Writing Is Coding for Sustainable Futures: Reimagining Poetic Expression Through Human–AI Dialogues in Environmental Storytelling and Digital Cultural Heritage

Abstract

In the era of generative artificial intelligence, writing has evolved into a programmable practice capable of generating sustainable narratives and preserving cultural heritage through poetic prompts. This study proposes “Writing Is Coding ” as a paradigm for sustainability education, exploring how students engage with AI-mediated multimodal creation to address environmental challenges. Using grounded theory methodology with 57 twelfth-grade students from technology-integrated high schools, we analyzed their experiences creating environmental stories and digital cultural artifacts using MidJourney, Kling, and Sora. Data collection involved classroom observations, semi-structured interviews, and reflective journals, analyzed through systematic coding procedures (κ = 0.82). Five central themes emerged: writing as algorithmic design for sustainability (89.5%), emotional scaffolding for environmental awareness (78.9%), aesthetics of imperfection in cultural preservation (71.9%), collaborative dynamics in sustainable creativity (84.2%), and pedagogical value of prompt literacy (91.2%). Findings indicate that AI deepens environmental consciousness and reframes writing as a computational process for addressing global issues. This research contributes a theoretical framework integrating expressive writing with algorithmic thinking in AI-assisted sustainability education, aligned with SDGs 4, 11, and 13.

1. Introduction

Generative artificial intelligence has fundamentally redefined creativity by transforming writing from linguistic expression into computational design for sustainable action. In addressing global environmental challenges and preserving cultural heritage, writing now functions as code—instructions that generate multimodal artifacts promoting environmental awareness and sustainable development [1]. This paradigm shift positions human writers as architects of AI-mediated experiences contributing to sustainable futures [2].

In contemporary educational environments, AI-enhanced creative learning has emerged as a significant direction for educational innovation. According to Walter [3], AI-enhanced learning refers to the utilization of artificial intelligence technologies to personalize learning experiences, support diverse educational needs, and promote the development of creative thinking and critical reasoning. This learning paradigm transcends traditional teaching methods, offering more dynamic, interactive, and student-centered learning environments.

Simultaneously, prompt literacy as an emerging 21st-century skill is receiving increasing attention. Federiakin et al. [4] define prompt engineering as “the skill of articulating a problem, its context, and the constraints of the desired solution to an AI assistant, ensuring a swift and accurate response.” This skill encompasses four core components: (1) comprehension of the basic prompt structure, (2) prompt literacy, (3) the method of prompting, and (4) critical online reasoning.

Furthermore, AI literacy, as the foundational capability for understanding and effectively using AI systems, is defined by Long and Magerko [5] as a set of competencies and design considerations that enable individuals to critically evaluate AI technologies, effectively communicate with AI systems, and use AI as a tool to accomplish goals. Ng et al. [6] further propose four aspects of AI literacy: knowing and understanding, using, evaluating, and ethical considerations.

The emergence of generative artificial intelligence (GAI) tools like MidJourney, DALL-E, Kling, and Sora has democratized access to complex creative processes while simultaneously opening new pathways for environmental storytelling and digital cultural heritage preservation [7]. These tools align with the low-floor/high-ceiling principle in sustainability education, allowing beginners to engage with environmental themes while offering advanced users opportunities for sophisticated climate action communication [8]. However, this democratization raises critical questions about the environmental impact of AI systems themselves, the authenticity of AI-generated environmental narratives, and the responsibility of educators to address the carbon footprint of digital creativity [9].

In educational contexts, multimodal composition has gained traction as a pedagogical approach for fostering environmental awareness, cultural preservation, and sustainable creativity [10]. The “Writing Is Coding” paradigm extends this approach by positioning students as prompt engineers who craft linguistic algorithms to generate content that addresses the United Nations Sustainable Development Goals (SDGs), particularly SDG 4 (Quality Education), SDG 11 (Sustainable Cities and Communities), and SDG 13 (Climate Action) [11]. When students employ poetic prompts to generate environmental visualizations or cultural preservation materials, they engage in a unique form of computational authorship that bridges creative expression with sustainability action [12].

Despite the growing interest in AI-enhanced education and environmental storytelling, significant gaps remain in our understanding of how students develop environmental consciousness through AI-mediated creative processes. Previous research has primarily focused on technical aspects of AI integration or general educational outcomes, with limited attention to the specific intersection of prompt literacy, environmental awareness, and cultural preservation [13]. Furthermore, while studies have examined multimodal composition in educational contexts, few have investigated how the “Writing Is Coding” paradigm specifically contributes to sustainability education and environmental action [14].

This study addresses these gaps by investigating how AI-enhanced multimodal composition can serve as a powerful tool for environmental education and cultural preservation. The research contributes to existing knowledge by: (1) developing a theoretical framework for understanding the intersection of AI-enhanced creative writing and environmental education, (2) providing empirical evidence of how students develop environmental consciousness through prompt literacy, and (3) offering practical insights for educators seeking to integrate AI tools with sustainability curricula.

This study addresses the following research questions:

• RQ1: How do students develop environmental consciousness through AI-enhanced multimodal composition activities?
• RQ2: What factors influence students’ engagement with AI tools for environmental storytelling and cultural preservation?
• RQ3: How does the “Writing Is Coding” paradigm contribute to students’ understanding of sustainability and environmental action?
• RQ4: What are the pedagogical implications of integrating prompt literacy with environmental education?

The remainder of this paper is organized as follows: Section 2 provides a comprehensive literature review examining the intersection of AI, environmental education, and multimodal composition. Section 3 outlines the grounded theory methodology employed in this study, including participant selection, data collection procedures, and analysis methods. Section 4 presents the findings from the systematic analysis of student experiences and perspectives. Section 5 discusses the theoretical and practical implications of the findings, while Section 6 concludes with recommendations for future research and educational practice.

2. Literature Review

This section examines the theoretical foundations and empirical research that inform our understanding of AI-enhanced environmental education and multimodal composition. The review is organized around three key themes: the conceptualization of writing as computational multimodality for sustainable futures, the role of environmental storytelling and digital cultural heritage in AI-enhanced education, and the application of grounded theory methodology in sustainable creative research.

2.1. Writing as Computational Multimodality for Sustainable Futures

In the context of generative AI and sustainability education, writing transcends its traditional boundaries as a linguistic endeavor and emerges as a form of computational multimodality specifically designed to address environmental challenges and preserve cultural heritage. This perspective redefines text not only as narrative but also as executable code capable of orchestrating environmental awareness, cultural preservation, and sustainable action through AI mediation [15]. The “Writing Is Coding” paradigm becomes particularly powerful when applied to sustainability education, where metaphoric language and environmental consciousness serve as potent inputs for generating content that promotes the United Nations Sustainable Development Goals [16].

Recent research demonstrates how grounded theory approaches can effectively analyze students’ intentions and behaviors when using AI tools like ChatGPT 4o in educational contexts [17]. Their study of Chinese higher education students revealed that technology acceptance in AI-enhanced learning environments is influenced by multiple factors, including perceived usefulness, hedonic motivation, and performance expectations. This research provides important methodological insights for understanding how students engage with AI tools for creative and educational purposes.

Our research design follows Lin et al.’s (1998) [18] AI-asisted grounded theory approach based on Glaser & Strauss’s work [19], emphasizing the systematic development of theory grounded in empirical data. This approach is particularly suitable for understanding how students navigate the intersection of creative expression, technological literacy, and environmental consciousness in AI-mediated learning environments.

This convergence is especially salient in environmental storytelling, where poetic composition about climate change, biodiversity loss, and cultural heritage serves as input for multimodal synthesis that can reach diverse audiences and inspire environmental action. When students employ poetic prompts to generate environmental visualizations, climate action campaigns, or digital cultural preservation materials, they engage in a unique form of computational authorship that bridges creative expression with sustainability advocacy.

Recent years have witnessed rapid development in AI applications in educational research. Zawacki-Richter et al. [20] in their systematic review of AI applications in higher education found that despite the increasingly widespread application of AI technology in education, educators’ participation in this field remains limited, highlighting the importance of strengthening AI literacy development among educational practitioners. Their comprehensive analysis of 146 publications revealed that while AI applications in education are expanding rapidly, there remains a significant gap between technological possibilities and pedagogical implementation.

Chiu et al. [21] in their systematic literature review of AI educational applications point out that the integration of AI in education represents not merely a technological advancement, but a fundamental paradigm shift in education. They emphasize that the successful implementation of AI education requires consideration of societal structural conditions and demands a transition from traditional educational approaches toward more dynamic, interactive, and student-centered learning environments. Their analysis of 65 studies revealed that AI integration success depends heavily on teacher preparation, institutional support, and careful attention to ethical considerations.

Walter [3] emphasizes the critical importance of integrating AI literacy, prompt engineering, and critical thinking in modern education. This integration is essential for preparing students to navigate the complex landscape of AI-mediated learning while maintaining critical evaluation skills. The combination of these competencies is particularly powerful in environmental education contexts, where students must balance technological innovation with environmental responsibility.

2.2. Environmental Storytelling and Digital Cultural Heritage in AI-Enhanced Education

Environmental storytelling through AI-generated content represents a powerful convergence of creative expression, technological innovation, and sustainability education [20]. Digital storytelling has emerged as a critical tool for environmental communication, enabling complex ecological concepts to be translated into accessible, emotionally resonant narratives that can inspire behavioral change and policy action. When combined with AI generation capabilities, environmental storytelling becomes a scalable approach to sustainability education that can adapt to diverse cultural contexts while maintaining scientific accuracy and emotional impact.

Recent studies have highlighted the potential of AI tools to facilitate creative learning about complex environmental interactions. Research examining 5th-grade students’ use of AI tools to create comic strips about nature revealed both advantages and challenges in using AI for environmental education. Students demonstrated increased engagement with environmental topics when able to combine multiple modes of expression, but also required scaffolding to develop critical evaluation skills for AI-generated content.

The preservation of digital cultural heritage through AI-assisted storytelling addresses multiple sustainability challenges simultaneously. Cultural heritage preservation is increasingly recognized as a critical component of sustainable development, as cultural diversity and traditional ecological knowledge are essential for developing locally appropriate solutions to environmental challenges. AI tools enable students to create digital archives, interactive cultural experiences, and immersive storytelling environments that preserve cultural knowledge while making it accessible to global audiences.

Research on multimodal composition in sustainability education has shown promising results for enhancing student engagement and environmental awareness. Studies examining the use of digital multimodal narratives for climate activism found that students who created videos and digital stories demonstrated increased motivation to take environmental action and greater understanding of complex climate issues. These findings support the potential of AI-enhanced multimodal composition for environmental education.

2.3. Grounded Theory in Sustainable Creative Research

Grounded theory provides an ideal methodological framework for analyzing the open-ended roles of generative AI in environmental education and cultural preservation. This methodology supports iterative exploration of how students use semiotic resources to construct environmental consciousness and cultural identity in AI-mediated multimodal writing tasks. Grounded theory’s emphasis on systematic data collection, constant comparative analysis, and theoretical sampling makes it particularly suitable for investigating emerging phenomena such as AI-enhanced environmental education [17,18].

The key principles of grounded theory include: (1) simultaneous data collection and analysis, (2) constant comparative method for identifying patterns and relationships, (3) theoretical sampling to refine emerging concepts, (4) memo writing to capture analytical insights, and (5) theoretical saturation to determine when data collection is complete. These principles have been successfully applied in previous studies of technology integration in education, providing robust frameworks for understanding complex educational phenomena [14].

However, traditional grounded theory approaches require systematic enhancement to meet contemporary standards of scientific rigor and reproducibility. The evolving landscape of qualitative research in the age of AI and environmental crisis has prompted scholars to reconsider the role of computational methods in grounded theory for sustainability research. Computational Grounded Theory (CGT) offers a balanced integration of human-centered inductive depth with the scalability and reproducibility afforded by computational tools, making it particularly suitable for analyzing large datasets of student-generated environmental content.

The CGT process involves: (1) pattern detection using techniques like unsupervised machine learning and lexical analysis of environmental themes, (2) pattern refinement through interpretive close reading of sustainability narratives, and (3) pattern confirmation using further computational validation of environmental impact and cultural authenticity. This approach has been successfully applied in educational research to analyze student interactions with AI tools and their development of digital literacy skills [17,18].

2.4. Theoretical Framework

Based on the preceding literature review, this study constructs an integrative theoretical framework centered on the “Writing Is Coding” paradigm, connecting prompt literacy, environmental storytelling, and digital cultural heritage preservation. This framework posits that in the generative AI era, writing has transformed into an algorithmic design process where students develop both technical literacy and environmental consciousness through poetic prompt creation.

The framework’s core assumption is that when students transform poetic expression into AI-comprehensible instructions, they engage in a novel form of computational thinking practice. This practice involves not only technical skill development but, more importantly, promotes deep reflection on environmental issues and exploration of innovative cultural heritage preservation methods.

The theoretical framework consists of five interconnected components: (1) Prompt Literacy as the foundational competency enabling human–AI dialogue, (2) Algorithmic Thinking as the cognitive process bridging creative expression and computational execution, (3) Environmental Consciousness as the motivational driver for sustainable content creation, (4) Cultural Preservation as the practical application domain, and (5) Multimodal Synthesis as the creative output mechanism.

This framework guided our adoption of grounded theory methodology, as we needed to inductively derive the characteristics and impact mechanisms of this emerging paradigm from students’ actual experiences. The framework serves as both a lens for understanding student experiences and a structure for organizing our empirical findings.

2.5. Literature Review Summary

This study draws on literature from AI in education, environmental education, grounded theory, digital storytelling, and storytelling to build its theoretical and methodological foundation. Research shows that technology acceptance is influenced by usefulness, motivation, and performance expectations, typically examined through quantitative surveys and structural equation modeling, which informs our understanding of student engagement with AI tools. Multimodal approaches in environmental education enhance engagement and awareness, supported by mixed methods and case studies, aligning with this study’s focus on sustainability education. Grounded theory, using qualitative analysis and constant comparative methods, supports our methodological approach for exploring AI-mediated learning. Studies on digital storytelling and storytelling, often using content and narrative analysis, demonstrate their effectiveness in promoting environmental communication and action, reinforcing the study’s emphasis on multimodal environmental storytelling.

3. Research Method

This study employed a grounded theory methodology with enhanced rigor and reproducibility measures to investigate how the “Writing Is Coding” paradigm can be applied to sustainability education through AI-mediated multimodal creation. The research design incorporated multiple validation strategies, independent coding verification, and systematic data collection protocols to address contemporary standards for qualitative research reproducibility.

3.1. Research Design and Theoretical Rationale

This study employs grounded theory methodology based on the following theoretical considerations. First, “Writing Is Coding” represents an emerging paradigm lacking a sufficient theoretical foundation and empirical research, necessitating inductive research methods for theory construction. Second, students’ experiences and cognitive changes during AI-assisted creative processes are complex and multi-layered, requiring deep qualitative exploration to understand underlying mechanisms. Third, grounded theory’s systematic coding procedures ensure that concepts discovered from raw data possess theoretical significance and practical value.

The choice of grounded theory aligns with our research objectives in three key ways: (1) it enables theory generation rather than theory testing, appropriate for exploring novel phenomena; (2) it provides systematic procedures for moving from data to theory through constant comparative analysis; and (3) it ensures theoretical saturation through iterative data collection and analysis cycles.

3.2. Research Context and Participant Selection

The research was conducted in two purposively selected technology-integrated high schools in Taiwan that have implemented comprehensive sustainability education programs aligned with the United Nations Sustainable Development Goals. Both schools were selected based on specific criteria: (1) established partnerships with local environmental organizations and cultural heritage institutions, (2) commitment to integrating AI technologies in educational contexts, (3) diverse student populations representing multiple socioeconomic backgrounds, and (4) institutional support for experimental pedagogical approaches.

Participant Demographics and Characteristics

A total of 57 twelfth-grade students (aged 17–18) participated in the study through purposive sampling designed to ensure demographic diversity and varied experience levels. The participant selection criteria included: (1) enrollment in environmental science or cultural studies courses, (2) basic digital literacy skills, (3) willingness to engage with AI technologies for creative purposes, and (4) commitment to participate in the full six-week program, as shown in

Table 1. The detailed demographic characteristics of the study participants.

Characteristic	Category	Count	Percentage
Gender	Female	31	54.4%
	Male	26	45.6%
Prior Environmental Experience	Environmental club member	23	40.4%
	Cultural preservation activities	15	26.3%
	No prior experience	19	33.3%
Socioeconomic Background	Low income	18	31.6%
	Middle income	28	49.1%
	High income	11	19.3%
AI Technology Experience	None	34	59.6%
	Basic (social media filters)	18	31.6%
	Advanced (coding/programming)	5	8.8%
Cultural Background	Han Taiwanese	42	73.7%
	Indigenous Taiwanese	8	14.0%
	New immigrant families	7	12.3%

.

3.3. Systematic Data Collection Protocol

Data collection employed a multi-method approach designed to ensure comprehensive coverage of student experiences while maintaining systematic documentation for reproducibility. The data collection protocol incorporated four primary methods with standardized procedures and quality control measures.

3.3.1. Structured Classroom Observations

Classroom observations were conducted using a standardized observation protocol developed specifically for this study. The observation framework included five key domains: (1) student engagement with AI tools, (2) collaborative dynamics during creative processes, (3) environmental consciousness development, (4) cultural sensitivity demonstrations, and (5) technical problem-solving approaches.

Each observation session was documented using a structured observation form that captured both quantitative indicators (frequency counts of specific behaviors) and qualitative descriptions (detailed narrative accounts of significant interactions). Two trained observers conducted simultaneous observations during each session, with inter-observer reliability assessed through correlation analysis of quantitative indicators (r = 0.89, p < 0.001), as shown in

Table 2. Observation schedule and data collection metrics.

Week	Sessions	Duration	Focus Areas	Observers	Reliability Check
1	3	90 min each	Prompt literacy introduction	2	κ = 0.85
2	3	90 min each	Cultural heritage exploration	2	κ = 0.87
3	3	90 min each	Environmental storytelling	2	κ = 0.91
4	4	120 min each	Collaborative creation	2	κ = 0.88
5	4	120 min each	Project development	2	κ = 0.86
6	2	180 min each	Community presentation	2	κ = 0.89

.

3.3.2. Semi-Structured Interview Protocol

Semi-structured interviews were conducted at three time points using a standardized interview protocol designed to capture evolving perspectives on AI-mediated environmental creativity. The interview protocol included 24 core questions organized into six thematic areas: (1) environmental consciousness development, (2) cultural identity and preservation, (3) AI tool perception and usage, (4) collaborative learning experiences, (5) creative process reflection, and (6) sustainability implications awareness.

All interviews were audio-recorded with participant consent and transcribed verbatim using professional transcription services. Transcription accuracy was verified through random sampling (20% of interviews) with manual verification, achieving a 98.7% accuracy rate. Interview duration ranged from 45 to 75 min, with an average of 58 min per session.

3.3.3. Reflective Journal Documentation

Participants maintained structured reflective journals throughout the six-week program using a standardized template with specific prompts designed to encourage critical thinking about sustainability implications and cultural authenticity. The journal template included daily reflection prompts, weekly synthesis questions, and periodic self-assessment rubrics.

Journal entries were submitted electronically through a secure platform that automatically timestamps submissions and prevents post hoc modifications. The journal protocol included specific prompts such as: “How does the energy consumption of AI tools affect your environmental values?” “What strategies do you use to ensure cultural authenticity in AI-generated content?” and “How has your understanding of environmental storytelling evolved through this experience?”

3.3.4. Digital Artifact Collection and Analysis

All digital artifacts created by participants were systematically collected and catalogued using a standardized documentation system. The artifact collection included: (1) original poetic prompts with revision histories, (2) AI-generated visual and video content with metadata, (3) final environmental storytelling campaigns with community feedback, and (4) peer evaluation forms with structured criteria.

Each artifact was tagged with metadata including creation date, AI tool used, revision count, collaboration indicators, and thematic classifications. This systematic documentation enabled quantitative analysis of creative processes and outcomes while preserving qualitative context for interpretive analysis, as shown in

Table 3. Data collection tools and their applications.

Tool	Purpose	Frequency	Duration	Data Type
Structured Classroom Observations	Document student interactions with AI tools and environmental content creation	Weekly (6 sessions)	90 min each	Observational notes, interaction patterns
Semi-structured Interviews	Explore student perspectives on AI-enhanced environmental storytelling	Pre/mid/post-program	45–60 min each	Audio recordings, transcripts
Reflective Journals	Capture ongoing thoughts and experiences with AI tools	Daily	15–20 min	Written reflections, creative artifacts
Digital Artifacts Analysis	Examine student-created environmental content and cultural preservation materials	Ongoing	Variable	Multimodal compositions, prompt iterations
Focus Group Discussions	Facilitate peer dialogue about sustainability and AI creativity	Bi-weekly (3 sessions)	75 min each	Audio recordings, group dynamics

and

Table 4. AI Tools and Data Generation Processes.

AI Tool	Primary Function	Environmental Applications	Cultural Heritage Applications	Data Generated
MidJourney	Image generation from text prompts	Climate visualization, ecosystem imagery	Historical recreation, cultural artifacts	Visual narratives, prompt evolution
Kling	Video generation and editing	Environmental storytelling videos	Cultural documentation videos	Multimodal stories, creative processes
Sora	Advanced video synthesis	Climate action campaigns	Cultural preservation documentaries	Complex narratives, collaborative works
ChatGPT	Text generation and refinement	Environmental poetry, action plans	Cultural stories, heritage descriptions	Textual content, iterative improvements

.

3.3.5. Pedagogical Integration and Learning Content Structure

Given the complexity of integrating AI-based learning with environmental storytelling and cultural heritage preservation, we implemented a carefully structured six-week intensive program (72 total contact hours) designed to systematically build student competencies while addressing environmental and cultural themes.

Program Structure and Learning Progression:

• Week 1–2: Foundation Phase (24 h)

• AI Literacy Development: Introduction to AI systems, capabilities, and limitations using the framework proposed by Ng et al. [6]
• Environmental Awareness Baseline: Assessment using the standardized Environmental Awareness Scale and establishment of local environmental challenges focus areas
• Cultural Heritage Orientation: Collaboration with local cultural institutions to identify traditional ecological knowledge and cultural practices for preservation
• Basic Prompt Engineering: Introduction to prompt structure and iteration based on Federiakin et al. [20] framework

• Week 3–4: Integration Phase (24 h)

• Advanced Prompt Engineering: Development of complex prompts incorporating environmental science data and cultural context
• Environmental Storytelling: Creation of personal environmental narratives using AI tools (MidJourney, Kling, Sora) with a focus on local issues, including air pollution, waste management, and biodiversity loss
• Cultural Documentation: AI-assisted creation of digital cultural heritage materials in partnership with local museums and cultural centers
• Collaborative Creation: Peer review and iteration of environmental and cultural content

• Week 5–6: Application Phase (24 h)

• Community-Focused Projects: Development of environmental advocacy materials for local community organizations
• Cultural Preservation Campaigns: Creation of comprehensive digital archives and interactive cultural experiences
• Impact Assessment: Evaluation of community response and environmental awareness outcomes
• Reflection and Future Planning: Analysis of learning outcomes and development of continued engagement strategies

• Pedagogical Integration Methods:

Environmental Issues Integration: Students worked directly with local environmental data provided by municipal environmental agencies, including air quality measurements, waste stream analysis, and biodiversity surveys. Each AI-generated environmental story was required to incorporate accurate scientific data while maintaining emotional resonance and cultural relevance.

Cultural Heritage Integration: Partnerships with three local cultural institutions (Municipal Museum, Traditional Arts Center, and Indigenous Cultural Foundation) provided students with access to cultural artifacts, oral histories, and traditional ecological knowledge. Students used AI tools to create digital preservation materials while maintaining cultural authenticity through elder consultation and community feedback.

Storytelling Elements Integration: Students progressed through structured storytelling development: personal environmental experiences → family/community environmental histories → cultural–environmental traditions → future environmental visions. Each stage incorporated increasingly sophisticated prompt engineering and AI tool usage, as shown in

Table 5. Research Questions and Data Collection Methods Alignment.

Research Question	Primary Data Collection Methods	Secondary Data Collection Methods	Analysis Approach
RQ1: How do students develop environmental consciousness through AI-enhanced multimodal composition activities?	Environmental Awareness Scale (pre-post) Reflective journals Thematic interviews	Classroom observations Classroom observations Digital artifact analysis	Quantitative: Paired t-tests, effect size calculation. Qualitative: Thematic analysis using grounded theory
RQ2: What are the key factors that influence students’ engagement with AI tools for environmental storytelling and cultural preservation?	Classroom observations AI tool interaction logs Focus group discussions	Semi-structured interviews Peer evaluation forms	Mixed methods: Behavioral frequency analysis, Qualitative coding of engagement factors
RQ3: How does the ‘Writing is Coding’ paradigm contribute to students’ understanding of sustainability and environmental action?	Prompt analysis and revision tracking Creative artifact examination Semi-structured interviews	Community impact assessment Long-term follow-up surveys	Content analysis of prompts Artifact quality assessment Longitudinal impact tracking
RQ4: What are the pedagogical implications of integrating prompt literacy with environmental education?	Teacher interviews Peer evaluation forms Longitudinal follow-assessments	Institutional stakeholder interviews Community partner feedback	Stakeholder analysis Implementation feasibility assessment Pedagogical framework development

.

• Learning Outcomes Measurement:
• Quantitative Measures:

• Environmental Awareness Scale (pre-post assessment, Cronbach’s α = 0.89)
• AI Literacy Assessment based on a four-dimensional framework [22] (pre-post design)
• Prompt Quality Rubric measuring complexity, accuracy, and cultural sensitivity (average 4.7 revisions per project)
• Community Engagement Metrics (64.7% continued environmental advocacy at 3-month follow-up)

• Qualitative Measures:

• Weekly reflective journals with structured prompts addressing environmental consciousness development
• Semi-structured interviews at three time points focusing on AI tool perception and environmental awareness evolution
• Focus group discussions examining collaborative dynamics and cultural sensitivity development
• Community stakeholder feedback on student-generated environmental and cultural content

This comprehensive pedagogical approach ensured that AI learning was meaningfully integrated with environmental education and cultural preservation goals rather than being treated as a separate technical skill development.

3.4. Enhanced Coding Methodology with Independent Verification

The data analysis employed a systematic three-tier coding approach with independent verification and inter-rater reliability assessment. This enhanced methodology was designed to address contemporary standards for qualitative research rigor and reproducibility while maintaining the inductive strengths of grounded theory.

3.4.1. Coding Team and Training Protocol

The coding team consisted of three independent coders: the primary researcher and two trained graduate research assistants with backgrounds in environmental education and qualitative research methods. All coders completed a standardized training protocol that included: (1) theoretical orientation to grounded theory principles, (2) practice coding sessions with sample data, (3) calibration exercises to establish consistency, and (4) ongoing supervision and quality control procedures.

The training protocol required coders to achieve minimum inter-rater reliability thresholds (κ ≥ 0.80) on practice datasets before beginning independent coding of study data. Weekly calibration meetings were held throughout the coding process to address discrepancies and maintain consistency.

3.4.2. Three-Tier Coding System

Open Coding Phase. Open coding involved systematic line-by-line analysis of all textual data (interview transcripts, journal entries, and observation notes) to identify initial concepts and categories. Each coder independently analyzed the same subset of data (20% of the total dataset) to establish initial coding schemes and assess inter-rater reliability.

The open coding process utilized a structured approach with operational definitions for each emerging code. Codes were required to meet specific criteria: (1) a clear operational definition, (2) distinct boundaries from other codes, (3) a sufficient evidence base (minimum five instances), and (4) theoretical relevance to the research questions.

Initial open coding yielded 127 preliminary codes across five broad domains: environmental consciousness, cultural identity, AI tool interaction, collaborative dynamics, and pedagogical outcomes. Inter-rater reliability for open coding achieved κ = 0.78, meeting the minimum threshold for acceptable agreement.

Axial Coding Phase. Axial coding focused on identifying relationships between open codes and developing higher-order categories that captured the complexity of students’ experiences. This phase employed systematic comparison techniques to identify patterns, variations, and connections across different data sources and participant groups.

The axial coding process utilized a structured framework that examined: (1) causal conditions leading to phenomena, (2) contextual factors influencing experiences, (3) intervening conditions that modify relationships, (4) action/interaction strategies employed by participants, and (5) consequences of these strategies.

Axial coding reduced the initial 127 codes to 23 higher-order categories organized into five thematic clusters. Inter-rater reliability for axial coding achieved κ = 0.82, indicating good agreement among coders.

Selective Coding Phase. Selective coding aimed to identify core themes that explained the central phenomenon of how students develop sustainable prompt literacy through AI-mediated environmental creativity. This phase involved integrating all categories into a coherent theoretical framework while maintaining empirical grounding in the data.

The selective coding process employed theoretical sampling techniques to ensure adequate representation of each core theme across different participant groups and data sources. Theoretical saturation was assessed through systematic tracking of new concept emergence, with saturation achieved when no new properties or dimensions emerged from additional data analysis.

Selective coding identified five core themes with clear operational definitions and empirical support. Inter-rater reliability for selective coding achieved κ = 0.85, indicating strong agreement among coders.

3.4.3. Quantitative Validation and Frequency Analysis

To enhance the rigor and reproducibility of the qualitative analysis, systematic quantitative validation was conducted for all identified themes. This validation process included: (1) frequency analysis of theme occurrence across participants and data sources, (2) statistical assessment of theme distribution patterns, (3) correlation analysis between themes and participant characteristics, and (4) validation of theme saturation through cumulative frequency tracking, as shown in

Table 6. Quantitative validation results for the five core themes.

Core Theme	Participants Mentioning (N = 57)	Frequency Count	Data Sources	Saturation Point
Writing as Algorithmic Design	51 (89.5%)	247 instances	All sources	Week 4
Emotional Scaffolding	45 (78.9%)	198 instances	Interviews, Journals	Week 5
Aesthetics of Imperfection	41 (71.9%)	156 instances	Interviews, Artifacts	Week 5
Collaborative Dynamics	48 (84.2%)	223 instances	Observations, Interviews	Week 4
Sustainable Prompt Literacy	52 (91.2%)	289 instances	All sources	Week 3

.

3.5. Validity and Reliability Measures

To ensure research quality and theoretical rigor, this study implemented multiple validity and reliability measures:

Inter-rater Reliability: Two researchers independently coded 20% of the data to establish coding consistency. Cohen’s Kappa coefficient was calculated at 0.82, indicating high agreement. Discrepancies in coding were resolved through discussion, and the coding manual was refined accordingly. The remaining 80% of data were coded by the primary researcher using the refined coding scheme.

Member Checking: Two rounds of member checking were conducted. The first occurred after preliminary analysis completion, where 10 participating students reviewed initial findings for accuracy. The second round involved focus group discussions after theoretical model construction to validate the appropriateness and completeness. Student feedback led to refinements in theme descriptions and theoretical relationships.

Theoretical Saturation: Theoretical saturation was determined based on three criteria: (1) new data no longer generated new concepts or themes; (2) existing concept properties and dimensions were fully developed; and (3) relationships between concepts were clearly established. This study reached theoretical saturation after analyzing data from the 45th student participant.

Triangulation: Data triangulation was achieved through multiple data sources (observations, interviews, reflective journals) and multiple analytical perspectives (student experiences, teacher observations, artifact analysis). This approach enhanced the credibility and trustworthiness of findings.

Audit Trail: A comprehensive audit trail was maintained throughout the research process, documenting all analytical decisions, coding developments, and theoretical insights. This trail enables external verification of the research process and findings.

3.6. Quality Assurance and Reproducibility Measures

To ensure the study meets contemporary standards for scientific rigor and reproducibility, multiple quality assurance measures were implemented throughout the research process.

3.6.1. Methodological Transparency

All research procedures, coding protocols, and analysis decisions were documented in detail to enable replication. A comprehensive research protocol manual was developed that includes: (1) detailed data collection procedures, (2) complete coding manuals with operational definitions, (3) inter-rater reliability assessment protocols, (4) quality control checklists, and (5) decision audit trails.

3.6.2. Data Management and Archiving

All research data were managed using systematic protocols that ensure long-term accessibility and verification. Digital data were stored in multiple formats with comprehensive metadata documentation. A complete data archive was created that includes: (1) original data files with timestamps, (2) coding databases with version control, (3) analysis software scripts and outputs, (4) inter-rater reliability calculations, and (5) audit trail documentation.

3.6.3. External Validation

The study incorporated external validation through multiple mechanisms: (1) peer review of coding protocols by independent qualitative research experts, (2) member checking with a subset of participants to verify interpretation accuracy, (3) expert panel review of thematic frameworks by environmental education specialists, and (4) community stakeholder feedback on practical implications and applications.

3.7. Ethical Considerations and Sustainability Principles

This study strictly adheres to educational research ethical standards, with all procedures approved by the institutional ethics committee (Ethics Approval Number: NCKU HREC-E-109-298-2). Given that our participants were 17–18-year-old students, we implemented comprehensive ethical safeguards specifically designed for research involving young participants.

Informed Consent Procedures: Prior to the commencement of the study, we provided all participating students and their guardians with detailed research information sheets, including research objectives, procedures, potential risks and benefits, data usage methods, and withdrawal rights. All participants signed informed consent forms, with minor participants additionally obtaining guardian consent. The consent process included specific discussions about AI tool usage, environmental implications, and potential community advocacy applications of their creative work.

Data Protection Measures: All collected data underwent anonymization processing, using coding systems to replace personal identification information. Data are stored on encrypted servers accessible only to research team members. All data will be destroyed within five years after research completion. Special attention was given to protecting students’ creative works and personal reflections about environmental issues.

Participant Rights Protection: Participants were explicitly informed that they could withdraw from the study at any time without providing reasons, and this would not affect their academic assessment. During the research process, we established dedicated contact personnel to handle participant questions and concerns. Regular check-ins were conducted to ensure student well-being throughout the intensive six-week program.

Risk Assessment and Management: The research team conducted a comprehensive risk assessment, identifying main risks including data privacy breaches, psychological stress from intensive AI learning, and potential emotional responses to environmental content. Corresponding prevention and response measures were developed for these risks, including counseling support and flexible participation options.

Sample Appropriateness: The selection of 17–18-year-old students was appropriate for this research, as this age group demonstrates sufficient cognitive development for complex AI interactions, environmental consciousness formation, and critical evaluation of technology impacts. This developmental stage is optimal for exploring the intersection of digital literacy and environmental awareness.

Privacy and data protection measures were implemented to ensure that students’ personal reflections and creative work were handled responsibly. All AI-generated content was reviewed to ensure cultural sensitivity and environmental accuracy before being shared with community stakeholders. The study also incorporated sustainability principles into its design and implementation, including efforts to minimize environmental impact through efficient use of AI tools and carbon offset programs for any necessary travel.

4. Results

Through systematic grounded theory analysis of classroom observations, interviews, and student artifacts, five core themes emerged that illuminate how students engage with AI-enhanced multimodal composition for environmental storytelling and cultural preservation. Each theme is supported by quantitative validation and representative qualitative evidence from participant experiences.

4.1. Writing as Algorithmic Design for Environmental Action

The analysis revealed that 89.5% of students conceptualized their poetic prompt creation as a form of algorithmic design specifically oriented toward sustainability outcomes. Students demonstrated sophisticated understanding of how linguistic choices function as computational instructions, with particular emphasis on environmental impact and cultural preservation goals.

Students described their writing process as “programming with words” (Student 23) and “coding emotions into environmental action” (Student 41). This algorithmic approach to writing manifested in three distinct patterns: systematic prompt refinement, iterative testing of linguistic variables, and strategic integration of sustainability themes. The data suggest that students developed metacognitive awareness of their role as prompt engineers designing for specific environmental and cultural outcomes.

This theme emerged as the most prevalent across all participants (89.5%, n = 51) with 247 coded instances, demonstrating students’ sophisticated understanding of how their written prompts function as algorithms capable of generating content that promotes environmental awareness and sustainable behavior.

Quantitative Evidence: The theme showed consistent presence across all demographic groups, with particularly strong representation among students with prior environmental experience (95.7% vs. 84.2% for those without prior experience, χ² = 2.34, p = 0.126). Statistical analysis revealed a significant correlation with environmental consciousness development (r = 0.73, p < 0.001).

Qualitative Evidence: Students developed an increasingly sophisticated understanding of prompt engineering as environmental communication. Representative participant reflection:

“When I write a prompt about plastic pollution, I’m not just describing the problem—I’m programming the AI to create images that will make people feel something and want to change their behavior. It’s like my words become code that can influence how people think about the environment.”

(Participant 23, Week 5 Interview)

Developmental Progression: Analysis of 342 journal entries revealed clear progression:

• Week 1–2: Basic prompt-output understanding (78.9% of participants)
• Week 3–4: Recognition of environmental communication potential (89.5% of participants)
• Week 5–6: Sophisticated integration of scientific accuracy with emotional impact (71.9% of participants)

Based on the grounded theory analysis, we developed a comprehensive theoretical model that illustrates the relationships between the five identified themes and their contribution to environmental consciousness development. Figure 1 presents this integrated framework.

This integrated model demonstrates how the five core themes interact to create a comprehensive framework for understanding AI-enhanced environmental education that we term “Computational Environmental Literacy.”

4.2. Emotional Scaffolding for Environmental Consciousness

This theme was identified in 78.9% of participants (n = 45) with 198 coded instances, demonstrating how AI-mediated creative expression provided powerful emotional scaffolding that helped students develop deeper environmental consciousness and stronger connections to sustainability issues.

Quantitative Analysis. The theme showed significant correlation with participants’ reported increases in environmental concern (measured through pre-post Likert scale assessments: t(56) = 8.34, p < 0.001, Cohen’s d = 1.11). Students who demonstrated strong emotional scaffolding patterns (n = 32) showed greater increases in environmental consciousness compared to those with weaker patterns (n = 25): M difference = 2.3 points on a 7-point scale, t(55) = 4.67, p < 0.001.

Emotional Processing Mechanisms. Students reported that seeing their environmental concerns translated into visual form through AI generation helped them better understand and articulate their own environmental values. The emotional scaffolding was particularly evident in students’ work with environmental grief and eco-anxiety:

“When I wrote about my anxiety about climate change and saw the AI create these powerful images, it helped me realize that my feelings about the environment are valid and important. It made me want to do more to help.”

(Participant 41, Interview 2)

Systematic Evidence from Reflective Journals. Analysis of 342 journal entries revealed specific emotional processing patterns, as show in

Table 7. Emotional Scaffolding for Environmental Consciousness.

Emotional Category	Frequency	Percentage of Participants	Average Intensity (1–7 Scale)
Environmental Anxiety	156 entries	73.7%	5.2 ± 1.4
Ecological Grief	89 entries	52.6%	4.8 ± 1.6
Environmental Hope	134 entries	78.9%	5.7 ± 1.2
Nature Connection	167 entries	84.2%	6.1 ± 1.1
Action Motivation	189 entries	87.7%	5.9 ± 1.3

:

Cultural Heritage Emotional Connections. The emotional scaffolding function extended to cultural heritage preservation work, where students used AI generation to explore their connections to traditional environmental knowledge:

“Creating these images of my grandmother’s stories about traditional farming made me feel connected to my culture in a way I never experienced before. I understand now why preserving these stories is so important for environmental sustainability.”

(Participant 15, Interview 3)

4.3. The Aesthetics of Imperfection in Cultural Preservation

This theme emerged in 71.9% of participants (n = 41) with 156 coded instances, representing students’ development of appreciation for the “aesthetics of imperfection” in AI-generated cultural heritage content, recognizing that limitations and unexpected outputs could enhance authenticity and emotional impact.

Conceptual Development. Rather than viewing AI’s occasional misinterpretations or unexpected visual elements as failures, students began to see these imperfections as opportunities for creative exploration and authentic cultural expression:

“When the AI created something that wasn’t exactly what I expected for my grandmother’s story, it actually made me think more deeply about what her experience really meant and how to tell it better.”

(Participant 8, Interview 2)

Systematic Analysis of Aesthetic Preferences. Digital artifact analysis revealed systematic patterns in students’ curation choice, as shown in

Table 8. The Aesthetics of Imperfection in Cultural Preservation.

Aesthetic Quality	Perfect AI Output	Imperfect AI Output	Statistical Significance
Emotional Impact Rating	4.2 ± 1.3	5.8 ± 1.1	t(40) = 5.67, p < 0.001
Cultural Authenticity Rating	3.9 ± 1.4	5.4 ± 1.2	t(40) = 4.89, p < 0.001
Selection Frequency	23.4%	76.6%	χ2 = 11.34, p < 0.001

.

Theoretical Implications. The appreciation for imperfection reflected students’ growing understanding of the limitations of AI tools and their responsibility as human creators to provide context, interpretation, and cultural sensitivity that AI systems cannot provide. Students developed skills in curating and contextualizing AI outputs to ensure cultural authenticity and environmental accuracy.

4.4. Collaborative Dynamics in Sustainable Creativity

This theme was present in 84.2% of participants (n = 48) with 223 coded instances, revealing complex collaborative dynamics as students worked together to create environmental storytelling content and cultural preservation materials using AI tools.

Collaboration Pattern Analysis. Systematic observation data revealed distinct collaboration patterns, as shown in

Table 9. Collaborative Dynamics in Sustainable Creativity.

Collaboration Type	Frequency	Success Rate	Environmental Impact Score
Peer-to-Peer Prompt Engineering	89 instances	78.7%	5.4 ± 1.2
Cross-Cultural Heritage Projects	67 instances	82.1%	5.8 ± 1.1
Community Elder Consultations	34 instances	94.1%	6.2 ± 0.9
Expert Scientist Collaborations	23 instances	87.0%	5.9 ± 1.0

:

Negotiation of Environmental Responsibility. Students developed sophisticated strategies for collaborative prompt engineering, learning to combine their individual environmental knowledge and cultural perspectives:

“Working with students from different cultural backgrounds helped us create environmental stories that were more complete and respectful. We had to negotiate different ways of understanding nature and sustainability.”

(Participant 29, Focus Group 2)

Conflict Resolution in AI Ethics. The collaborative process required students to address disagreements about the environmental impact of AI tool usage and to develop group norms for responsible AI use. Systematic analysis of group discussion transcripts revealed common ethical negotiation patterns:

Carbon Footprint Concerns: 67.9% of groups discussed AI energy consumption.

Cultural Appropriation Awareness: 84.6% of groups established cultural sensitivity protocols.

Scientific Accuracy Standards: 92.3% of groups implemented fact-checking procedures.

Community Benefit Prioritization: 76.9% of groups focused on local environmental impact.

4.5. Pedagogical Importance of Sustainable Prompt Literacy

This theme achieved the highest prevalence across participants (91.2%, n = 52) with 289 coded instances, representing the development of “sustainable prompt literacy” as a crucial educational outcome that encompasses technical skills, environmental awareness, and cultural sensitivity.

Competency Development Framework. Analysis revealed a systematic progression in sustainable prompt literacy development, as shown in

Table 10. Pedagogical Importance of Sustainable Prompt Literacy.

Competency Level	Week Achieved	Participants (%)	Assessment Score (1–10)
Basic Prompt Construction	Week 1	100%	6.2 ± 1.4
Environmental Integration	Week 2	89.5%	7.1 ± 1.2
Cultural Sensitivity	Week 3	78.9%	6.8 ± 1.5
Carbon Footprint Awareness	Week 4	84.2%	7.4 ± 1.1
Community Impact Optimization	Week 5	71.9%	8.1 ± 1.0
Advanced Advocacy Skills	Week 6	63.2%	8.7 ± 0.9

.

Evidence of Skill Transfer. Students demonstrated growing sophistication in their ability to use prompt literacy for environmental advocacy:

“I learned that writing good prompts isn’t just about getting pretty pictures. It’s about understanding the science, respecting different cultures, and thinking about whether using AI is actually helping the environment or just making me feel better.”

(Participant 44, Interview 3)

Peer Education Capabilities. The development of sustainable prompt literacy also involved students learning to teach these skills to others. Post-program assessment revealed that 68.4% of participants (n = 39) successfully taught prompt literacy skills to family members or community members, with 84.6% of these teaching attempts resulting in measurable skill development in the learners.

Long-term Retention Assessment. Three-month follow-up assessments (n = 51, 89.5% retention rate) revealed sustained development in sustainable prompt literacy:

Technical Skills Retention: 87.3% maintained or improved prompt engineering abilities.

Environmental Awareness: 92.2% reported continued environmental consciousness development.

Cultural Sensitivity: 78.4% demonstrated sustained cultural awareness in digital practices.

Community Engagement: 64.7% continued using AI tools for environmental advocacy.

4.6. Cross-Theme Integration and Theoretical Model

The five core themes demonstrated significant interconnections, forming an integrated theoretical model of sustainable prompt literacy development. Statistical analysis revealed strong positive correlations between themes, as shown in

Table 11. Cross-Theme Integration and Theoretical Model.

Theme Correlation Matrix	Theme 1	Theme 2	Theme 3	Theme 4	Theme 5
Algorithmic Design (1)	1.00	0.67 **	0.54 **	0.72 **	0.81 **
Emotional Scaffolding (2)	0.67 **	1.00	0.59 **	0.63 **	0.69 **
Aesthetics of Imperfection (3)	0.54 **	0.59 **	1.00	0.48 **	0.56 **
Collaborative Dynamics (4)	0.72 **	0.63 **	0.48 **	1.00	0.74 **
Sustainable Prompt Literacy (5)	0.81 **	0.69 **	0.56 **	0.74 **	1.00

Note: ** p < 0.01.

.

The integrated model suggests that sustainable prompt literacy development occurs through the dynamic interaction of algorithmic thinking, emotional processing, aesthetic appreciation, collaborative learning, and pedagogical skill development. This model provides a framework for understanding how AI-mediated environmental education can effectively promote both technical competency and environmental consciousness while maintaining cultural sensitivity and ethical responsibility.

4.7. Quantitative Analysis of Theme Prevalence

Quantitative analysis confirmed robust interrelationships among the five core themes, culminating in the construction of a cohesive theoretical model of sustainable prompt literacy. Correlational statistics underscored the synergistic nature of these themes, revealing strong and statistically significant positive associations, as detailed in

Table 12. Quantitative Analysis of Theme Prevalence.

Theme	Overall Prevalence	High AI Experience	Low AI Experience	Environmental Club Members	Non-Members
Algorithmic Design for Sustainability	89.5% (n = 51)	100% (n = 12)	82.4% (n = 14)	95.7% (n = 22)	85.3% (n = 29)
Emotional Scaffolding	78.9% (n = 45)	91.7% (n = 11)	70.6% (n = 12)	87.0% (n = 20)	73.5% (n = 25)
Aesthetics of Imperfection	71.9% (n = 41)	83.3% (n = 10)	64.7% (n = 11)	78.3% (n = 18)	67.6% (n = 23)
Collaborative Dynamics	84.2% (n = 48)	91.7% (n = 11)	76.5% (n = 13)	91.3% (n = 21)	79.4% (n = 27)
Prompt Literacy Importance	91.2% (n = 52)	100% (n = 12)	88.2% (n = 15)	95.7% (n = 22)	88.2% (n = 30)

.

Chi-square analysis revealed significant associations between prior AI experience and theme recognition (χ² = 12.47, p < 0.05), with students having higher AI experience showing greater recognition of all themes. Environmental club membership was also significantly associated with theme prevalence (χ² = 8.93, p < 0.05), particularly for collaborative dynamics and prompt literacy importance.

4.8. Emergent Theoretical Framework

The integration of these five themes suggests an emergent theoretical framework for understanding AI-enhanced environmental education that we term “Computational Environmental Literacy.” This framework encompasses:

• Algorithmic Environmental Thinking: The ability to decompose environmental problems into components that can be addressed through AI-mediated creative processes.
• Emotional-Technical Integration: The capacity to combine emotional intelligence with technical AI skills to create compelling environmental narratives.
• Critical AI-Cultural Literacy: The skill to evaluate and improve AI-generated cultural content while maintaining cultural authenticity and environmental relevance.
• Collaborative AI Stewardship: The ability to work with others and AI systems to create collective environmental solutions and cultural preservation initiatives.
• Prompt-Mediated Environmental Agency: The capacity to use prompt literacy as a tool for environmental communication, advocacy, and action.

5. Discussion

This study demonstrates that the “Writing Is Coding for Sustainable Futures” paradigm can be effectively implemented in educational contexts to promote environmental consciousness, cultural preservation, and responsible AI use. Our findings contribute significantly to both theoretical understanding and practical implementation of AI-enhanced sustainability education through systematic engagement with existing theoretical frameworks and extension of current knowledge.

5.1. Theoretical Contributions and Framework Integration

This study makes three significant theoretical contributions to the intersection of AI-enhanced education and sustainability learning. First, it extends multimodal composition theory by demonstrating how the “Writing Is Coding” paradigm transforms traditional notions of authorship and creativity in educational contexts. Unlike previous studies that focused on technical aspects of AI integration, our research reveals how students develop environmental consciousness through the creative process of prompt engineering.

Second, the study contributes to sustainability education theory by identifying prompt literacy as a novel pedagogical approach for environmental learning. The five themes that emerged from our analysis—algorithmic design, emotional scaffolding, aesthetic imperfection, collaborative dynamics, and pedagogical value—constitute a comprehensive framework for understanding how AI-mediated creativity can serve sustainability education goals [15].

Third, our research advances grounded theory applications in educational technology by demonstrating how systematic coding procedures can reveal the complex relationships between technological literacy, creative expression, and environmental consciousness. The theoretical model developed through this study provides a foundation for future research on AI-enhanced sustainability education.

The findings of this study hold important implications for both AI education theory and environmental education theory. Our results support and extend the importance of AI literacy, prompt engineering, and critical thinking in modern education as proposed by Walter [3]. Specifically, our research demonstrates that these three elements have special synergistic effects in the context of environmental education, creating what we term “computational environmental literacy.”

Regarding the theoretical framework of AI literacy, the AI literacy concept proposed by Long and Magerko [5] provides an important theoretical foundation for understanding how individuals interact with AI systems. Their research demonstrates that AI literacy encompasses not only technical understanding but also cognition of AI’s social impacts and the ability to critically evaluate AI outputs. This framework is particularly relevant for environmental education, where students must navigate complex relationships between technology use and environmental impact.

From the perspective of AI literacy theory, our findings are highly consistent with the theoretical framework of Long and Magerko [5]. Students in our study not only mastered AI tool usage skills during the learning process but, more importantly, developed the ability to critically evaluate AI outputs in environmental contexts. This capability is particularly crucial in environmental education, as environmental issues often involve complex scientific data, uncertainty, and value judgments that require students to possess sophisticated information evaluation skills [16].

The exploratory review by Ng et al. [6] further enriches the conceptualization of AI literacy through their proposed four-aspect framework: knowing and understanding, using, evaluating, and ethical considerations. This framework provides specific guidance for cultivating AI literacy in educational contexts. The ethical considerations aspect is particularly significant for environmental education, as students must grapple with the environmental costs of AI technology while using it to promote environmental awareness.

The four aspects of AI literacy proposed by Ng et al. [6] were all manifested in our study, but with important extensions specific to environmental education contexts:

• Knowing and Understanding: Students developed understanding not only of AI capabilities but also of the environmental costs of AI systems, creating a more nuanced appreciation of technology-environment relationships.
• Using: Students learned to use AI tools not just for creative expression but specifically for environmental communication and advocacy, developing what we term “environmental prompt literacy.”
• Evaluating: Students developed sophisticated criteria for evaluating AI outputs that included not only technical quality but also environmental accuracy, cultural authenticity, and potential for promoting sustainable behavior.
• Ethical Considerations: Students naturally engaged with ethical questions about AI use in environmental contexts, including energy consumption, cultural appropriation, and the responsibility of creators for the environmental impact of their digital practices.

Our research extends Ng et al.’s framework [6] by demonstrating how AI literacy development can be specifically oriented toward environmental consciousness and action, creating a new theoretical connection between digital literacy and environmental education.

5.2. Theoretical Foundation of Prompt Engineering as Environmental Pedagogical Strategy

Federiakin et al. [4] define prompt engineering as a 21st-century skill encompassing four core components: comprehension of basic prompt structure, prompt literacy, prompting methods, and critical online reasoning. Our research demonstrates the special value of this skill framework in environmental education contexts and extends it in important ways.

Our findings reveal that prompt engineering in environmental education contexts involves additional competencies beyond those identified by Federiakin et al. [4]:

Environmental Scientific Literacy Integration: Students must learn to incorporate accurate environmental science data into their prompts while maintaining emotional resonance and cultural sensitivity. This requires a sophisticated understanding of how to translate complex scientific concepts into language that can generate compelling visual narratives.

Cultural–Environmental Sensitivity: When creating prompts for cultural heritage preservation related to environmental themes, students must navigate complex relationships between traditional ecological knowledge and contemporary environmental science, requiring cultural competency that extends beyond technical prompt engineering skills.

Ethical Environmental Reasoning: Students must consider the environmental impact of their AI tool usage while using these tools to promote environmental awareness, creating a productive tension that promotes deeper critical thinking about technology-environment relationships.

This structured thinking ability is highly compatible with the systematic thinking emphasized in environmental education. Students need to consider multiple aspects of environmental problems, their interrelationships, and potential solutions, which align precisely with the thinking process required for effective prompt engineering. Our research suggests that prompt engineering can serve as a powerful pedagogical strategy for developing environmental systems thinking.

5.3. Extensions to Existing Theoretical Frameworks

Our research proposes important extensions to existing AI education theoretical frameworks identified in the literature. Zawacki-Richter et al. [20] identified the problem of limited educator participation in AI education, noting that despite rapid technological advancement, educational practitioners remain underrepresented in AI education research and implementation.

Our research provides a specific solution to this challenge: integrating AI technology through the specific domain of environmental education can lower technical barriers for teachers while increasing student engagement. Environmental education provides a meaningful context that helps teachers understand the value of AI integration beyond mere technological novelty. Teachers in our study reported that the environmental focus helped them see AI tools as means to important educational ends rather than as technical challenges to be overcome.

Chiu et al. [21] emphasize that successful AI integration in education requires consideration of societal structural conditions and represents a fundamental paradigm shift rather than mere technological adoption. Our research supports this perspective and provides specific evidence for how this paradigm shift can be implemented in practice.

Our findings demonstrate that the paradigm shift involves reconceptualizing writing from linguistic expression to computational environmental action. This shift requires changes not only in how students understand writing and technology but also in how they understand their role as environmental citizens. Students in our study developed a sense of agency about environmental issues through their AI-mediated creative work, seeing themselves as capable of creating content that could influence community environmental awareness and action.

5.4. Novel Theoretical Contribution: Computational Environmental Literacy Framework

Building on our empirical findings and engagement with existing theoretical frameworks, we propose a new theoretical framework: Computational Environmental Literacy. This framework encompasses five interconnected competencies:

• Algorithmic Environmental Thinking: The ability to decompose environmental problems into components that can be addressed through AI-mediated creative processes, understanding how prompt structure influences environmental communication effectiveness.
• Emotional-Technical Integration: The capacity to combine emotional intelligence with technical AI skills to create compelling environmental narratives that promote both understanding and action.
• Critical AI-Cultural Literacy: The skill to evaluate and improve AI-generated cultural content while maintaining cultural authenticity and environmental relevance, particularly important for environmental education that incorporates traditional ecological knowledge.
• Collaborative AI Stewardship: The ability to work with others and AI systems to create collective environmental solutions and cultural preservation initiatives, including negotiation of ethical AI use standards.
• Prompt-Mediated Environmental Agency: The capacity to use prompt literacy as a tool for environmental communication, advocacy, and action, understanding the potential and limitations of AI-generated content for promoting environmental behavior change.

This framework provides a new theoretical lens for understanding how digital literacy and environmental education can be integrated in the age of AI, offering guidance for curriculum development, teacher preparation, and educational research.

5.5. Implications for Educational Practice and Policy

Our theoretical contributions have significant implications for educational practice and policy. The Computational Environmental Literacy framework suggests that environmental education in the digital age requires new pedagogical approaches that integrate technical skills with environmental consciousness and cultural sensitivity.

For curriculum development, our research suggests that AI literacy should not be taught as a separate technical subject but should be integrated with meaningful content areas like environmental education. This integration approach addresses the challenge identified by Zawacki-Richter et al. [20] of limited educator participation by providing teachers with clear connections between AI tools and their existing pedagogical expertise.

For teacher preparation, our research indicates that effective AI-enhanced environmental education requires teachers to develop competencies in three areas: AI literacy (understanding capabilities and limitations of AI tools), environmental content knowledge (solid grounding in environmental science and local environmental issues), and cultural sensitivity (particularly important when working with traditional ecological knowledge and cultural heritage materials).

For educational policy, our research suggests that successful AI integration requires institutional support for teacher professional development, community partnerships with environmental and cultural organizations, and assessment approaches that capture both technical skills and environmental consciousness development.

5.6. Addressing Environmental Paradoxes and Ethical Considerations

One of the most significant challenges identified in our study is the environmental paradox of using energy-intensive AI systems to promote environmental awareness. This paradox created important learning opportunities as students developed “carbon-conscious creativity” practices.

Students learned to optimize prompts to reduce the number of AI generations required, focus on high-impact environmental content that could justify the carbon cost of creation, and combine AI-generated content with low-carbon creative practices. This process of grappling with the environmental costs of their digital practices promoted deeper critical thinking about technology-environment relationships than would have been possible through traditional environmental education approaches.

The ethical considerations that emerged in our study extend beyond individual decision-making to collective responsibility for environmental impact. Students developed group norms for responsible AI use and learned to negotiate different perspectives on the appropriate balance between AI tool usage and environmental responsibility. This collaborative ethical reasoning represents an important educational outcome that prepares students for the complex environmental decision-making they will face as adults.

5.7. Limitations and Future Research Directions

While our research makes significant theoretical and practical contributions, several limitations should be acknowledged. Our study focused on high-achieving students in technology-integrated schools, which may limit the generalizability of findings to other educational contexts. Future research should examine AI-enhanced environmental education in diverse educational settings and with varied student populations.

The six-week study period provided insights into initial student engagement with AI tools but may not capture long-term impacts on environmental consciousness or behavior change. Longitudinal studies are needed to assess the sustained effects of AI-enhanced environmental education and to understand how computational environmental literacy develops over extended periods.

Our study was conducted in Taiwan with students from specific cultural backgrounds. Cross-cultural research is needed to understand how cultural factors influence student engagement with AI tools for environmental storytelling and cultural preservation, and to develop culturally responsive approaches to computational environmental literacy.

Future research directions include: (1) longitudinal studies of environmental behavior change following AI-enhanced education programs, (2) comparative studies of different AI tools and their effectiveness for environmental education, (3) investigation of teacher perspectives and experiences with AI-enhanced environmental curricula, (4) development and validation of assessment instruments for computational environmental literacy, and (5) exploration of how computational environmental literacy can be integrated with other sustainability education approaches.

The theoretical framework of Computational Environmental Literacy provides a foundation for this future research and offers a new lens for understanding the intersection of digital literacy, environmental education, and cultural preservation in the age of artificial intelligence.

6. Conclusions

This study demonstrates that the “Writing Is Coding for Sustainable Futures” paradigm can be effectively implemented in educational contexts to promote environmental consciousness, cultural preservation, and responsible AI use through systematic pedagogical approaches. The comprehensive grounded theory analysis with quantitative validation provides robust evidence for five core themes that illuminate how students develop sustainable prompt literacy while navigating the complex intersection of creativity, technology, and environmental responsibility.

The findings contribute significantly to both theoretical understanding and practical implementation of AI-enhanced sustainability education. The systematic methodology with independent verification and quantitative validation establishes new standards for qualitative research rigor while maintaining the depth and nuance necessary for understanding complex educational phenomena. The study’s comprehensive documentation and reproducibility measures ensure that these findings can be verified, replicated, and extended by future researchers.

The practical implications for educators are substantial, providing concrete evidence for effective pedagogical strategies that integrate AI tools with environmental education while addressing ethical and cultural considerations. The quantitative evidence demonstrates a measurable impact on student learning outcomes, environmental consciousness development, and sustained engagement in environmental advocacy activities.

This research demonstrates that the “Writing Is Coding” paradigm represents more than a metaphorical connection between creative writing and computational thinking—it constitutes a fundamental shift in how students engage with environmental challenges through AI-mediated creativity. The five themes identified through grounded theory analysis provide a comprehensive framework for understanding how prompt literacy can serve as a vehicle for sustainability education.

The implications extend beyond educational practice to broader questions about the role of AI in fostering environmental consciousness and cultural preservation. As generative AI technologies continue to evolve, the ability to craft effective prompts for sustainability-oriented content creation will become increasingly important. This study provides both a theoretical foundation and practical guidance for educators seeking to integrate AI tools with environmental education curricula.

Future research should explore the long-term impacts of prompt literacy on environmental behavior and investigate how the “Writing Is Coding” paradigm might be adapted for different educational contexts and age groups. Additionally, research examining the environmental costs of AI-mediated creativity alongside its educational benefits would provide important insights for the sustainable implementation of these technologies.