Future-proof learning: redesigning ADDIE for Generation Alpha with explainable AI and human-computer integration

The COVID-19 pandemic accelerated the integration of educational technologies in higher education institutions, prompting a rapid shift toward digital learning environments (Schneider and Council, 2021). As e-learning becomes mainstream and expands access to open and distance learning (ODL), integrating advanced frameworks such as human-computer integration (HInt) and explainable artificial intelligence (XAI) has become essential for fostering engagement, personalisation and transparency in instruction. These frameworks are particularly relevant for digitally fluent learners who expect interactive and adaptive experiences. While human-computer interaction (HCI) focuses on usability and interface efficiency, HInt emphasises co-agency, where humans and systems continuously adapt and learn from each other (Cornelio et al., 2022). This conceptual shift underpins FLEX-ADDIE's learner-AI collaboration.

Instructional design remains critical in shaping these experiences, with the ADDIE model (analyse, design, develop, implement and evaluate) continuing as a foundational framework (Patel et al., 2018). Originally developed for military training, ADDIE has informed various learner-centred models but now requires re-examination to stay relevant in AI-enhanced education. To support learner autonomy in ODL contexts, HInt and XAI provide the necessary collaborative architecture and transparent algorithmic feedback. HInt enables deeper human–technology collaboration, while XAI strengthens trust through interpretable decision-making (Gohel et al., 2021). This study is therefore situated within ODL contexts, where large-scale participation, asynchronous interaction and learner autonomy make transparency and adaptive feedback critical design considerations in AI-mediated environments.

These design considerations are particularly urgent for Generation Alpha. Generation Alpha, born between 2010 and 2025, is growing up immersed in intelligent systems and automation (Glumova et al., 2021; Ziatdinov and Cilliers, 2021). Their digital fluency demands redesign of learning systems to sustain engagement and equity.

This paper introduces FLEX-ADDIE, a conceptual model that re-envisions ADDIE through HInt and XAI integration. Rather than offering empirical validation, it aims to establish theoretical foundations for transparent, adaptive and ethically governed learning systems aligned with Generation Alpha's emerging needs and the principles of ODL.

The literature review adopted a content-driven, semi-systematic and thematic synthesis approach (Snyder, 2019), supported by structured searches using Google Scholar as the primary database. A total of 206 articles were initially identified. Following a preliminary screening of titles and abstracts, 122 articles were excluded. A full-text review of the remaining 84 articles resulted in 16 being retained for the final synthesis. This relevance screening evaluated thematic alignment with instructional design, applicability to training and learning contexts and publication timeframe. Google Scholar was chosen as a primary database to ensure diverse and inclusive sourcing. This approach reflects an effort to capture open-access scholarship across varied educational research contexts, which might otherwise be excluded when relying strictly on subscription-based repositories such as Scopus or Web of Science. Search strings combined the keywords “ADDIE,” “instructional design,” “AI,” “XAI,” “HInt” and “Generation Alpha,” limited to English-language publications (2019–2024). Relevant studies were screened for their contribution to instructional design, AI integration, or learner adaptability in ODL. The synthesis consolidated conceptual insights and identified research gaps informing the FLEX-ADDIE model. Such inclusive sourcing aligns with the ODL ethos of promoting accessible and equitable knowledge dissemination (Kukulska-Hulme et al., 2024).

Following manual screening, studies not meeting inclusion criteria were excluded. Inclusion criteria included:

1. ADDIE-based studies: Focused on ADDIE's effectiveness in education or training, peer-reviewed and published from 2019 onwards, corresponding to the pre- and post-COVID-19 transition period.

2. Generation Alpha studies: Focused on traits and learning characteristics of Generation Alpha in technological or educational contexts, drawn from reputable peer-reviewed or published sources.

Article selection combined systematic keyword searching with authorial judgement of conceptual relevance, consistent with the semi-systematic review tradition, where interpretive decisions are made transparently and methodologically reasoned (Snyder, 2019). This study is positioned as a conceptual design research contribution informed by a semi-systematic literature review. These studies were not analysed as a fixed dataset but were synthesised alongside supporting literature to inform conceptual development.

To enhance conceptual grounding, brief exploratory reflections were gathered from 20 digitally fluent undergraduates, selected based on their familiarity with AI-mediated learning environments, through open-ended questions. An initial pool of 24 participants was considered, of which four responses were excluded due to lack of relevance. These reflections were not intended as empirical validation but served as illustrative inputs to examine how the proposed model aligns with current learner expectations in AI-mediated environments (formal qualitative coding procedures were not applied). This approach provides reflective input from learners familiar with AI-mediated learning, aligning with the study's conceptual scope.

Exploratory reflections enhance conceptual and practical understanding in literature-based research by encouraging critical self-evaluation and contextual insight (Slimani-Rolls and Kiely, 2018; Hanks, 2019). The growing integration of generative AI in education further highlights the need for reflective evaluation and adaptation of AI-mediated learning resources (Cooper, 2023). Collectively, these studies support the inclusion of exploratory reflections as a means of refining and contextualising the FLEX-ADDIE model.

The FLEX-ADDIE model is derived from theoretical synthesis rather than empirical testing. Learner reflections are included to examine conceptual resonance with current digital learner expectations and do not constitute formal qualitative validation. Empirical evaluation of the model remains a direction for future research.

The ADDIE model emerged from military training design and evolved into a widely adopted systematic instructional framework, comprising analysis, design, development, implementation and evaluation (Molenda, 2022).

The ADDIE model comprises five interrelated phases – analysis, design, development, implementation and evaluation – that structure systematic instructional planning, as seen in Figure 1. The analysis phase identifies learner needs and performance gaps (Hamzah et al., 2022); design defines instructional strategies and assessment; development produces learning materials; IXAI implementation delivers instruction and evaluation assesses effectiveness for iterative improvement. While initially framed as linear, contemporary applications emphasise flexibility through iterative refinement and integration with agile techniques. ADDIE therefore remains a structured yet adaptable framework for instructional development.

Contemporary extensions of ADDIE, such as Rapid Prototyping, Agile ID and Merrill's Principles of Instruction, strengthen its iterative and learner-centred dimensions (Badali et al., 2022; Abuhassna and Alnawajha, 2023; Shakeel et al., 2023). However, their emphasis on speed, flexibility and rapid deployment often foregrounds workflow efficiency over interpretability and transparent design reasoning. As AI-mediated systems increasingly influence instructional decision-making, this relative underemphasis on explainability becomes a structural limitation.

ADDIE remains a robust and widely applied framework for iterative development and evaluation (Spatioti et al., 2022). Empirical applications across domains reaffirm its procedural strength and adaptability. However, most implementations focus on delivery efficiency and engagement rather than addressing AI-driven feedback logic, algorithmic bias, or structured human-AI co-agency (Almelhi, 2021; Jonnalagadda et al., 2022; Zardosht et al., 2023). Multimedia and game-based applications further demonstrate its use in e-learning and educational gaming contexts, contributing to retention and engagement. Recently, mobile applications and augmented reality (AR) have also been proven to support higher-order thinking and digital competence (Abuhassna et al., 2024; Adeoye et al., 2024).

ADDIE has also demonstrated sustained relevance across vocational, blended and technology-enhanced learning environments. Applications in Technical and Vocational Education and Training (TVET), web-centric and problem-based learning contexts and blended Islamic studies highlight its adaptability to practice-oriented and hybrid pedagogies (Stapa and Mohammad, 2019; Hamzah et al., 2022; Shakeel et al., 2023).

Empirical applications demonstrate ADDIE's adaptability across diverse contexts, including Learning Management System (LMS) -based instruction, vocational education, multimedia development, clinical training and blended learning. These applications highlight its systematic structure and capacity for iterative refinement. However, most implementations focus on delivery efficiency, engagement and content alignment rather than interrogating algorithmic mediation, interpretability, or structured human-AI co-agency. As AI increasingly shapes feedback systems and learning analytics, this omission suggests a structural gap within conventional ADDIE applications, one that FLEX-ADDIE seeks to address through explicit cross-phase integration of HInt and XAI.

Generation Alpha (born approximately between 2010 and 2025) is frequently described as the most digitally immersed cohort to date, shaped by early and sustained exposure to networked technologies (Ziatdinov and Cilliers, 2021). Continuous interaction with digital devices has reshaped their learning environments and media engagement patterns. Despite these advantages, concerns have been raised regarding shorter attention spans and reduced face-to-face social interaction in some contexts. While generational descriptors provide heuristic value, they risk overgeneralising heterogeneous learner populations across socio-economic and cultural contexts.

Raised in technology-rich environments, this cohort engages comfortably with mobile applications and online learning platforms, particularly following the COVID-19 shift to remote education (Glumova et al., 2021). Interactive digital tools support engagement and problem-solving (Jha, 2020), while AI-driven and immersive technologies strengthen personalised learning experiences (Fan and Zhong, 2022). Yet technological familiarity does not automatically imply critical understanding of algorithmic systems or digital governance structures.

Interactive, multimedia and gamified approaches have been shown to enhance engagement and cognitive development among digitally immersed learners (Tootell et al., 2014; Kovácsné Pusztai, 2021). Visual and dynamic media enhance comprehension (Gupta et al., 2022), while augmented and virtual reality enable immersive simulation-based learning, particularly in professional training contexts (Danry et al., 2021). This synthesis indicates that future instructional frameworks should move beyond static content delivery toward structured interactivity and adaptive feedback mechanisms embedded within the design architecture.

A strong orientation toward collaboration has been observed in both physical and digital learning environments. Online platforms and game-based systems support communication and may reduce dropout rates (Tootell et al., 2014; Kovácsné Pusztai, 2021). While learner preferences vary, flexible instructional approaches remain necessary to accommodate diverse levels of independence and interaction.

AI-enabled systems increasingly mediate student-instructor engagement and expand access to content (Rathore, 2023). Tools such as ChatGPT may provide personalised feedback; however, algorithmic transparency is critical to sustaining trust and responsible use (Dwivedi et al., 2023). Despite technological mediation, teachers remain central in guiding motivation and critical engagement (Serinikli, 2019). Experiential and project-based strategies further support teamwork, problem-solving and higher-order thinking (Miller, 2023). These patterns reinforce the need for instructional frameworks that deliberately structure human-AI collaboration rather than treating technology as a replacement for pedagogical oversight.

Continuous exposure to digital media provides Generation Alpha with rapid access to information and strong online search proficiency but may reduce tolerance for extended text-based instruction formats (Kovácsné Pusztai, 2021). Modular, scaffolded learning structures have therefore been recommended to sustain attention. Instructional models such as ADDIE have been adapted to support personalised, technology-aligned curricula (Nagpal and Kumar, 2020; Almelhi, 2021), while educators are encouraged to integrate disciplinary expertise with digital pedagogy to maintain engagement in evolving learning environments (Miller, 2023).

Informal learning through tutorials, videos, podcasts and social media increasingly complements formal education (Ziatdinov and Cilliers, 2021). Integrating such elements into structured curricula may enhance digital literacy, credibility and psychological engagement (Glumova et al., 2021). These shifts highlight the need for instructional frameworks that intentionally coordinate formal and informal learning pathways rather than treating them as parallel domains. Although multimedia and gamified approaches enhance engagement, they rarely address how AI-mediated feedback mechanisms shape learner autonomy or data governance.

Exposure to digital environments and online entrepreneurship has also been associated with adaptability, innovation and problem-solving skills relevant to future employment (Serinikli, 2019; Ziatdinov and Cilliers, 2021). Supportive parental and educational guidance further strengthens these capacities and promotes balanced development (Jha, 2020). However, high digital fluency does not eliminate socio-emotional risks. Excessive screen exposure has been linked to emotional and social challenges (Höfrová et al., 2024). Instructional design must therefore integrate emotional intelligence, ethical awareness and sustainable practices alongside technological innovation (Kantarcıoğlu, 2025).

The collective insights from these studies outline the major traits and educational tendencies of Generation Alpha, acknowledging the influence of cultural, institutional and individual factors. A summary of the reviewed studies is presented in Table 1.

ADDIE provides a structured framework adaptable to digitally immersed learners. However, this adaptability largely operates at the level of content delivery and engagement, rather than addressing deeper challenges related to algorithmic mediation, learner autonomy and transparency in AI-mediated environments. In the analysis phase, identifying media habits, attention patterns and digital fluency enables alignment between learner characteristics and instructional goals. Design and development phases can incorporate multimedia, mobile-first structures, immersive technologies and adaptive mechanisms to address interactive preferences while mitigating cognitive overload. Implementation may increasingly rely on AI-supported feedback systems and data-informed adjustment, while evaluation leverages analytics and transparent reporting to sustain engagement and trust.

However, although ADDIE can accommodate such adaptations, it does not inherently embed algorithmic transparency, human-AI co-agency, or structured governance mechanisms. These omissions highlight the need for a model that systematically integrates HInt and XAI across all five phases, an objective addressed by FLEX-ADDIE.

HCI has traditionally emphasised usability and interface efficiency, while recent developments highlight a shift toward HInt, in which humans and systems act as co-agents in shared, adaptive environments (Rodrigues Barbosa et al., 2024). As such, HCI alone does not fully address the evolving demands of AI-mediated learning environments. This shift extends beyond simple interaction to embodied and continuous collaboration, as noted by Rodrigues Barbosa et al. (2024) and Cornelio et al. (2022), whose studies emphasise the growing importance of integration-based design in immersive and AI-enhanced systems. In this study, HInt is used to represent these co-adaptive dynamics that characterise Generation Alpha's lived experience with intelligent technologies.

Within instructional design, this shift becomes critical as AI increasingly mediates feedback, assessment and content adaptation, necessitating explainable systems to support trust, accountability and informed decision-making. Embedding HInt and XAI into the ADDIE framework is proposed to strengthen experiential learning by enabling real-time feedback, adaptive pathways and transparent decision-making. XAI is intended to foster trust through interpretable feedback, while HInt promotes engagement through embodied, multimodal learning such as AR/VR-based simulations and responsive platforms. Recent advances in XAI emphasise context-sensitive explanations suitable for microlearning, an approach particularly compatible with Generation Alpha's digital fluency.

While alternative design frameworks offer valuable orientations, FLEX-ADDIE differs in scope and integration logic. Models such as successive approximation model (SAM) emphasise iterative prototyping but do not embed AI transparency or human-computer co-agency as structural principles. Universal dlearning (UDL) foregrounds accessibility and inclusivity but does not systematically integrate AI-mediated feedback across instructional phases. Agile instructional design and design-based research (DBR) prioritise responsiveness and research–practice integration, yet neither provides a phase-based architecture for embedding explainable AI and human-centred analytics. Similarly, agile-blended learning highlights flexibility, learner autonomy and technology-mediated instruction but remains focused on delivery rather than explainability or human-AI co-agency (Tang et al., 2025).

The proposed FLEX-ADDIE model integrates these principles by aligning ADDIE's systematic structure with HInt-XAI elements across all five phases: analysis, design, development, implementation and evaluation, as shown in Figure 2. It translates cognitive, social and technological traits of Generation Alpha into design logic, such as bite-sized and multimodal development to manage cognitive load, explainable AI to support co-agency in implementation and transparent analytics in evaluation. As illustrated in Figure 2, each phase is redefined to support adaptive co-agency, explainable feedback and transparent learning pathways aligned with digitally fluent learners. FLEX-ADDIE thus operates as a structured instructional architecture that embeds these mechanisms across the design cycle in an adaptive and interpretable design structure.

Recent efforts to modernise instructional design frameworks reflect the need for adaptive, data-driven learning environments. However, few models explicitly embed XAI and HInt across all design phases. The proposed FLEX-ADDIE model retains ADDIE's systematic rigour while introducing transparency, adaptability and ethical governance mechanisms aligned with contemporary ODL contexts.

The SAM and Agile instructional design prioritise iterative prototyping and learner-centred responsiveness. Empirical work has shown that SAM improves engagement and conceptual gains among distance learners compared with ADDIE, highlighting its flexibility (Ali et al., 2021). Nevertheless, SAM lacks the structured evaluation and explainability mechanisms needed for AI-enhanced learning ecosystems.

Similarly, the UDL framework advances inclusivity and learner variability. Studies of UDL in inclusive classrooms and MOOCs demonstrate its value but also reveal interpretational inconsistencies and a lack of adaptive intelligence or transparent algorithmic mediation (Almumen, 2020; Iniesto et al., 2023).

Contemporary DBR approaches emphasise iterative innovation and contextual responsiveness. While DBR contributes to scalable educational improvement (Tinoca et al., 2022), it remains largely research-driven and difficult to operationalise at institutional scale.

In contrast, FLEX-ADDIE integrates HInt and XAI across the five ADDIE phases: analysis, design, development, implementation and evaluation, to enable learner-system co-agency. It maintains methodological rigour while embedding explainable data flows, adaptive feedback and human-centred decision support aligned with ethical AI guidelines (Dwivedi et al., 2023; Rodrigues Barbosa et al., 2024; Filiz et al., 2025). This reconceptualisation positions FLEX-ADDIE as a systematic, interpretable and ethically aligned conceptual framework for trustworthy, inclusive digital transformation in ODL.

To reinforce the conceptual grounding of the FLEX-ADDIE model, exploratory reflections were collected from 20 Gen Z undergraduates through open-ended questions. The feedback highlighted strong preferences for multimodal learning, personalised digital content and adaptive platforms, aligning closely with the model's HInt and XAI components. Students highlighted the value of multimedia tools such as videos and podcasts, real-time feedback and intuitive interfaces for sustaining engagement and comprehension. They also suggested improvements for e-learning systems, including embedded feedback mechanisms and better integration with external resources.

These reflections, while limited in scope, offer illustrative alignment signals regarding learner priorities in AI-mediated learning environments. They informed refinement of FLEX-ADDIE by highlighting expectations related to adaptability, system trust and human-computer co-agency. The patterns presented in Table 2 are descriptive and serve to examine conceptual resonance.

Students described orchestrating multiple tools and modalities; noted LMS/e-learning friction around access, stability and navigation; reported frequent use of AI and support tools as study scaffolds; emphasised the need for timely, actionable feedback and requested personalisation, better search and mobile/offline modes.

These insights informed FLEX-ADDIE's HInt-XAI integration: co-agency features for learner control, platform design that reduces friction and explainable assistance and dashboards that make recommendations, progress and evaluations transparent. The reflections were used not as empirical validation, but as illustrative exploratory feedback informing the conceptual refinement of the FLEX-ADDIE model.

These student insights were subsequently mapped across the five phases of the FLEX-ADDIE model to clarify how HInt and XAI mechanisms operate within each instructional stage. Each phase specifies inputs, processes, outputs and roles that guide adaptive, transparent and learner-centred design. A concise operational summary is provided in Table 3.

Collectively, these operational elements illustrate how HInt and XAI mechanisms are embedded within each ADDIE phase, supporting the adaptive and transparent learning processes visualised in Figure 2.

FLEX-ADDIE is a conceptual instructional design framework that addresses limitations in traditional models by embedding XAI and HInt as structural principles. Rather than centring automation alone, it emphasises transparency, interpretability and human-AI co-agency to support adaptive and ethically grounded learning environments.

A key feature is the conceptual integration of SHapley Additive exPlanations and Local Interpretable Model-agnostic Explanations to enhance interpretability in AI-assisted instructional processes. These approaches can make algorithmic decisions in content sequencing, feedback and analytics more accessible to educators and learners. Such transparency may strengthen trust, reduce bias and support responsible AI use in education (Kalusivalingam et al., 2021; Salih et al., 2024).

The model is designed to be adaptable across contexts. In K-12, HInt applications such as gamified or immersive tools may support engagement. In higher education, explainable dashboards may guide personalised pathways and performance tracking. In MOOCs and rural ODL, lightweight analytics may promote broader participation despite infrastructural constraints (Sharma, 2024; Filiz et al., 2025).

FLEX-ADDIE aligns HInt and XAI mechanisms across the ADDIE phases: interpretive dashboards in analysis, explainable structuring in design, traceable adaptive tools in development, transparent feedback in implementation and analytics-informed reflection in evaluation. Future research should examine the model through measures of learner trust, interpretability comprehension, engagement indicators and educator confidence in AI-supported instruction.

The model provides preliminary implementation guidance. Core components include phase-specific explainable mechanisms, explicit co-agency structures and transparency protocols that strengthen trust and interpretability. Optional enhancements such as adaptive dashboards or predictive analytics may be adopted according to institutional capacity. In resource-constrained contexts, initial focus may prioritise the analysis and evaluation phases. Minimum requirements include AI-capable infrastructure, faculty AI literacy and governance frameworks addressing data ethics and consent.

FLEX-ADDIE thus offers a structured foundation for exploring scalable and ethically governed ODL implementation, guided by privacy-by-design, informed consent, bias auditing and algorithmic transparency.

FLEX-ADDIE is proposed as a conceptual extension of ADDIE for AI-mediated ODL environments, systematically integrating HInt and XAI across all five design phases. Its contribution is not in the mere introduction of AI tools but in integrating structured transparency, interpretability and human-AI collaboration within the framework of instructional design. The model is particularly relevant for ODL environments, where transparency, scalability and adaptive feedback are essential, further highlighting the need for flexible and scalable instructional models (Husain, 2025).

As a conceptual framework, FLEX-ADDIE requires empirical validation. Future research should prioritise foundational studies examining learner trust, interpretability comprehension and measurable indicators of human-AI co-agency. Longitudinal and cross-cultural investigations are necessary to assess scalability, contextual adaptability and ethical implications across diverse higher education settings.

This study did not involve human subjects requiring institutional ethics clearance. Student feedback was obtained voluntarily with informed consent, and no personal or identifiable data were collected. The author declares no competing interests and receives no external funding.