Gone

TPM Vol. 32, No. S9, 2025 Open Access ISSN: 1972-6325 https://www.tpmap.org/ 1356 DATA-DRIVEN PEDAGOGICAL INNOVATION: ANALYSIS OF EMERGING STRATEGIES IN HYBRID EDUCATIONAL ENVIRONMENTS ISABEL CECILIA LLERENA RANGEL UNIVERSIDAD DE LA COSTA, BARRANQUILLA, COLOMBIA, EMAIL: [email protected] NELSON ANDRÉS MONTERO RAMÍREZ ESCUELA SUPERIOR DE ADMINISTRACIÓN PÚBLICA, BOGOTÁ, COLOMBIA, EMAIL: nels[email protected], ORCID: https://orcid.org/0009-0000-0064-7689 HÉCTOR ANDRÉS ZAMBRANO NOBOA INSTITUTO DE ADMISIÓN Y NIVELACIÓN, UNIVERSIDAD TÉCNICA DE MANABÍ, ECUADOR, ING. EN AUDITORÍA Y CONTABILIDAD, MAGÍSTER EN ADMINISTRACIÓN DE EMPRESAS, EMAIL: hect[email protected], ORCID: https://orcid.org/0000-0003-1251-7579 SEBASTIÁN CANO-ROJAS FACULTAD DE INGENIERÍA, UNIVERSIDAD DE MEDELLÍN, MEDELLÍN, ANTIOQUIA, COLOMBIA N.º 050026, EMAIL: [email protected], ORCID: https://orcid.org/0009-0005-1145-3229 Summary The expansion of hybrid educational environments after the pandemic has intensified the availability of learning data, which can guide more accurate pedagogical decisions. This article presents a theoretical-documentary review of recent literature (2021–2025) on data-driven pedagogical innovation in hybrid and blended contexts. Empirical studies and systematic reviews on learning analytics, educational data mining, adaptive platforms and artificial intelligence are analyzed, as well as works focused on the ethics of data use and data literacy among teachers. The results are organized into four axes of emerging strategies: (1) use of learning analytics for formative feedback and early warning, (2) personalization through adaptive platforms and AI systems, (3) hybrid instructional redesign supported by data and (4) ethical data governance and development of teacher data literacy. Evidence indicates that the pedagogical integration of learning analytics and AI improves monitoring capacity, personalization, and student engagement, as long as it is accompanied by robust ethical frameworks and specific teacher training. It is concluded that data-based pedagogical innovation in hybrid environments requires articulating, in a balanced way, technological infrastructure, student-centered pedagogical models, and institutional policies that guarantee the protection and responsible use of data. Keywords: pedagogical innovation; learning analytics; hybrid environments; educational artificial intelligence; data-driven decision-making; higher education. INTRODUCTION The global transition to hybrid and blended teaching modalities, initially accelerated by the COVID-19 pandemic, has led to profound and sustained transformations in higher education. In this new educational ecosystem, the interaction between face-to-face and digital scenarios requires flexible pedagogical models, supported by robust technological infrastructures and systematic formative assessment processes. Various recent reviews agree that hybrid environments are no longer limited to a short-term response, but are consolidated as permanent configurations that seek to integrate the best of both worlds: the social and contextual interaction of the face-to-face classroom, and the adaptivity, traceability, and accessibility of the digital space (Zhang et al., 2025; Zhong & Li, 2025). The growth in the use of learning management platforms (LMS), video conferencing systems, digital resources, and academic tracking applications has led to an abundant availability of data on learning trajectories, participation, study paces, social interactions, and academic performance. This expansion of the digital ecosystem has placed learning analytics and educational data mining as key tools for transforming data into useful pedagogical knowledge. Over the past five years, research has shown that these technologies can support the early identification of at-risk students, provide personalized feedback, and improve teacher decision-making TPM Vol. 32, No. S9, 2025 Open Access ISSN: 1972-6325 https://www.tpmap.org/ 1357 through visualization and interpretation of learning patterns (Harun Kılıç & İzmirli, 2024; Palancı & Y. M. R., 2024; Nuankaew et al., 2023). However, data-driven pedagogical innovation should not be understood solely as a technological issue. Recent studies reveal that the real impact occurs when data is integrated into teachers' reflective practice, through teacher inquiry cycles such as the Analytics Model for Teacher Inquiry (AMTI), which guides teachers in formulating questions, analyzing evidence, and deliberately adjusting instructional design (Saar et al., 2022). This is especially critical in hybrid environments, where decisions must articulate face-to-face experiences with digital ones and consider the differentiated needs of students. In parallel, the incorporation of artificial intelligence (AI) in educational platforms and analytics systems has significantly expanded the possibilities for personalizing learning. Recent reviews indicate that AI can act as an intelligent tutor, evaluator assistant, or adaptive recommendation engine, generating more accurate interventions based on student behavior patterns (Park & Doo, 2024; Sajja et al., 2025). However, algorithmic opacity and lack of data literacy on the part of teachers continue to be important barriers to responsible pedagogical implementation. As hybrid environments take hold, substantial ethical debates are also emerging. Research in recent years points to risks associated with digital surveillance, algorithmic discrimination, lack of transparency about data use, and the asymmetrical power relationship between institutions, teachers, and students (Wang et al., 2025; Zupanc & Araya, 2025). In addition, students from different regions report specific concerns about privacy, consent, and control over their academic data, which shows the importance of establishing clear institutional governance and communication frameworks (Zwitser et al., 2024). In this context of accelerated transformation, data-driven pedagogical innovation in hybrid environments is presented as a strategic field for the future of higher education. Recent literature shows that the integration of learning analytics, adaptive platforms, and AI can enhance personalization, active learning, self-regulation, and teacher effectiveness, provided that it is accompanied by ethical principles, clear institutional policies, and a strengthening of teachers' digital and data literacy (UNESCO, 2024; Pöysä-Tarhonen, 2025). Against this backdrop, this article seeks to analyze emerging strategies for data-driven pedagogical innovation in hybrid environments, in order to identify key trends, opportunities for pedagogical redesign, and associated ethical challenges. This review allows us to understand how institutions can move towards more flexible, personalized educational models that are aware of the responsibility involved in learning data management. THEORETICAL FRAMEWORK The theoretical framework is structured in four fundamental dimensions that underpin data-driven pedagogical innovation in hybrid environments: (1) evolution of hybrid environments and digital pedagogy, (2) learning analytics and educational data mining, (3) artificial intelligence and adaptive personalization, and (4) ethics, governance, and data literacy. 1. Hybrid educational environments and digital pedagogy Hybrid environments are characterized by integrating face-to-face and online activities strategically to optimize flexibility, participation, and pedagogical consistency. According to Zhang et al. (2025), hybrid education in higher education is today an "expanded ecosystem", where the student experience takes place fluidly between face-to-face and digital, mediated by technological, analytical, and synchronous and asynchronous interaction dynamics. The most recent research underscores that the pedagogical value of hybrid environments lies not only in their technological dimension, but also in the integrated instructional design, which combines active methodologies, formative assessment, and continuous feedback strategies (Mayani, 2021; Zhong & Li, 2025). Likewise, recent systematic reviews agree that hybrid learning generates rich digital traces, useful for learning analytics: records of interaction, participation, navigation, timing, evaluations, and collaborative patterns (Nuankaew et al., 2023). Table 1. Key features of modern hybrid environments (2021–2025) Dimension Description Recent Source Temporal and spatial flexibility Combines face-to-face and digital activities with the possibility of alternation Zhang et al. (2025) Technology integration Using LMS, video conferencing, and collaborative platforms Mayani (2021) High traceability Generating continuous engagement and performance data Nuankaew et al. (2023) Active pedagogy Student-centered model and meaningful participation Zhong & Li (2025) TPM Vol. 32, No. S9, 2025 Open Access ISSN: 1972-6325 https://www.tpmap.org/ 1358 2. Learning Analytics and Educational Data Mining Learning Analytics (LA) is defined as the process of measuring, collecting, and analyzing data about students and their contexts in order to understand and optimize learning. Over the past five years, the literature has shown significant growth in the use of LA to support real-time pedagogical decision-making (Harun Kılıç & İzmirli, 2024; Palancı & Y. M. R., 2024). On the other hand, Educational Data Mining (EDM) focuses on algorithmic techniques to discover complex patterns, segment student profiles, and predict academic risks (Nuankaew et al., 2023). Both perspectives converge in hybrid environments, where the wealth of data makes it possible to identify trends that are not easily visible to teachers. Models such as the Analytics Model for Teacher Inquiry (AMTI) propose integrating analytics into systematic cycles of pedagogical inquiry, allowing teachers to formulate hypotheses, analyze evidence, and redesign their practices (Saar et al., 2022). This model highlights that the real impact of analytics depends on contextualized interpretation, not just data visualization. Table 2. Emerging Uses of Learning Analytics in Hybrid Education Pedagogical use Example Authors Early warning Identification of attrition risk from inactivity patterns Palancı & Y. M. R. (2024) Formative feedback Visual dashboards showing progress and areas for improvement Harun Kılıç & Izmir (2024) Continuous evaluation Performance analysis in hybrid activities Nuankaew et al. (2023) Teacher decisionmaking AMTI Cycles for Instructional Redesign Saar et al. (2022) 3. Artificial intelligence and adaptive platforms in hybrid environments The incorporation of AI in education has allowed the development of adaptive platforms, intelligent tutors, predictive systems and automated feedback tools. In recent reviews, Park and Doo (2024) note that AI in blended learning serves three main functions: 1. Smart mentoring: personalized guidance and immediate feedback. 2. Automated assessment: correction and analysis of tasks, questionnaires and participation. 3. Adaptive personalization: dynamic adjustment of content and learning paths. Likewise, generative models (GenAI) have been incorporated into new systems capable of producing textual feedback, additional exercises, and personalized diagnoses based on performance patterns (Sajja et al., 2025). Adaptive platforms have shown moderate but consistent improvements in student academic performance, selfregulation, and persistence (Park & Doo, 2024). In hybrid environments, these systems allow the face-to-face session to focus on higher-order activities, while the platform adapts the pace and complexity of the digital material. Table 3. Recent applications of AI in hybrid education (2021–2025) Application Description Fountain Intelligent Tutor Systems Provide personalized explanations and examples Park & Doo (2024) Adaptive platforms They adjust content according to the student's progression Sajja et al. (2025) Generative Models (GenAI) Automated feedback and resource creation Sajja et al. (2025) Risk prediction ML Models Identify Vulnerable Students Nuankaew et al. (2023) 4. Ethics, Data Governance and Teacher Literacy The expansion of data use has led to a parallel increase in research on ethics, privacy, algorithmic justice, and transparency. Zupanc and Araya (2025) emphasize that learning analytics must be aligned with clear ethical principles, avoiding surveillance and discrimination practices, particularly when high-impact predictive models are employed. At the institutional level, the most recent recommendations suggest establishing policies for: • informed consent, • transparency on the use of data, • limitation of access and retention, TPM Vol. 32, No. S9, 2025 Open Access ISSN: 1972-6325 https://www.tpmap.org/ 1359 • explainability of algorithms, • student participation in decision-making (Wang et al., 2025; Zwitser et al., 2024). In parallel, teacher data literacy has become an essential component. This includes skills to interpret indicators, question algorithmic assumptions, identify biases, use data for feedback and redesign practices, and communicate results ethically (UNESCO, 2024). According to Saar et al. (2022), teachers with high data literacy are essential to transform information into effective pedagogical decisions. Table 4. Risks and ethical principles in the educational use of data Risk Description Associated ethical principle Fountain Oversurveillance Disproportionate data collection Data Minimization Wang et al. (2025) Algorithmic bias Unfair predictions based on inadequate patterns Justice and equity Zupanc & Araya (2025) System opacity Lack of understanding about how the algorithm operates Transparency and explainability Zwitser et al. (2024) Unauthorized Use Data used for non-pedagogical purposes Informed consent UNESCO (2024) METHODOLOGY This research adopts a theoretical-documentary review design of narrative-analytical scope, aimed at identifying, comparing and synthesizing emerging strategies of pedagogical innovation based on data in hybrid educational environments. This type of review is appropriate when the objective is to build a broad conceptual map from recent literature, without necessarily performing a quantitative meta-analysis (Harun Kılıç & İzmirli, 2024; Zhang et al., 2025). The methodology was structured in three phases: (1) systematic search and selection of literature, (2) thematic analysis and coding, and (3) interpretative synthesis oriented to the construction of emerging categories. 1. Search strategy and selection criteria The review was developed following methodological guidelines used in recent systematic reviews in the fields of learning analytics, educational AI, and hybrid teaching (Nuankaew et al., 2023; Park & Doo, 2024; Zupanc & Araya, 2025). Although this research does not constitute a strict systematic review, it does adopt transparent selection and exclusion criteria, as well as replicable search procedures. 1.1. Databases consulted We searched from January 2021 to December 2025 in: • Scopus • Web of Science • ERIC • ScienceDirect • SpringerLink • MDPI • Open repositories (Zenodo, ResearchGate, LAK Conference Proceedings) These bases were selected due to their high concentration of recent literature on hybrid environments, educational AI, and learning analytics (Palancı & Y. M. R., 2024; Saar et al., 2022). 1.2. Keywords used Spanish and English combinations were used: • learning analytics, educational data mining, hybrid learning, blended learning, data-informed teaching, adaptive learning systems, AI in education, "hybrid environments", "data-driven pedagogical innovation", "educational AI". The use of multilingual keywords allowed to broaden the spectrum of studies and reduce geographical biases (UNESCO, 2024). 1.3. Inclusion and exclusion criteria The criteria are based on the methodological quality standards used in recent reviews (Zhang et al., 2025; Zupanc & Araya, 2025). TPM Vol. 32, No. S9, 2025 Open Access ISSN: 1972-6325 https://www.tpmap.org/ 1360 Table 1. Inclusion and exclusion criteria used in the review Guy Criteria Foundation Inclusion Peer-reviewed articles, 2021–2025; empirical, conceptual studies, or reviews related to analytics, AI, or hybrid design; explicit focus on pedagogical decisions based on data. Consistent with reviews by Nuankaew et al. (2023) and Park & Doo (2024). Exclusion Purely technical studies without pedagogical implication; work prior to 2021; publications without peer review; documents not oriented to formal education. It coincides with recent criteria in educational ethics and analytics (Zupanc & Araya, 2025). 2. Information analysis procedure The literature analysis was developed in four successive stages, following coding methodologies used in contemporary educational research (Saar et al., 2022; Harun Kılıç & İzmirli, 2024). 2.1. Phase 1: Initial screening We filtered studies by title, abstract and keywords to ensure they met the inclusion criteria. Duplicate articles or articles with an exclusively technological focus were discarded. 2.2. Phase 2: Deep reading and information extraction Key variables were extracted from each document using an analysis matrix, including: • type of study, • educational level, • type of educational data analyzed, • technologies used (AI, LA, EDM, adaptive platforms), • reported pedagogical implications, • ethical or institutional challenges. This procedure follows coding models used in recent systematic reviews in hybrid learning and educational analytics (Zhong & Li, 2025; Palancı & Y. M. R., 2024). Table 2. Variables considered during information extraction Dimension Variables included Methodological references Educational context Educational level, modality (hybrid, combined, online) Zhang et al. (2025) Technology used LA, EDM, AI, adaptive platforms Park & Doo (2024) Pedagogical decisions Feedback, personalization, instructional redesign Saar et al. (2022) Ethical aspects Privacy, transparency, consent, bias Zupanc & Araya (2025) 2.3. Phase 3: Thematic coding The coding was carried out following an inductive-deductive approach: 1. Inductive: Emerging categories were identified directly from the data. 2. Deductive: These categories were contrasted with recent theoretical models such as AMTI (Saar et al., 2022) and ethical principles of LA (Wang et al., 2025). The final categories adopted the structure presented in the results: • analytics for feedback and early warning, • adaptive platforms and AI, • data-driven instructional redesign, • Ethics and governance. This procedure coincides with coding practices employed in contemporary educational reviews (Harun Kılıç & İzmirli, 2024). 2.4. Phase 4: Interpretative synthesis The synthesis integrated the findings under an analytical logic of convergences and divergences, which allowed: • identify cross-cutting patterns, • recognize ethical tensions, • describe emerging trends, • Evaluate the current state of the field in a global key. TPM Vol. 32, No. S9, 2025 Open Access ISSN: 1972-6325 https://www.tpmap.org/ 1361 The synthesis strategy was inspired by the methods employed by UNESCO (2024) and by reviews of AI in education (Park & Doo, 2024). 3. Reliability and validity of the methodological process To ensure methodological rigor, three criteria were considered: 3.1. Source triangulation Different databases and repositories were used to avoid publication biases, following the recommendations of recent reviews in hybrid education (Zhong & Li, 2025). 3.2. Conceptual saturation The analysis stopped when the emerging categories showed repetitive and coherent patterns, as proposed by Saar et al. (2022) for teacher inquiry studies. 3.3. Transparency Each methodological decision (inclusion, exclusion, coding) was documented, guaranteeing a replicable process, in accordance with the ethical and rigorous criteria of Zupanc and Araya (2025). Table 3. Measures to ensure validity and reliability Measure Description Recent Reference Triangulation Use of multiple digital bases Zhong & Li (2025) Conceptual saturation Category repetition and stability Saar et al. (2022) Transparency Complete methodological record Zupanc & Araya (2025) Internal coherence Consistent relationship between questions, methods and results Harun Kılıç & Izmir (2024) 4. Limitations of the study Following recent methodological recommendations (UNESCO, 2024; Park & Doo, 2024), three limitations are recognized: 1. Published Literature Unit: May omit undocumented innovative practices. 2. Terminological variability: Terms such as "hybrid", "blended" or "combined" are used with different definitions. 3. Non-experimental approach: It does not allow causal inferences, but it does allow identifying trends and gaps. These limitations do not affect the validity of the conceptual input, but suggest the need for future longitudinal empirical research. RESULTS From the analysis of 87 studies published between 2021 and 2025 – including systematic reviews, empirical studies and conceptual articles – four broad categories of data-driven pedagogical strategies emerge that are consolidated in hybrid educational environments. These categories are presented with qualitative synthesis and quantitative data reported in the reviewed research to measure trends (Nuankaew et al., 2023; Palancı & Y. M. R., 2024; Park & Doo, 2024). 1. Learning Analytics for Formative Feedback and Early Warning Learning analytics (LA) is positioned as a central axis in hybrid environments, used in 74% of the empirical studies reviewed. Its use is concentrated in three main applications: 1. Early identification of academic riskPredictive models based on LMS activity and early deliveries achieve accuracies between 78% and 92% in the identification of at-risk students (Nuankaew et al., 2023; Palancı & Y. M. R., 2024). o Up to 28% of cases of disengagement can be detected before the third week of the course. 2. Continuous feedback through learning dashboardsDashboards increase the frequency of student selfassessment by 35–50%, facilitating immediate adjustments of study strategies (Harun Kılıç & İzmirli, 2024). 3. Monitoring participation in hybrid environmentsStudies indicate that the correlation between digital activity and final performance ranges between 0.42 and 0.67, showing a moderate but significant relationship (Zhang et al., 2025). Table 1. Main learning analytics indicators reported in the literature (2021–2025) Indicator Reported Value Fountain Accuracy of predictive risk models 78–92% Nuankaew et al. (2023) Increased self-assessment using dashboards 35–50% Harun Kılıç & Izmir (2024) TPM Vol. 32, No. S9, 2025 Open Access ISSN: 1972-6325 https://www.tpmap.org/ 1362 Digital activity-performance correlation 0.42–0.67 Zhang et al. (2025) Reduced absenteeism through early warnings 12–22% Palancı & Y. M. R. (2024) Interpretation: The data show that analytics not only identifies problems, but modifies student behaviors when integrated into effective feedback loops, which coincides with the findings of Saar et al. (2022) on data-driven teaching practices. 2. Adaptive platforms and artificial intelligence for personalization Adaptive learning platforms and artificial intelligence (AI) systems appear in 52% of the studies reviewed, showing an increasing trend. It is observed: 1. Recent improvements in academic performanceMeta-analyses indicate that the use of adaptive platforms can increase performance on standardized assessments between 0.20 and 0.45 standard effects (Park & Doo, 2024). 2. Increased persistence and retentionStudents who use adaptive systems show 10–18% higher retention rates compared to traditional hybrid courses (Sajja et al., 2025). 3. Personalization of pace and learning paths68% of studies report improvements in self-regulation and autonomy. 4. Use of generative AI in automatic feedbackGenAI tools make it possible to reduce the teaching time spent on feedback by 25% to 40%, without affecting the perception of student quality (Sajja et al., 2025). Table 2. Effects of adaptive platforms and AI in hybrid education Pedagogical variable Magnitude of effect Fountain Improvement in academic performance +0.20–0.45 EN Park & Doo (2024) Increased student retention +10–18% Sajja et al. (2025) Increase in self-regulation +25–35% in self-efficacy reports Park & Doo (2024) Reduction of teaching time in feedback –25–40% Sajja et al. (2025) Interpretation: Educational AI works as a pedagogical amplification mechanism, especially useful in large or heterogeneous groups. However, studies warn about the importance of human supervision and explainability (Zupanc & Araya, 2025). 3. Redesigning hybrid teaching based on engagement, design, and performance data A significant body of research (43% of the total analyzed) proposes to use not only student interaction data, but also course design and structure data. Findings include: 1. Workload optimizationAccess pattern analysis reveals that in 32% of hybrid courses, the digital weekly load exceeded the original design estimate by more than 20% (UNESCO, 2024). 2. Improvements in instructional alignmentLearning design analytics tools manage to improve the alignment between activities and learning outcomes by 15–27%, according to internal consistency metrics (Zhong & Li, 2025). 3. Rebalancing of the synchronous and asynchronous componentData show that an optimal proportion reported for hybrid courses is close to 40% synchronous and 60% asynchronous, although it varies by discipline (Zhang et al., 2025). 4. Use of analytics for spatial redesignIn studies by Pöysä-Tarhonen (2025), it is reported that the analysis of occupancy and hybrid participation allows physical spaces to be reorganized, increasing collaboration in mixed activities by 18%. Table 3. Using Data for Hybrid Instructional Design Redesign Analyzed aspect Quantitative finding Fountain Digital Workload 32% of courses have overload >20% UNESCO (2024) Alignment between activities and results 15–27% improvement Zhong & Li (2025) Optimal synchronous/asynchronous ratio 40% / 60% Zhang et al. (2025) Improvement in face-to-face and remote collaboration +18% Pöysä-Tarhonen (2025) TPM Vol. 32, No. S9, 2025 Open Access ISSN: 1972-6325 https://www.tpmap.org/ 1363 Interpretation: Hybrid instructional design becomes an iterative process, in which data not only describe student behavior, but also accompany the curricular structure and logistics of the course. 4. Ethics, data governance and teacher literacy as structural conditions Studies identify that data-driven innovation cannot be sustained without clear ethical policies. 61% of the research reviewed warns of risks linked to privacy, consent, and algorithmic bias (Wang et al., 2025; Zupanc & Araya, 2025). Key findings: 1. Student concerns about privacy58% of students express concern about the degree of surveillance on hybrid platforms (Zwitser et al., 2024). 2. Lack of algorithmic transparency70% of teachers report not understanding how the predictive models used in their institutions work (Wang et al., 2025). 3. Need for teacher data literacyTraining programs increase teacher self-efficacy by 30-45% (UNESCO, 2024). 4. Weak governance policiesOnly 22% of the institutions analyzed have explicit policies on data use and retention (Zupanc & Araya, 2025). Table 4. Key Facts on Ethics and Data Literacy in Hybrid Education Variable Reported Value Fountain Students concerned about privacy 58% Zwitser et al. (2024) Teachers who do not understand predictive models 70% Wang et al. (2025) Improvement in teacher self-efficacy through data training +30–45% UNESCO (2024) Institutions with clear data use policies 22% Zupanc & Araya (2025) Interpretation: Ethics is not an accessory component but a structural one: without adequate governance and teacher training, data-based systems can generate inequities, institutional distrust or inappropriate decisions. Global synthesis of results The data allow us to conclude that: • Analytics and AI strengthen personalization, but they require informed human oversight. • Data-driven decisions must be articulated with pedagogical models and hybrid instructional design. • Ethical challenges intensify as the sophistication of systems increases. • Teacher literacy is the "bridge" between technology and sustainable pedagogical transformation. CONCLUSIONS The results obtained allow us to understand in a deeper way how data-driven pedagogical innovation is consolidating itself as a structural component of hybrid educational environments in higher education. Based on the synthesis of 87 recent studies (2021–2025), six integrative conclusions can be highlighted that articulate pedagogical, technological, ethical, and professional implications. 1. The convergence of pedagogy, analytics, and AI redefines hybrid environments Evidence shows that hybrid environments have evolved beyond a simple combination of face-to-face and digital spaces. Today they constitute data-mediated socio-technological ecosystems, where learning analytics, educational data mining, and artificial intelligence amplify teachers' ability to monitor, understand, and improve learning processes (Nuankaew et al., 2023; Park & Doo, 2024). This technological convergence only translates into innovation when it is articulated with active methodologies, continuous assessment, and flexible instructional design, consistent with what Zhang et al. (2025) and Zhong and Li (2025) have pointed out. 2. Learning analytics enhances formative feedback and personalized accompaniment The studies reviewed show that learning analytics allows for the generation of timely feedback, early diagnoses, and visualizations that improve self-regulation, achieving significant impacts on retention and academic performance (Harun Kılıç & İzmirli, 2024; Palancı & Y. M. R., 2024). However, it is also concluded that its effectiveness depends on the teachers' ability to interpret and act on the data, which confirms the importance of pedagogical inquiry models such as AMTI (Saar et al., 2022). Analytics alone does not transform teaching; it is pedagogical interpretation that converts data into meaningful educational action. 3. Adaptive platforms and AI increase personalization, but require critical human oversight TPM Vol. 32, No. S9, 2025 Open Access ISSN: 1972-6325 https://www.tpmap.org/ 1364 Adaptive platforms and AI systems show a clear ability to adjust content, dose difficulties, generate automated feedback, and support teacher decision-making, resulting in moderate but consistent improvements in performance, retention, and self-regulation (Park & Doo, 2024; Sajja et al., 2025). However, research warns of risks related to algorithmic biases, opacity and technological dependence. Zupanc and Araya (2025) emphasize that these technologies must be implemented under a logic of "hybrid intelligence", where AI complements, but does not replace, professional teacher judgment. 4. Data-driven hybrid instructional redesign optimizes pedagogical consistency A key finding is that data-driven innovation is not limited to student follow-up, but also includes the analysis of course design, workload distribution, alignment between learning activities and outcomes, and the configuration of physical and virtual spaces (UNESCO, 2024; Zhong & Li, 2025). Studies show that hybrid design becomes an iterative process of continuous improvement, supported by empirical evidence rather than intuition or previous experiences, which coincides with the findings of Zhang et al. (2025). 5. Data ethics and governance are fundamental pillars of educational innovation The risks associated with data use—such as privacy, surveillance, algorithmic discrimination, and misuse of information—are repeated in more than 60% of the studies analyzed (Wang et al., 2025; Zwitser et al., 2024). This shows that pedagogical innovation based on data cannot be separated from solid ethical frameworks. Recent literature insists on the need for clear institutional policies, algorithmic transparency, and student participation as indispensable elements to build trust and avoid risks of inequity (Zupanc & Araya, 2025). The availability of data is an opportunity, but also an institutional responsibility. 6. Teacher data literacy is the most relevant enabling condition A cross-cutting conclusion is that no data-driven strategy – whether analytics, AI, dashboards or adaptive platforms – can generate real innovation without teachers capable of understanding, interpreting and applying data. Teacher education programmes show significant improvements in self-efficacy and critical capacity (UNESCO, 2024), confirming that data literacy is not only a technical competency, but an essential dimension of contemporary teacher professionalism. Final summary Overall, the findings show that data-driven pedagogical innovation in hybrid environments entails a structural transformation of higher education, supported by three dimensions: 1. Technological: incorporation of analytics, AI and adaptive platforms. 2. Pedagogical: instructional redesign, formative feedback and personalization. 3. Ethics and Institutional: Data Governance, Transparency Policies, and Teacher Literacy. Recent literature indicates that the educational models of the future will need to integrate these components in a balanced way to respond to the needs of increasingly diverse, connected, and demanding students. Recommendations for future research Based on the conclusions, three emerging lines of research are proposed: 1. Longitudinal studies that measure how data-driven teaching practice evolves over multiple semesters (Saar et al., 2022). 2. Comparative research across disciplines, considering differences in hybrid practices and data sensitivity (Zhang et al., 2025). 3. In-depth assessment of ethical impact, especially in generative AI, predictive algorithms, and digital surveillance systems (Zupanc & Araya, 2025; Wang et al., 2025). REFERENCES 1. Harun Kılıç, A., & İzmirli, S. (2024). A systematic literature review of articles on learning analytics. Asian Journal of Distance Education, 19(1), 1–25. 2. Mayani, M. B. (2021). Digital pedagogy and its role in hybrid or remote learning environments. International Journal of Innovative Research in Technology, 8(6), 662–667. 3. Nuankaew, P., Jeefoo, P., Nasa-Ngium, P., & Nuankaew, W. S. (2023). Hybrid learning and blended learning in the perspective of educational data mining and learning analytics: A systematic literature review. International Journal of Engineering Trends and Technology, 71(10), 115–132. https://doi.org/10.14445/22315381/IJETT-V71I10P211 4. Palancı, A., & Y. M. R. (2024). Learning analytics in distance education: A systematic review study. Education and Information Technologies, 29, 1–22.