI processing.
Maintain human review for culturally sensitive outputs, redact PII, and archive audit logs for ethics and compliance.