Integrating Artificial Intelligence in the teaching of modern Greek literature: An educational proposal for secondary education

 Corresponding author: Koralia Trikalioti Copyright © 2025 Author(s) retain the copyright of this article. This article is published under the terms of the Creative Commons Attribution License 4.0. Integrating Artificial Intelligence in the teaching of modern Greek literature: An educational proposal for secondary education Koralia S. Trikalioti * and Ioanna Katsiampoura Department of Pedagogy and Primary Education, National and Kapodistrian University of Athens, Greece. World Journal of Advanced Research and Reviews, 2025, 28(01), 1618-1630 Publication history: Received on 05 September 2025; revised on 20 October 2025; accepted on 22 October 2025 Article DOI: https://doi.org/10.30574/wjarr.2025.28.1.3583 Abstract This article derives from the author’s Master’s thesis conducted within the postgraduate program Digital Culture, Smart Cities, IoT and Advanced Digital Technologies at the University of Piraeus, yet it extends beyond a mere presentation of the thesis to situate its findings within the broader context of rapidly evolving educational technologies. The study proposes an innovative instructional approach that integrates artificial intelligence (AI)-generated images into the teaching of Modern Greek literature, with a focus on poetry in the second year of senior high school. Its primary objective is to provide an alternative mode of interpreting literary texts that complements traditional instruction, fostering deeper comprehension, engagement, creativity, and inspiration among students. By visualizing the abstract concepts, words, and verses of poems through AI-generated images, learners are encouraged to connect with literary meaning on both cognitive and affective levels. Crucially, the approach seeks not only to align images with the poems but also to surpass conventional textbook illustrations by initiating a dialogic relationship with the texts and enhancing interpretive depth. Within the proposed scenario, students develop critical thinking, imagination, creativity, and collaborative skills as they create and discuss their own AI-generated images, thereby participating in a process consistent with STEAM education that integrates science, technology, engineering, art, and mathematics. The study also acknowledges limitations, including technical challenges, the absence of classroom-based pilot implementation, and the lack of empirical evaluation with student samples. Future directions emphasize refining and expanding the approach in light of the rapid advancement of AI tools, which offer new opportunities for innovative and evidence-based pedagogical applications. Keywords: Artificial Intelligence; Literature Education; STEAM; Visualization; Poetry; Digital Pedagogy 1. Introduction Within the specialization of Digital Culture in this Master’s program, we extensively examined technological applications and methods of digitizing cultural heritage. This included the creation of digital collections as well as the development of digital spaces (e.g., virtual museums) that hosted exhibitions of artworks across diverse themes. The aim behind each such initiative was to facilitate and expand the means of disseminating Art and Knowledge to a broader audience— users, viewers, and readers who could gain direct, cost-free, and easy access to a promising digital experience—without undermining the importance of physical visits, hospitality, and guided tours in libraries or museums [1]. Following the same rationale, and observing the rapid advances in Artificial Intelligence (AI) across all fields and their positive impacts, we drew inspiration to explore innovative, contemporary methods in educational practice. In recent years, generative AI has been increasingly explored as a pedagogical tool in art and humanities education [2]. Similar directions have also been highlighted in recent research that integrates poetry, AI, and STEAM pedagogy [3]. For more than a decade, such initiatives have been steadily developing abroad, and more cautiously in Greece [4]. By focusing on World Journal of Advanced Research and Reviews, 2025, 28(01), 1618-1630 1619 the beneficial outcomes of technology integration in teaching—namely the enhancement of understanding, the activation of student interest, the cultivation of critical thinking, and the stimulation of imagination—the intention arose to design a teaching scenario. This scenario would act as a complementary tool for both literature teachers and their students, helping to transmit poetic concepts and transforming a conventional lesson into an experience enriched with images, colors, and creativity [2]. Specifically, this teaching scenario aims to foster deeper comprehension of Surrealist and Symbolist poetry, with a focus on the “Modern Poetry” module taught in upper secondary education. Teachers often report that these poetic movements are difficult for students to understand, due to the prevalence of abstract ideas, imageries, and symbols, as well as the challenge of interpreting deeper meanings [3]. In addition to comprehension difficulties, many students perceive such texts as outdated, disconnected from their 21st-century realities, and anchored in past epochs with different social contexts. Yet the richness and timelessness of these poems emerge vividly when integrated into modern methods of learning that employ digital technologies and AI, proving their enduring value as educational resources [5]. The challenge extends further: within this teaching scenario, students are expected to acquire the technical skills needed to use AI tools and to generate their own prompts so that a digital image is created from a chosen poetic verse. Text-to-image models allow users to generate visual art from natural language prompts, bridging verbal and visual modes of meaning-making [5]. Previous work shows how AI-based “Poetry Machines” can influence students’ poetic writing and engagement [6]. Also, through collaborative work, they will select and vote on the best creation, thereby cultivating imagination, teamwork, and cooperation, under the spirit of constructive competition [6]. 2. Theoretical Framework The proposed teaching scenario is grounded in two fundamental educational and scientific pillars, both of which highlight how students can most effectively acquire knowledge within the learning process. On the one hand, it draws upon Allan Paivio’s Dual Coding Theory (1990, Mental Representations: A Dual Coding Approach, Oxford University Press) [7]. According to Paivio, human cognition operates through two distinct yet interconnected systems: the verbal system (responsible for processing language, words, and symbols) and the non-verbal system (responsible for processing images, visual perception, and spatial information). The theory rests on three key ideas about learning: Complementarity: When learners are presented with both verbal and visual information, the two systems interact to generate multiple memory traces, thereby increasing the likelihood of retention and recall. Concrete vs. Abstract concepts: Concrete words (e.g., “tree”) are easier to process because they activate both systems simultaneously. Abstract concepts (e.g., “justice”) are more difficult, but become more accessible when supported by imagery or metaphor. Pedagogical implications: Instruction is more effective when verbal explanations are accompanied by visual representations such as diagrams, photographs, or digital media. Figure 1 Paivio’s Dual Coding Theory (1990): Interaction between verbal and non-verbal systems. World Journal of Advanced Research and Reviews, 2025, 28(01), 1618-1630 1620 Within the development of this teaching scenario, Paivio’s theory finds direct application: poetic language is combined with AI-generated digital images, created through specific prompts based on the abstract concepts and meanings central to the poem. Dual Coding thus directly supports the claim that students can engage with abstract poetry more effectively when visual representations are introduced—particularly when artificial intelligence is used to “translate” abstract verses into tangible visual forms. On the other hand, the framework also builds on Richard E. Mayer’s Cognitive Theory of Multimedia Learning [8]. Mayer extends Paivio’s ideas within the context of modern cognitive psychology and instructional design. His theory rests on three core assumptions: Dual channels: Learners process information through two channels—auditory/verbal and visual/pictorial. Limited capacity: Each channel has limited capacity in working memory, which makes overload a significant concern. Active processing: Genuine learning occurs when learners actively select, organize, and integrate information from both channels. From these assumptions, Mayer formulates a set of design principles, several of which are highly relevant here: Modality principle: Spoken explanations paired with images are more effective than written text combined with images within the same channel. Redundancy principle: Unnecessary repetition (e.g., identical narration and on-screen text) should be avoided. Coherence principle: Extraneous material must be excluded to prevent cognitive overload. Figure 2 Mayer’s Cognitive Theory of Multimedia Learning (2001): Dual channels, limited capacity, and active processing. When applied to pedagogy, multimedia resources—when carefully designed to integrate words and images—enhance comprehension and long-term learning. In the present proposal, which introduces digital AI-generated images as supplementary material in teaching Modern Greek poetry, Mayer’s theory explains why such visualizations, when appropriately embedded, can increase understanding and engagement. At the same time, it cautions that overly complex or poorly chosen images risk causing cognitive overload—an important limitation to acknowledge. By integrating these two theoretical perspectives, the scenario advances toward a third layer: the educational framework of STEAM (Science, Technology, Engineering, Arts, Mathematics). STEAM emphasizes interdisciplinarity, creativity, and critical thinking. Modern Greek literature—particularly twentieth-century poetry—primarily belongs to the Arts dimension of STEAM. The proposed framework first links Paivio’s Dual Coding Theory to STEAM, then Mayer’s Multimedia Learning Theory, and finally unites both within a STEAM-oriented teaching design. Paivio’s Dual Coding Theory demonstrates that artistic and visual elements (such as AI-generated imagery) enhance memory and comprehension. Within STEAM, the Arts component enriches the sciences with imagination and creativity—paralleling Paivio’s non-verbal system. Mayer’s principles align with the multimodal and collaborative learning central to STEAM: students actively combine multiple channels (visual, verbal, interactive). The proposed classroom practice—students designing prompts and discussing the generated images—serves as a clear example of active cognitive processing. Thus, through Paivio, abstract poetry becomes more accessible via visualizations; through Mayer, the integration of words World Journal of Advanced Research and Reviews, 2025, 28(01), 1618-1630 1621 and images is optimized for learning; and through STEAM, poetry instruction enriched with AI imagery unites Art and Technology, fostering creativity, critical thinking, and collaboration among learners. Figure 3 Integration of Paivio, Mayer, and STEAM within the proposed teaching scenario. Finally, it is worth clarifying how this proposal aligns with each of the five components of STEAM: Science: Both Paivio’s and Mayer’s theories highlight the cognitive mechanisms by which visualization improves understanding of abstract concepts, grounding the approach in scientific accounts of memory and comprehension. Technology: The use of AI image generators represents an innovative technological application within education. Engineering: The design of a teaching scenario itself reflects an engineering design process, involving goals, inputs, outputs, and iterative testing. Arts: Poetry and its visual interpretation through AI-generated images sit at the heart of artistic creation and expression. Learners’ active role in creating imagery cultivates aesthetic sensitivity and critical engagement with texts. Mathematics: Although less explicit, the AI models used to generate images (e.g., diffusion models, GANs) function on mathematical algorithms, linking the artistic outcome back to mathematical underpinnings. 3. Literature Review 3.1. AI in Education Recent scholarship highlights the rapid expansion of generative artificial intelligence (AI) in education, with an emphasis on both pedagogical opportunities and methodological challenges. Noroozi et al. [9] outline the main theoretical and practical dimensions of generative AI, emphasizing its potential to reshape teaching and learning while also highlighting concerns about ethics and quality. In a similar direction, Su and Yang [10] suggest a framework for the use of large language models, such as ChatGPT, in educational contexts, showing how they can provide personalization, structured guidance, and adaptive feedback. At a broader level, Giannakos et al. [11] discuss the promise and challenges of generative AI, noting both its capacity to enhance learner engagement and the risks of overreliance or inequitable access. Other studies have focused on AI-generated educational resources and point to the importance of establishing clear standards of quality [12]. At the same time, reviews of immersive technologies such as virtual and augmented reality [13] [14] show similar advantages and drawbacks, reinforcing the argument that technology use in education should be guided by sound cognitive and pedagogical principles rather than technological novelty. 3.2. Applications in Language and Literature Teaching Although the majority of studies on AI in education focus on STEM fields, a growing body of research addresses its role in language and literature. Kouvara et al. [3] explore how poetry can be integrated into STEAM education through the use of AI, arguing that such interdisciplinary approaches foster creativity and broaden the scope of the “A” (Arts) within the STEAM framework. Tang and Leong [15] report on the integration of AI-generated art styles into illustration teaching, underscoring how generative technologies can support artistic literacy and student expression. Likewise, Cornelia Berg et al. [16] examine the motivational and cognitive effects of student-produced AI-generated images, concluding that such practices enhance both engagement and satisfaction. These studies provide evidence that AI can World Journal of Advanced Research and Reviews, 2025, 28(01), 1618-1630 1622 enrich the teaching of literature and the arts, particularly when students are encouraged to take an active role in the production and interpretation of AI-generated content. Figure 4 Conceptual space of the literature review, showing the intersection of AI in education, language and literature teaching, and multimodal resources, which frames the focus of this thesis. 3.3. Use of Images and Multimodal Resources in Education The pedagogical value of images and multimodal resources has been long established in educational research, often drawing on theories such as Paivio’s Dual Coding Theory and Mayer’s Cognitive Theory of Multimedia Learning. More recent studies extend these frameworks into the domain of artificial intelligence. Aktay [17], for instance, investigates the usability of AI-generated images in educational contexts, finding that visual materials created through AI can serve as effective mediators of abstract concepts when carefully designed. Other reviews of immersive and multimodal resources [13] [14] highlight their capacity to stimulate deeper learning and support diverse cognitive styles. Taken together, these findings confirm the importance of multimodality in learning and suggest that the addition of AIgenerated imagery represents a natural evolution of established practices in educational design. Table 1 Overview of key literature categorized into three thematic areas: AI in education, applications in language and literature teaching, and the use of images and multimodal resources. Category Reference Key Contribution AI in Education Noroozi et al. (2024) – Generative AI in Education Overview of pedagogical & methodological perspectives AI in Education Su & Yang (2023) – Framework for ChatGPT in Education Framework for LLMs in diverse settings AI in Education Giannakos et al. (2025) – Promise and challenges of generative AI Benefits and risks of generative AI AI in Education Huang et al. (2025) – AI-generated educational resources Quality evaluation of AI educational resources World Journal of Advanced Research and Reviews, 2025, 28(01), 1618-1630 1623 Language & Literature Teaching Kouvara et al. (2024) – Expanding the ‘A’ in STEAM Poetry + AI in STEAM; creativity & interdisciplinarity Language & Literature Teaching Tang & Leong (2025) – AI-generated art in illustration teaching AI art styles in art/literature teaching Language & Literature Teaching Cornelia Berg et al. (2024) – AI-generated images & motivation Impact of AI images on motivation & learning Images & Multimodality Aktay (2022) – Usability of AI-generated images in education AI images as mediators of abstract concepts 3.4. Summary Current research indicates that AI has been widely explored in education through frameworks such as intelligent tutoring systems and generative models, though most applications remain in STEM. Applications in language and literature are emerging but less developed, with promising evidence that AI-generated texts and images can enhance engagement and comprehension. Meanwhile, the established role of multimodal resources in learning suggests that integrating AI-generated images into the teaching of poetry directly addresses a gap in the literature: the use of generative AI to support the interpretation of abstract and complex literary texts. 4. AI-Based Digital Tools, Prompt Creation, and Teaching Scenario 4.1. AI Tools for the Creation of Digital Images Within the framework of this teaching proposal, three different AI applications (Stable Diffusion, Deep Dream Generator, and Pixlr) were examined and evaluated with regard to their potential for classroom use, based on criteria relevant to educators: • Ease of use (user interface) • Simplicity of prompt design and execution • Free access or low-cost availability • Speed of image generation • Options for editing or refining generated images The following section summarizes the main findings regarding these applications in order to provide a clearer picture of their affordances: Stable Diffusion is supported by more advanced, state-of-the-art models, but it requires higher levels of technical expertise, installation, and in many cases subscription fees. As such, it is not easily accessible for teachers of literature or general educators seeking direct classroom integration. Deep Dream Generator encourages creative experimentation but is limited in its free version, both in the number of daily generations and in prompt flexibility. Pixlr, by contrast, is an “all-in-one” solution: browser-based, intuitive, fast, and equipped with ready-to-use tools. This makes it a more suitable option for educators who aim to produce visual material quickly and without engaging with complex technical requirements. Since the present article was written a year after the completion and publication of the master’s thesis, it was deemed appropriate to update the previous findings with the latest developments in these tools: Stable Diffusion 3.5 now offers enhanced prompt comprehension and significantly improved rendering of text and typography. It can also be accessed via cloud (Amazon Bedrock), eliminating the need for local installation. Additional extensions, such as ControlNet and LoRA fine-tuning, provide more precise control over outputs, though they remain technically demanding for non-specialists. Deep Dream Generator has introduced an AI Upscaler (producing images up to 120MP, suitable for classroom printing) and a new AI Video Generator, which transforms an image or text prompt into a short 8-second video. However, the free version still allows only three generations per day. Pixlr has integrated several new features, including AI Generative Fill (enabling teachers to add elements directly relevant to literary analysis, e.g., symbols or objects tied to a poem), AI Remove Background/Object (useful for adapting generated visuals to teaching materials), and AI Expand (extending images for presentations). Furthermore, tools such as AI Face Swap and AI Image to Video broaden its creative capacity. Its highly user-friendly interface—available both in browser and mobile application—makes Pixlr especially attractive for educators, offering a low learning curve and immediate integration into classroom practice. World Journal of Advanced Research and Reviews, 2025, 28(01), 1618-1630 1624 Table 2 AI Tools for Image Generation (Update 2024–2025) Tool New Features Usefulness for Educators Limitations Stable Diffusion 3.5 - Improved prompt understanding - Better text/typography rendering - Cloud access via Amazon Bedrock - Extensions: ControlNet, LoRA fine-tuning Enables high-quality image generation with great accuracy for advanced users Requires technical expertise; less accessible for nonspecialist educators Deep Dream Generator - AI Upscaler (up to 120MP) - AI Video Generator (image/text → 8-second video) Produces high-resolution images and short videos that can enrich classroom material Free version limited to 3 images per day Pixlr - AI Generative Fill - AI Remove Background/Object - AI Expand - AI Face Swap – AI Image to Video Very user-friendly (browser & mobile app); ideal for quickly integrating visuals into teaching resources Some features restricted in the free version 4.2. Prompt Design Image generation through AI relies on transforming natural language prompts into visual representations, where largescale text–image models “translate” verbal descriptions into visual features. The effectiveness of a prompt depends greatly on its level of specificity and descriptiveness: abstract or highly poetic phrases (e.g., “the melancholy of time”) often fail to produce coherent visual results, since models cannot easily associate intangible concepts with concrete visual data. In contrast, when such ideas are reformulated into explicit, recognizable imagery (e.g., “an old clock melting in a dark landscape”), the system is able to approximate the intended meaning with greater fidelity. This demonstrates the challenges involved in translating poetic language into images and highlights the importance of simplifying and concretizing prompts to generate outputs that remain closely aligned with the creator’s original intention. In practice, the main challenge in designing prompts for the selected applications was to distill the deeper meaning of each poem while avoiding excessive reliance on surreal phrasing, symbolism, or abstract concepts, without omitting them altogether. For each poem, we composed one or more concise sentences that encapsulated its essence in simplified terms, making it feasible for the system to generate an image that faithfully reflected “what the poet intended to say.” Each of the twelve poems in the corpus was therefore accompanied by a caption functioning both as a prompt and as a pedagogical aid: on the one hand, the caption helped students grasp the meaning of the poem more directly; on the other hand, it facilitated more accurate visual representation of the poem’s ideas. Example: “Echo” by Andreas Embirikos (From the collection The Castles of the Wind, 1934 [18]) «Our footsteps still resound within the forest, with the drone of insects and the heavy drops of frost that drip upon the foliage of the trees. And behold, reverberates within the caves World Journal of Advanced Research and Reviews, 2025, 28(01), 1618-1630 1625 the vibration of every blow of the woodcutters, as they thin the trunks with their axes, holding in their mouths the songs they had learned when they were children, playing hide-and-seek in the forest». Prompt rationale: Although Embirikos is among the central figures of Greek surrealism, this particular poem is relatively accessible, as it begins with vivid imagery that conveys a clear emotional state. The adult speaker’s wandering through the forest evokes memories of childhood, with the forest functioning as a symbolic space of nostalgia and recollection. Caption: “The footsteps of the now elderly poet in the forest of memories follow the echoes of his childhood, when he once played freely in the same place.” Application used: Deep Dream Generator Prompt: An old man is walking in the forest and sees a small child running in front of him. Generated image: Figure 5 The Generated image inspired by the A. Embiriko’s poem: “Echo” 4.3. Development of the Teaching Scenario The teaching scenario developed for these twelve poems of modern Greek poetry followed the official format of the Institute of Educational Policy (ΙΕΠ), the scientific and advisory body of the Greek Ministry of Education. The scenario was structured in twelve stages and included all necessary information for classroom implementation. 4.3.1. Teaching Scenario • Title: Literature through the Lens of Artificial Intelligence: Creating Images to Interpret Poems • Author: Koralia S. Trikalioti – Philologist • Subject Area: Modern Greek Literature • Class/Level: Grade 11 (Upper Secondary School) • Duration: 2 teaching hours • Prerequisites: Familiarity with reading and interpreting poetry; basic digital literacy (computer and internet use). World Journal of Advanced Research and Reviews, 2025, 28(01), 1618-1630 1626 Learning Objectives: Cognitive: Understanding and analyzing poetic motifs (e.g., sea, journey, memory); connecting literary symbols with visual representations. Skills: Enhancing creative expression through AI tools; developing critical thinking about visualizing poetry. Attitudes/Values: Raising awareness of technology as an educational aid; appreciating the aesthetic dimension of literature. Materials and Resources: Computers/tablets with internet access; AI applications (Pixlr as the primary tool, Stable Diffusion, Deep Dream Generator); projector/interactive whiteboard. Class Organization: Group work (3–4 students); plenary presentations. Teaching Process: Introduction (10’): Presentation of selected poems (e.g., Seferis, Kavvadias); discussion of recurring motifs (sea, journey, exile). Activity 1 (20’): Groups design prompts based on motifs; use Pixlr and other tools to generate images. Activity 2 (20’): Students edit their images using Pixlr features (Generative Fill, Remove Background, Expand). Activity 3 (20’): Groups present their images and discuss how these visually translate the poem. Evaluation (15’): Each group relates their image to specific verses or themes; teacher facilitates reflection on AI’s contribution to literary interpretation. Worksheets: Instructions for prompt design; comprehension questions; group notes linking text to image. Comments: The scenario can be extended to additional texts. It enables critical discussion on AI and art, fosters active participation, and enhances engagement through visualization. In this way, the scenario constitutes a complete teaching resource: it supports the comprehension of poetic meaning through visualization, integrates innovative AI technologies, and cultivates students’ creativity alongside collaboration and teamwork. 5. Summary of Contribution This teaching proposal constitutes an innovative contribution to the teaching of Modern Greek Literature, as it represents the first attempt in the Greek context to integrate Artificial Intelligence (AI) tools for the generation of digital images that accompany the interpretation of poetry. By combining traditional archival material, namely twentiethcentury poetry, with state-of-the-art AI technologies, the proposal creates a new pedagogical resource with multiple affordances. Visualization enhances learners’ comprehension of poetic meaning, in line with theories of multimedia learning [8] and dual coding [7] which stress that the joint activation of verbal and visual channels strengthens learning outcomes. For educators, the proposal provides flexible tools while simultaneously fostering engagement with advanced technological applications that are expected to significantly shape the future of education [19] [20]. For students, working with AI-generated images stimulates imagination by translating abstract or symbolic ideas into concrete forms; cultivates critical thinking through evaluation of the relevance of generated images to poetic texts and promotes collaboration, as students design prompts and present their outputs in groups [21]. The proposal thus bridges traditional and innovative practices, demonstrating education as a space where culture and technology co-exist creatively. 5.1. Limitations Despite its considerable advantages, the proposal has certain limitations. Most importantly, the teaching scenario has been developed exclusively at a theoretical level and has not yet been piloted under real classroom conditions. Consequently, no empirical data exist regarding its effectiveness or its acceptance by teachers and students. At the technical level, Stable Diffusion requires significant expertise and configuration; Deep Dream Generator is highly