PAIDEIA: A Human-Centered Pedagogical Framework for Responsible Generative AI Integration in Ancient Greek Literature in Translation

Synthesis: Lymperi (2026) proposes PAIDEIA, a conceptual framework for integrating Generative AI into Ancient Greek literature in translation at the lower secondary level. It joins five dimensions: classical paideia as holistic formation, Human-Centered Artificial Intelligence, AI Literacy, Critical Thinking, and dialogic learning. The central principle is that AI output should function as an epistemic stimulus rather than an authoritative answer, so students question, compare, verify, interpret, and justify while the teacher designs the task and keeps disciplinary and ethical aims primary. An illustrative sequence, built on the author's earlier Euripides' Helen scenario for Grade 9 and estimated at two teaching hours, runs through four stages: students' own interpretation, an AI-generated reading, critical and dialogic examination, and evidence-based reconstruction. The paper states plainly that it has no empirical validation and reports no classroom data. The distinction worth keeping is between AI competence and educational purpose: AI literacy is embedded in interpretation, dialogue, and ethical judgment rather than treated as the endpoint.

Key Findings

1. Five dimensions, one humanistic orientation. PAIDEIA combines classical paideia, Human-Centered AI, AI Literacy, Critical Thinking, and dialogic learning, with paideia supplying the overarching purpose and the other four contributing complementary contemporary pedagogical dimensions.
2. Four stated contributions. The paper claims a framework, a synthesis of paideia with Human-Centered AI and literacy scholarship, a pedagogical role for AI output as an object of inquiry, and an illustrative classroom application with research directions.
3. Two UNESCO texts anchor the ethics. The framework's normative claims rest on UNESCO's 2023 guidance for generative AI in education, which stresses human oversight and learner agency, and its 2021 Recommendation on the Ethics of Artificial Intelligence.
4. Grade 9, two teaching hours. The illustrative application draws on the author's 2025 E.DI.V.E.A. scenario on Euripides' Helen, designed for Grade 9 and estimated at two teaching hours.
5. Four stages keep the student first. Students interpret before any AI text appears, then examine an AI-generated reading against the play, discuss it, and reconstruct a justified reading of their own.
6. No empirical validation. The 2026 preprint states that the framework has not been empirically validated and that the Helen sequence is illustration, not evidence of effectiveness.

Five dimensions and what each supplies

The framework is an architecture for Humanities and Social Science Education, not a lesson plan. Classical paideia, following Jaeger, is treated as holistic intellectual, ethical, cultural, and civic formation, not the simple transmission of knowledge, and it sets the purpose the other dimensions serve. Human-Centered Artificial Intelligence (HCAI), following Shneiderman, supplies the design stance: systems should augment human capability while people keep control, accountability, and responsibility. AI Literacy supplies the evaluative capacity, extending past functional use of tools to judging accuracy, bias, and limits. Critical Thinking supplies the criteria students apply to an AI reading: accuracy, coherence, assumptions, evidence, and interpretative validity. Dialogic learning supplies the mechanism, since meaning is negotiated in classroom talk rather than received from a model. Human agency sits at the center: the teacher decides when and why AI enters, frames the questions, and sets the criteria for judging output, while students retain interpretation, judgment, and argument.

The Helen sequence

The illustrative application reuses a teaching scenario the author published in a 2025 E.DI.V.E.A. collection, designed for Grade 9 and estimated at two teaching hours. The paper is explicit that this is illustration, not evidence. Stage 1 asks students to read selected passages and commit to an interpretation with textual support before any AI text appears, protecting their interpretative agency. Stage 2 presents an AI-generated response to a crafted literary question, framed as a text to examine rather than correct information. Stage 3 compares that response with the play and with students' own readings, testing which claims the text supports, which are unsupported, what context is missing, and how rhetorical fluency may shape perceptions of credibility. Stage 4 asks for a revised interpretation that corrects, extends, or rejects the AI reading and justifies the choice. The stated sequence runs from text to student interpretation, AI stimulus, critical examination, dialogue, evidence-based reconstruction, and Reflection.

Design guidance, not tested effects

The paper reports no sample, no instrument, and no measured outcome, and it says the framework has not been empirically validated. Its claims are therefore design guidance: what a teacher should sequence and why, not what a sequence produces. Three constraints follow. The setting is narrow, one play in translation at one grade level in a Greek secondary collection, and the author proposes adaptation to other subjects and age groups as future work rather than claiming transfer. Assessment is left to the teacher, who sets criteria for evaluating AI output; the paper names no rubric or measure for its own aims. The ethical work is real but declarative, covering authorship, Academic Integrity, transparency, bias, and responsibility, without evidence about how students reason through those questions. Its value lies in the design argument and the distinction between AI competence and educational purpose, which future mixed-methods studies could test.

What this means for practice

• Instructors. Have students commit to an evidence-backed reading before showing them an AI interpretation, then require them to name which AI claims the text supports and which it does not.
• Instructors. Ask students to judge the AI response on assumptions and missing context, not only factual errors, so rhetorical fluency does not pass for reliability.
• Curriculum designers. Treat PAIDEIA as a design frame for teacher-written prompts and criteria, not a ready-made unit; the paper supplies only an outline.
• Administrators. Extend the conversation beyond one lesson to school-level norms on accuracy, bias, authorship, and privacy, which the paper treats as whole-school questions.

Limitations

• No empirical validation. The paper states the framework has not been validated and offers no classroom data, so its stated aims, including critical thinking and AI literacy, remain untested.
• The illustration is a reused scenario. The Helen sequence comes from a 2025 E.DI.V.E.A. collection, designed for Grade 9 at two teaching hours, and the paper does not treat it as evidence.
• One narrow setting. Lower-secondary Ancient Greek literature in translation, a single play, and a Greek curriculum context; transfer to other texts, ages, subjects, or educational contexts is proposed, not shown.
• No instruments. The paper assigns criteria-setting to teachers but names no rubric, outcome measure, or analytic method for testing the framework.
• No named tool or model. GenAI is discussed generically, so the design cannot be tied to a specific system's behavior or version.

Citation

Lymperi, T. (2026). PAIDEIA: A Human-Centered Pedagogical Framework for Responsible Generative AI Integration in Ancient Greek Literature in Translation. EdArXiv preprint (v1).