Submitted:
25 May 2026
Posted:
26 May 2026
You are already at the latest version
Abstract

Keywords:
1. Introduction
1.1. The Enduring Hegemony of the Textbook
1.2. The Promise and Limitations of Open Educational Resources
1.3. The Emergence of Generative AI in Education
1.4. The Research Gap: Integrating AI and OER
1.5. Research Questions and Significance
- RQ1: What design principles enable AI-powered, dynamic OER that are both pedagogically sound and practically usable by teachers?
- RQ2: How can generative AI systems maintain pedagogical quality (accuracy, alignment, cultural sensitivity) and ensure compliance with open licences?
- RQ3: What are the measurable benefits and risks of AI-dynamic OER compared to traditional static textbooks and static OER, in terms of student learning outcomes, engagement, and teacher perceptions?
2. Literature Review
2.1. Theoretical Foundations: UDL and Constructivism
2.2. The Textbook as an Artefact: History, Economics, and Inequity
2.3. Open Educational Resources: Definitions, Adoption, and Barriers
2.4. Adaptive Learning Technologies: From Rule-Based to Generative
2.5. Generative AI in Education: Opportunities and Risks
2.6. Prior Work at the Intersection of AI and OER
2.7. Synthesis and Contribution
3. Methodology
3.1. Research Design: Design-Based Research
3.2. Phase 1: Framework Development through Expert Focus Group
3.3. Phase 2: Prototype Development AI-OER Studio
- Topic (free text, e.g., “the water cycle”)
- Target grade level (dropdown: K2, 35, 68, 912, undergraduate)
- Language (dropdown: English, Spanish, French, Mandarin more languages could be added later)
- Cultural context (free text, e.g., “examples from agriculture in sub-Saharan Africa”)
- Learning objectives (free text, up to three)
- Output length (short ≈ 500 words, medium ≈ 1,500 words, long ≈ 3,000 words)
3.4. Phase 3: Evaluation
3.4.1. Expert Review
- Factual accuracy: Absence of factual errors, hallucinations, or misleading statements.
- Pedagogical alignment: How well the content matched typical learning objectives for the grade level, including appropriate depth and clarity.
- Bias / cultural sensitivity: Whether the content avoided stereotypes, included diverse perspectives where relevant, and used culturally inclusive language.
- Technical usability: How easy it was to view, navigate, and (in principle) edit the HTML output.
3.4.2. Classroom Pilot
3.5. Data Analysis: multilevel model
4. Findings / Results
4.1. Design Principles for AI-Powered Dynamic OER (RQ1)
4.2. Technical Feasibility and Licensing Compliance (RQ2)
4.3. Expert Evaluation of AI-Generated OER Quality
4.4. Classroom Pilot Outcomes (RQ3)
4.4.1. Matching of Groups
4.4.2. Learning Outcomes
4.4.3. Engagement (Time on Task)
4.4.4. Teacher Perceptions and Time Use
4.4.5. Multilevel results and adaptive feature usage
| Fixed effect | Coefficient | SE | t | p | 95% CI |
| Intercept | 32.4 | 5.1 | 6.35 | <0.001 | [22.4, 42.4] |
| Pre-test score | 0.52 | 0.09 | 5.78 | <0.001 | [0.34, 0.70] |
| Condition (AI-OER vs. static) | 7.8 | 2.9 | 2.69 | 0.009 | [2.1, 13.5] |
4.5. Summary of Key Findings
5. Discussion
5.1. Answering the Research Questions
5.2. The Death of the Static Textbook But Not the Teacher
5.2.1. Acknowledging the Novelty Effect
5.2.2. Legal Disclaimer Regarding AI-Generated OER
5.3. Comparison with Prior Adaptive Systems
5.4. A Detailed Risk Matrix and Mitigation Strategies
5.5. Theoretical Implications
5.6. Practical Implications for Different Stakeholders
6. Limitations and Future Research
6.1. Limitations of This Study
6.2. Directions for Future Research
7. Conclusions and Implications
7.1. Summary of Contributions
7.2. The Textbook Transformed, Not Erased
7.3. A Call to Action for the Educational Community
7.4. Final Reflection
Appendix A. Rubric for Evaluating AI-Generated OER Quality
| Dimension | Score 1 (Poor) | Score 3 (Acceptable) | Score 5 (Excellent) |
| Factual accuracy | Multiple major factual errors (e.g., wrong dates, false scientific claims). | 12 minor errors that do not undermine core understanding. | No errors; all statements can be verified from authoritative sources. |
| Pedagogical alignment | Content does not match stated grade level or learning objectives; inappropriate depth. | Partially matches; some objectives are addressed but not all. | Fully matches; appropriate vocabulary, examples, and assessment for the grade level. |
| Bias / cultural sensitivity | Stereotypical, exclusionary, or offensive language; only one cultural perspective presented. | Neutral but lacks diversity; no explicit bias but also no inclusion of diverse perspectives. | Actively inclusive; multiple cultural perspectives; language avoids stereotypes. |
| Technical usability | Cannot edit or export; broken formatting; not accessible (e.g., missing alt text). | Editable with some effort; basic accessibility features present. | One-click edit/export; fully responsive; meets WCAG 2.1 AA accessibility. |
Appendix B. Sample Prompt Template for the AI-OER Studio
| You are an AI assistant that generates Open Educational Resources (OER). You must release all output under the Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) license. Do not claim copyright on behalf of yourself or any other entity. Topic: {topic} Target grade level: {grade_level} (adjust vocabulary, sentence length, and concept depth accordingly) Language: {language} Cultural context: Please use examples primarily from {cultural_context}. If the context is not specified, use a globally diverse set of examples. Learning objectives: {learning_objectives} Generate the following sections: 1. An introductory paragraph that engages the learner and states the key idea. 2. A list of 3-5 key terms with student-friendly definitions. 3. Two to three multiple-choice questions, each with four options, an answer key, and a brief explanation of why the correct answer is correct. 4. One short-answer prompt that asks students to apply the concept. Format the output as JSON with keys: "intro", "key_terms", "mc_questions", "short_answer". After the JSON, add a footer in HTML: "<footer>This resource was generated by AI (GPT-4) and should be reviewed by an educator before use. Licensed CC BY-SA 4.0.</footer>" |
Appendix C. Teacher Interview Protocol (Semi-Structured)
- General experience: Overall, how did you find the process of using AI-generated OER compared to your usual textbook or static OER?
- Time impact: Approximately how much time did you spend each week on preparing materials using the AI tool? How does that compare to your usual preparation time?
- Quality of content: Can you give me a specific example of a time when the AI-generated content was excellent? A time when it was problematic (factual error, bias, unclear)?
- Student reactions: How did students respond to the ability to get different versions (reading level, language, examples)? Did any student use that feature?
- Trust and validation: How much did you trust the AI content before your own review? After review? What would increase your trust?
- Features and improvements: What is the single most important feature you would add or change?
- Sustainability: Would you continue using this tool if it were available? Would you recommend it to colleagues? Why or why not?
- Risks: What concerns, if any, do you have about the broader adoption of AI-generated OER in schools?
Appendix D. Example of Static vs. AI-Dynamic OER (Textual Comparison)
References
- Aleven, V., McLaughlin, E. A., Glenn, R. A., & Koedinger, K. R. (2016). Instruction based on adaptive learning technologies. In R. E. Mayer & P. A. Alexander (Eds.), Handbook of research on learning and instruction (2nd ed., pp. 522560). Routledge. [CrossRef]
- Baidoo-Anu, D., & Ansah, L. O. (2023). Education in the era of generative artificial intelligence (AI): Understanding the potential benefits of ChatGPT in promoting teaching and learning. Journal of AI, 7(1), 5262. [CrossRef]
- Bazzul, J., & Sykes, H. (2011). The secret identity of a biology textbook: Straight and naturally sexed. Cultural Studies of Science Education, 6(2), 265286. [CrossRef]
- Birhane, A., Ruane, E., Laurent, T., Brown, M. S., Flowers, J., Ventresque, A., & Dancy, C. L. (2022). The forgotten margins of AI ethics. Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency, 948958. [CrossRef]
- Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77101. [CrossRef]
- Brown, A. L., & Brown, K. D. (2010). Strange fruit indeed: Interrogating contemporary textbook representations of racial violence toward African Americans. Teachers College Record, 112(1), 3167. [CrossRef]
- CAST. (2018). Universal Design for Learning guidelines version 2.2. https://udlguidelines.cast.org.
- Clark, R. C., & Mayer, R. E. (2024). E-learning and the science of instruction (6th ed.). Wiley. [CrossRef]
- Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Lawrence Erlbaum.
- Colvard, N. B., Watson, C. E., & Park, H. (2018). The impact of open educational resources on various student success metrics. International Journal of Teaching and Learning in Higher Education, 30(2), 262276. https://www.isetl.org/ijtlhe/pdf/IJTLHE3386.pdf.
- Common Corpus. (2024). Common Corpus: A large-scale open dataset for training LLMs on public domain and openly licensed texts. https://commoncorpus.com.
- Fleiss, J. L. (1971). Measuring nominal scale agreement among many raters. Psychological Bulletin, 76(5), 378–382. [CrossRef]
- Hake, R. R. (1998). Interactive-engagement versus traditional methods: A six-thousand-student survey of mechanics test data for introductory physics courses. American Journal of Physics, 66(1), 6474. [CrossRef]
- Hedges, L. V. (1981). Distribution theory for Glass’s estimator of effect size and related estimators. Journal of Educational Statistics, 6(2), 107128. [CrossRef]
- Hilton, J. (2016). Open educational resources and college textbook choices: A review of research on efficacy and perceptions. Educational Technology Research and Development, 64(4), 573590. [CrossRef]
- Hilton, J. (2020). Open educational resources, student efficacy, and user perceptions: A synthesis of research published between 2015 and 2018. Educational Technology Research and Development, 68(3), 853876. [CrossRef]
- Holstein, K., McLaren, B. M., & Aleven, V. (2019). Designing for complementarity: Teacher and AI needs for the future of K-12 learning. Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems, 112. [CrossRef]
- International Council for Open and Distance Education (ICDE). (2023). AI and OER: A scoping report. https://www.icde.org/ai-oer-report.
- Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Bang, Y., Dai, W., Madotto, A., & Fung, P. (2023). Survey of hallucination in natural language generation. ACM Computing Surveys, 55(12), 138. [CrossRef]
- Kasneci, E., Seßler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., Gasser, U., Groh, G., Günnemann, S., Hüllermeier, E., Krusche, S., Kutyniok, G., Michaeli, T., Nerdel, C., Pfeffer, J., Poquet, O., Sailer, M., Schmidt, A., Seidel, T., … Kasneci, G. (2023). ChatGPT for good? On opportunities and challenges of large language models for education. Learning and Individual Differences, 103, 102274. [CrossRef]
- Kitchin, R. (2022). The data revolution: A critical analysis (2nd ed.). Sage. [CrossRef]
- Knox, J. (2019). What does the ‘postdigital’ mean for education? Three critical perspectives. Postdigital Science and Education, 1(2), 357–370. [CrossRef]
- McKenney, S., & Reeves, T. C. (2019). Conducting educational design research (2nd ed.). Routledge. [CrossRef]
- Mollick, E. R., & Mollick, L. (2023). Using AI to implement effective teaching strategies in classrooms: Five strategies, including prompts. SSRN Electronic Journal. Advance online publication. [CrossRef]
- Montgomery, D. C., Hinton, K. D., & Kirby, S. D. (2021). Teacher time use and instructional differentiation: A national survey. Journal of Educational Research, 114(3), 245259. [CrossRef]
- Open Education Group. (2024). AI-OER prototype: Demonstration and design notes. https://openedgroup.org/ai-oer-prototype.
- Reiser, R. A. (2018). A history of instructional design and technology. In R. A. Reiser & J. V. Dempsey (Eds.), Trends and issues in instructional design and technology (4th ed., pp. 2033). Pearson.
- Rose, D. H., & Gravel, J. W. (2010). Universal Design for Learning. In P. Peterson, E. Baker, & B. McGaw (Eds.), International encyclopedia of education (3rd ed., pp. 119124). Elsevier. [CrossRef]
- Rosenberg, J. M., Borchers, C., & Gibbons, B. (2024). Large language models and educational equity: A systematic bias audit of six LLMs. Computers & Education, 215, 105032. [CrossRef]
- Seaman, J. E., & Seaman, J. (2021). Digital texts in the time of COVID: A survey of faculty. Bay View Analytics. https://www.bayviewanalytics.com/reports/digitaltextsinthetimeofcovid.pdf.
- Siemens, G. (2005). Connectivism: A learning theory for the digital age. International Journal of Instructional Technology and Distance Learning, 2(1), 310. http://www.itdl.org/Journal/Jan_05/article01.htm.
- Steenbergen-Hui, S., & Cooper, H. (2014). A meta-analysis of the effectiveness of intelligent tutoring systems on K-12 students’ mathematical learning. Journal of Educational Psychology, 106(2), 331347. [CrossRef]
- Tlili, A., Huang, R., Chang, T. W., & Burgos, D. (2023). Open educational resources in the age of artificial intelligence: A Delphi study. British Journal of Educational Technology, 54(2), 543567. [CrossRef]
- Tlili, A., Huang, R., Burgos, D., & Stracke, C. M. (2024). Sustainability of Open Educational Resources in the age of generative AI. British Journal of Educational Technology, 55(3), 892–912. [CrossRef]
- Tomlinson, C. A. (2014). The differentiated classroom: Responding to the needs of all learners (2nd ed.). ASCD.
- UNESCO. (2019). Recommendation on Open Educational Resources (OER). https://www.unesco.org/en/legal-affairs/recommendation-open-educational-resources-oer.
- U.S. Copyright Office. (2023). Copyright Registration Guidance for Works Containing Material Generated by Artificial Intelligence. Federal Register, 88(51), 1619016194. https://www.federalregister.gov/d/2023-05321.
- U.S. Government Accountability Office. (2019). College textbooks: Students have greater access to textbook information, but most instructors do not adopt free digital materials. GAO-19-299. https://www.gao.gov/products/gao-19-299.
- Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes. Harvard University Press.
- Walden University. (2024). Generative AI for writing feedback: A pilot study. https://waldenu.edu/research/ai-writing-pilot.
- Wiley, D., & Hilton, J. (2018). Defining OER-enabled pedagogy. International Review of Research in Open and Distributed Learning, 19(4), 133147. [CrossRef]
- Williamson, B. (2021). Double, double, toil and trouble: The digital education data cycle. Learning, Media and Technology, 46(1), 15. [CrossRef]
| Principle | Description | Example implementation from prototype | Expert quote illustrating need |
| P1: Prompt transparency | Every AI-generated output must clearly disclose that it was generated by AI, including the model name and version, the training data cutoff date, and any human post-editing. | Footer: “This text generated by GPT-4 (trained on data up to May 2024). Reviewed by [teacher name] on [date].” | “If I don’t know where this came from, I cannot trust it with my students. A textbook has a publisher and authors. AI needs something equivalent.” (Teacher, female, middle school) |
| P2: Human-in-the-loop validation | Teachers must have the final say. The AI should never distribute content directly to students without an explicit human approval step. | “Approve all” button after editing; option to reject and regenerate any paragraph. | “I would never hand over my class to a bot. But if you give me a fast way to check and fix, that’s a huge time saver.” (Instructional designer, male, university) |
| P3: Adaptive granularity | The system should be able to generate multiple versions of the same core content at different reading levels, in different languages, and with culturally varied examples, with one click. | Dropdown to switch reading level from “grade 5” to “grade 9” regenerates the entire page with adjusted vocabulary and sentence length. | “I have students reading at a second-grade level and others at high school level in the same room. One version will never work. The AI must give me three or four quickly.” (Teacher, female, elementary) |
| P4: License preservation | The AI’s output must automatically inherit an open licence (preferably CC BY-SA or CC BY) and include machine-readable metadata. The system must not claim copyright on behalf of the AI. | HTML export includes CC BY-SA logo, legal code link, and JSON-LD metadata. | “The whole point of OER is that you can share and remix. If the AI makes it impossible to tell what the licence is, it defeats the purpose.” (OER librarian, female) |
| Dimension | Biology module (Ecosystems) | History module (Industrial Revolution) | Combined mean (SD) | Range |
| Factual accuracy | 4.3 (0.6) | 4.0 (0.8) | 4.15 (0.7) | 3.05.0 |
| Pedagogical alignment | 4.5 (0.5) | 4.3 (0.7) | 4.40 (0.6) | 3.55.0 |
| Bias / cultural sensitivity | 4.1 (0.9) | 3.8 (1.0) | 3.95 (0.95) | 2.05.0 |
| Technical usability | 4.7 (0.4) | 4.6 (0.5) | 4.65 (0.45) | 4.05.0 |
| Characteristic | Treatment (n=39) | Control (n=39) | p-value (t-test or chi-square) |
| Age (years, mean ± SD) | 12.4 ± 0.5 | 12.5 ± 0.6 | 0.44 |
| Female (%) | 51% | 54% | 0.81 |
| English language learner (%) | 18% | 21% | 0.76 |
| Individualised education plan (%) | 13% | 10% | 0.71 |
| Prior semester science grade (%) | 85.2 ± 11.3 | 84.9 ± 10.8 | 0.76 |
| Home computer access (%) | 92% | 87% | 0.48 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2026 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license.