This study employed an updated systematic review and meta-analysis to synthesize quantitative evidence concerning the effects of ethnoscience, indigenous knowledge, local wisdom, and culturally contextualized instructional approaches in science and STEM education. An updated review design was adopted because the current synthesis incorporated one study retained from the previous version of the review and additional studies identified through an updated literature search. The review was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 guidelines (Fitriyah et al., 2025; Fitriyah, Erlangga, et al., 2026b; Page et al., 2021). PRISMA 2020 was used to structure and document the identification, screening, eligibility assessment, and inclusion of studies, thereby supporting transparency, reproducibility, and completeness in the reporting process.

A review protocol was developed before the literature search and screening process. The protocol specified the review objectives, eligibility criteria, information sources, search strategy, study-selection procedures, data-extraction framework, moderator coding, risk-of-bias assessment, effect-size calculation, and statistical-analysis plan. The protocol was not formally registered in an external review registry. The review addressed three main objectives. First, it estimated the overall effect of ethnoscience- and culturally contextualized instructional interventions on science and STEM educational outcomes. Second, it evaluated the magnitude of between-study heterogeneity. Third, it examined whether variation in effect sizes was associated with educational level, outcome type, instructional model, science subject, cultural context, research design, intervention duration, and instructional media. The unit of analysis was the independent study-level comparison between an intervention group and a control or comparison group. Where multiple reports were derived from the same participants or dataset, the reports were linked and treated as a single underlying study to avoid double-counting participants.

Studies were selected using predefined inclusion and exclusion criteria established before the screening process. These criteria were designed to ensure that the included studies were conceptually relevant to ethnoscience and science education and provided sufficient quantitative information for calculating standardized effect sizes. Eligible studies examined educational interventions that integrated ethnoscience, indigenous knowledge, indigenous science, local wisdom, traditional ecological knowledge, culturally grounded science, culture-based learning, culturally responsive science education, or ethno-STEM into science or STEM instruction. Participants could include primary school students, secondary school students, higher education students, pre-service teachers, or in-service teachers involved in science- or STEM-related education.

The quantitative synthesis included studies employing experimental, quasi-experimental, or controlled comparative designs. Studies were required to provide sufficient statistical information to calculate a standardized mean difference, including sample sizes, means, standard deviations, standard errors, confidence intervals, or other convertible statistics. Eligible educational outcomes included academic achievement, scientific knowledge, conceptual understanding, scientific literacy, thinking and reasoning skills, motivation, awareness, attitudes, self-efficacy, and science-related identity. Studies were excluded when they were conceptual, qualitative, descriptive, correlational, non-intervention, or review studies, or when they did not provide sufficient quantitative information for effect-size calculation. The detailed inclusion and exclusion criteria are presented in Table 1.

When multiple publications reported findings from the same participants, intervention, or dataset, they were treated as a single study. The publication containing the most complete methodological description and quantitative data was retained. This procedure was used to maintain the statistical independence of the included effect sizes and prevent the same participants from being counted more than once.

A comprehensive literature search was conducted across international bibliographic databases, research registers, relevant websites, organisational sources, and citation networks. Database and register searching was supplemented by backward reference-list screening, forward citation searching, targeted website searching, and searches of relevant educational and research organisations. The search strategy consisted of five conceptual components: cultural knowledge, science or STEM education, educational participants, intervention research, and educational outcomes. Terms related to cultural knowledge included “ethnoscience,” “indigenous knowledge,” “indigenous science,” “local wisdom,” “traditional ecological knowledge,” “culturally grounded science,” “culturally responsive science,” “culture-based science,” and “ethno-STEM.”

These terms were combined with science-education terms, including “science education,” “science learning,” “STEM education,” “biology education,” “physics education,” “chemistry education,” “earth science,” and “environmental science.” Participant-related terms included “student,” “learner,” “pupil,” “teacher,” “pre-service teacher,” and “in-service teacher.” Research-design terms included “experiment,” “intervention,” “quasi-experiment,” “control group,” “comparison group,” “pretest,” and “posttest.” Outcome-related terms included “achievement,” “knowledge,” “conceptual understanding,” “scientific literacy,” “reasoning,” “thinking skills,” “motivation,” “attitude,” “awareness,” “self-efficacy,” and “identity.”

The general search structure was as follows:

(“ethnoscience” OR “indigenous knowledge” OR “indigenous science” OR “local wisdom” OR “traditional ecological knowledge” OR “culturally grounded science” OR “culturally responsive science” OR “culture-based science” OR “ethno-STEM”) AND (“science education” OR “science learning” OR “STEM education” OR biology OR physics OR chemistry OR “earth science” OR “environmental science”) AND (student* OR learner* OR pupil* OR teacher* OR “pre-service teacher*” OR “in-service teacher*”) AND (experiment* OR intervention* OR quasi-experiment* OR “control group” OR “comparison group” OR pretest OR posttest) AND (achievement OR knowledge OR “conceptual understanding” OR “scientific literacy” OR reasoning OR “thinking skills” OR motivation OR attitude* OR awareness OR self-efficacy OR identity)

The search syntax was adapted to the field codes, phrase-searching functions, truncation rules, and Boolean requirements of each information source. Searches were applied to titles, abstracts, and keywords whenever these fields were available. No restriction based on educational level was applied during the initial search. Broad search terms were used to maximize search sensitivity, while the relevance and eligibility of the retrieved publications were determined during title and abstract screening and full-text assessment. Reference lists of potentially eligible studies were examined to identify additional publications (Arztmann, 2023; Budniok et al., 2026). Forward citation searches were also conducted to locate more recent studies that cited relevant publications. All records retrieved from the different sources were exported to reference-management software before deduplication and screening. The search identified 3,360 records from bibliographic databases and 37 records from research registers. An additional 100 records were identified through supplementary methods, comprising 42 records from websites, 18 from organisations, 25 through citation searching, and 15 from other relevant sources.

The study-selection process followed the identification, screening, eligibility, and inclusion stages specified in PRISMA 2020. Records identified through databases, registers, websites, organisations, citation searching, and other supplementary sources were compiled into a single review database using reference-management software. Duplicate records were initially identified through automated matching of titles, authors, publication years, and digital identifiers. The automatically detected duplicates were subsequently checked manually to prevent the incorrect removal of publications with similar titles or authorship information (Ding, 2023; Zohar & Ben-Ari, 2022).

Before title and abstract screening, 412 duplicate records were removed. An automated screening procedure marked 150 records as ineligible based on predefined bibliographic and content-related criteria. A further 50 records were removed for other predefined reasons, including clearly ineligible publication types, incomplete records, and documents outside the scope of the review. Following these removals, 2,785 records remained for title and abstract screening. Title and abstract screening was conducted independently by two reviewers using systematic-review screening software. Each record was assessed against the predefined eligibility criteria. Records were excluded at this stage when they clearly failed to address ethnoscience, indigenous knowledge, local wisdom, culturally contextualized instruction, science education, STEM education, relevant participant groups, educational interventions, or measurable educational outcomes.

The reviewers completed the screening independently before comparing their decisions. Disagreements were resolved through discussion and consensus. When agreement could not be reached, a third reviewer adjudicated the eligibility decision. Of the 2,785 records screened, 2,618 were excluded during title and abstract screening. The remaining 167 reports were sought for full-text retrieval. Seven reports could not be retrieved because the full text was unavailable, resulting in 160 reports assessed for eligibility. Full-text assessment was also conducted independently by two reviewers. Each report was evaluated against all inclusion and exclusion criteria presented in Table 1. The reviewers recorded one primary reason for excluding each ineligible report to avoid double-counting exclusion reasons.

Among the 160 reports assessed from the database and register pathway, 133 were excluded. The reasons were an irrelevant topic in 58 reports, an ineligible research design in 39 reports, insufficient quantitative data in 25 reports, and other eligibility-related reasons in 11 reports. The supplementary search pathway identified 100 reports. Five reports could not be retrieved, leaving 95 reports for full-text eligibility assessment. Of these, 92 were excluded because 40 were off-topic, 27 employed an ineligible research design, 18 lacked sufficient quantitative information, and seven did not meet other eligibility criteria. The updated search produced 30 newly eligible studies. One study was retained from the previous version of the review, resulting in a total of 31 studies included in the systematic review and meta-analysis. The complete study-selection process is presented in the PRISMA 2020 flow diagram in Figure 1.

A standardized data-extraction form was used to record study identification, publication year, participant characteristics, educational level, research design, intervention characteristics, outcome measures, sample sizes, means, and standard deviations. When necessary, other reported statistics were converted into a format suitable for effect-size calculation. Data were extracted from the main text, tables, figures, and supplementary materials and were independently verified by a second reviewer. Discrepancies were resolved by rechecking the original publication. Eight moderators were coded to examine potential sources of variation in effect sizes: educational level, outcome type, instructional model, science subject, cultural context, research design, intervention duration, and media type. These moderators were selected because participant characteristics, pedagogical approaches, disciplinary content, cultural integration, methodological design, intervention intensity, and learning resources may influence intervention outcomes. The operational definitions and coding categories for these moderator variables are summarized in Table 2.

Coding was based on the primary participant group, principal outcome, dominant instructional model, main science subject, central cultural orientation, and main instructional medium reported in each study. Duplicate reports, conceptual studies, and studies unsuitable for quantitative synthesis were not treated as moderator categories.

The methodological quality of the included studies was assessed using a structured risk-of-bias framework appropriate for experimental and quasi-experimental educational research. The assessment covered participant selection, baseline comparability, group allocation, intervention implementation, missing outcome data, outcome measurement, and selective reporting. Each domain was rated as low risk, some concerns, or high risk of bias. An overall judgment was assigned based on the pattern and severity of domain-level concerns. Two reviewers independently completed the assessment, and disagreements were resolved through discussion or consultation with a third reviewer. Risk-of-bias judgments were not converted into numerical quality scores and were not used as statistical weights. Instead, they were considered in the interpretation of findings and in sensitivity analyses examining whether the pooled estimate changed after studies with high risk of bias were excluded.

The intervention effect was expressed as Hedges’ g, a bias-corrected standardized mean difference. This measure was used because the included studies assessed related educational outcomes using different instruments and scoring scales. Positive effect sizes indicated more favourable outcomes for the intervention group. For outcomes in which lower scores represented better performance, the effect direction was reversed. The quantitative synthesis included 31 independent study-level effects and 5,164 observations, consisting of 2,580 observations in intervention groups and 2,584 in comparison groups. Each study contributed one independent effect estimate to the primary model. When a study reported multiple relevant outcomes, the primary outcome was selected or closely related measures were combined to avoid overrepresentation of a single study. Reports based on the same sample or dataset were treated as one independent study (Gunnoo et al., 2023; Ribeiro et al., 2026).

A random-effects model was applied because the true intervention effects were expected to vary across educational levels, instructional models, science subjects, cultural contexts, research designs, intervention durations, media types, and outcome measures. Studies were weighted using the inverse-variance method, while between-study variance was estimated using restricted maximum likelihood. Confidence intervals for the heterogeneity variance were calculated using the Q-profile method. All effect estimates were reported with 95% confidence intervals, and statistical significance was set at p < .05.

Between-study heterogeneity was assessed using Cochran’s Q, I², τ², τ, and H. These statistics were interpreted jointly with the magnitude, direction, and precision of individual effect sizes. Substantial heterogeneity was considered evidence that study-level characteristics could contribute to variation in intervention effects. Potential sources of heterogeneity were examined using subgroup analyses and mixed-effects meta-regression (Maringa et al., 2025; Varah et al., 2025). The prespecified moderators were educational level, outcome type, instructional model, science subject, cultural context, research design, intervention duration, and media type. Each moderator was first examined separately using random-effects subgroup analysis or univariable meta-regression. Differences between moderator categories were evaluated using an omnibus test rather than by comparing the significance of individual subgroup estimates. Categories represented by too few studies were combined when conceptually appropriate or reported descriptively.

Because the meta-analysis included 31 studies, multivariable meta-regression was restricted to a limited number of theoretically relevant predictors to reduce the risk of overfitting. Educational level, outcome type, instructional model, cultural context, and intervention duration were treated as the principal substantive moderators. Science subject and media type were examined as additional instructional moderators, while research design was treated as a methodological moderator and was also considered in sensitivity analysis. Moderator findings were interpreted cautiously because subgroup sizes were unequal and several study characteristics could be correlated. Interpretation therefore considered effect magnitude, confidence intervals, residual heterogeneity, theoretical plausibility, and the distribution of studies across moderator categories.

All statistical analyses were conducted using R within the RStudio integrated development environment. The meta package was used to calculate Hedges’ g, estimate the pooled effect under the random-effects model, assess heterogeneity, conduct subgroup analyses, and generate forest and funnel plots. The metafor package was used for meta-regression, influence diagnostics, and sensitivity analyses. Studies were weighted using the inverse-variance method, and between-study variance was estimated using the restricted maximum-likelihood estimator (Wood & Hedges, 2024). Confidence intervals for τ² were calculated using the Q-profile method. Statistical tests were two-sided, with p < .05 considered statistically significant. All data-processing procedures, coding decisions, and statistical outputs were checked before interpretation to ensure consistency between the analytical dataset and the reported results.

The study-selection process is presented in Figure 1. The search identified 3,497 records, comprising 3,360 records from bibliographic databases, 37 from research registers, and 100 from supplementary sources. The supplementary records consisted of 42 records identified through websites, 18 through organisations, 25 through citation searching, and 15 through other sources. Before screening, 412 duplicate records were removed, 150 records were marked as ineligible through the automated screening procedure, and 50 records were removed for other predefined reasons. This process left 2,785 records for title and abstract screening, of which 2,618 were excluded because they did not meet the preliminary eligibility criteria.

From the database and register pathway, 167 reports were sought for retrieval. Seven reports could not be retrieved, leaving 160 reports for full-text eligibility assessment. Following full-text review, 133 reports were excluded because they addressed an irrelevant topic (n = 58), employed an ineligible research design (n = 39), lacked sufficient quantitative information (n = 25), or did not meet other eligibility requirements (n = 11). From the supplementary-search pathway, 100 reports were sought for retrieval. Five reports could not be obtained, and 95 were assessed in full text. Of these, 92 were excluded because they were off-topic (n = 40), used an ineligible research design (n = 27), lacked sufficient quantitative data (n = 18), or failed to meet other eligibility criteria (n = 7). The updated search yielded 30 newly eligible studies. Together with one study retained from the previous version of the review, 31 studies were included in the final systematic review and quantitative synthesis.

The quantitative synthesis included 31 study-level estimates and 5,164 observations, consisting of 2,580 observations in the intervention conditions and 2,584 in the comparison conditions. As shown in Figure 2, the random-effects model produced a positive pooled standardized mean difference of g = 1.17, 95% CI [0.88, 1.46], z = 7.83, p < .001. The average standardized difference therefore favoured the intervention conditions. Nevertheless, the pooled estimate represents an average across studies with substantially different effect sizes and should not be interpreted as a uniform effect across all settings. The individual effect estimates displayed in Figure 2 ranged from −0.03 to 4.74.

Several estimates had confidence intervals that included zero, indicating considerable statistical uncertainty at the individual-study level. For example, the estimate for Londoño-Proaño and Pérez-Gavilanes was close to zero, g = −0.03, 95% CI [−0.31, 0.24], whereas the largest estimate was reported for Ghazali et al., g = 4.74, 95% CI [3.62, 5.85]. Most plotted estimates were positive, but their magnitude and precision varied considerably. Substantial between-study heterogeneity was observed, Q(30) = 354.11, p < .001, I² = 91.5%, τ² = 0.5766, τ = 0.7593, and H = 3.44. Figure 2 reports I² in rounded form as 92%. These statistics indicate that the variability among the observed effect estimates was considerably greater than would be expected from sampling error alone. The high heterogeneity suggests that the magnitude of the standardized differences may vary according to study population, instructional approach, outcome construct, disciplinary context, research design, intervention duration, media characteristics, or other unmeasured study-level factors. Consequently, the pooled result was interpreted alongside the moderator analyses rather than as an effect expected to occur consistently in every educational context.

Moderator analyses were conducted to examine whether variation in effect estimates was associated with educational level, outcome type, instructional model, science subject, cultural context, media type, and intervention duration. Figure 3 presents the number of studies, subgroup-specific effect estimates, confidence intervals, and omnibus interaction tests for each moderator, while Table 3 summarizes the corresponding moderator results and their statistical interpretations.

3.3.1 Education level

The subgroup analysis included five estimates classified as primary education, nine as higher or teacher education, 15 as secondary education, and two as curriculum-conceptual studies. The pooled estimates were g = 1.54, 95% CI [−0.00, 3.09], for primary education; g = 0.89, 95% CI [0.60, 1.18], for higher or teacher education; g = 1.19, 95% CI [0.81, 1.57], for secondary education; and g = 1.34, 95% CI [0.18, 2.51], for curriculum-conceptual studies. The omnibus interaction test was not statistically significant, p = .51. Thus, although the subgroup point estimates differed numerically, the analysis did not provide sufficient evidence that effect magnitude varied systematically across educational levels. The primary-education estimate was also imprecise, as indicated by its wide confidence interval.

3.3.2 Outcome type

The pooled standardized mean differences were g = 0.77, 95% CI [0.39, 1.14], for achievement and knowledge; g = 1.24, 95% CI [0.87, 1.62], for thinking and reasoning skills; g = 1.27, 95% CI [0.43, 2.11], for conceptual literacy; g = 1.20, 95% CI [0.94, 1.45], for self-efficacy and identity; g = 1.00, 95% CI [0.86, 1.15], for professional curriculum outcomes; and g = 1.77, 95% CI [−1.09, 4.63], for motivation, awareness, and attitudes. The omnibus interaction test was not statistically significant, p = .37. The available data therefore did not indicate statistically detectable differences among outcome categories. The motivation, awareness, and attitudes subgroup showed the largest point estimate, but its wide confidence interval indicates substantial uncertainty.

3.3.3 Instructional model

The pooled estimates were g = 0.98, 95% CI [0.63, 1.32], for media-assisted learning; g = 0.65, 95% CI [0.15, 1.14], for collaborative experiential learning; g = 2.03, 95% CI [0.87, 3.20], for inquiry- or project-based STEM; g = 1.14, 95% CI [0.94, 1.33], for culture-context-based learning; and g = 1.00, 95% CI [0.86, 1.15], for professional curriculum approaches. Although inquiry- or project-based STEM had the largest subgroup point estimate, the omnibus interaction test was not statistically significant, p = .16. Accordingly, the observed numerical differences should not be interpreted as evidence that one instructional model was consistently more effective than the others.

3.3.4 Science subject

The subgroup estimates were g = 1.27, 95% CI [0.72, 1.81], for biology and environmental science; g = 1.21, 95% CI [0.77, 1.66], for general STEM; and g = 0.94, 95% CI [0.33, 1.56], for physics and earth science. The omnibus interaction test was not statistically significant, p = .71. The overlapping confidence intervals and nonsignificant test indicate that the analysis did not detect systematic variation in effect magnitude across science-subject categories.

3.3.5 Cultural context

The estimated effects were highly similar across cultural-context categories. General-context studies produced g = 1.15, 95% CI [0.60, 1.70]; local-context studies produced g = 1.19, 95% CI [0.64, 1.74]; and studies coded as indigenous ethnoscience produced g = 1.14, 95% CI [0.95, 1.34]. The omnibus interaction test was not statistically significant, p = .99. The current analysis therefore did not indicate that effect magnitude differed according to the cultural-context classification used in the dataset.

3.3.6 Media type

The media-type estimates were g = 1.22, 95% CI [0.88, 1.56], for AI and digital tools; g = 1.28, 95% CI [0.84, 1.73], for modules and activity materials; g = 1.04, 95% CI [0.33, 1.74], for simulations and animations; g = 0.97, 95% CI [0.60, 1.34], for conceptual curriculum frameworks; g = 1.12, 95% CI [0.73, 1.51], for teacher professional-development programmes; g = 2.62, 95% CI [−1.46, 6.69], for digital platforms and open educational resources; g = 0.22, 95% CI [−0.15, 0.60], for game- and narrative-based media; and g = 1.19, 95% CI [0.89, 1.49], for the survey-questionnaire category. The omnibus interaction test was statistically significant, p < .01, suggesting that effect estimates differed across the media categories as coded in the dataset. However, this result should be interpreted cautiously. Several subgroups contained only one to three estimates, and the digital-platform and open-educational-resource subgroup had a particularly wide confidence interval. The statistically significant omnibus test therefore indicates possible between-category variation but does not establish that a particular medium is consistently superior.

3.3.7 Intervention duration

The estimated effects were g = 1.19, 95% CI [0.79, 1.59], for medium-duration interventions; g = 1.32, 95% CI [0.74, 1.89], for the long-or-non-intervention category; and g = 0.89, 95% CI [0.25, 1.53], for short interventions. The omnibus interaction test was not statistically significant, p = .61. Thus, the analysis did not detect systematic differences across the duration categories. Nevertheless, the “long or non-intervention” category combines conceptually different conditions, so its estimate should be treated as provisional until long-duration interventions and non-intervention studies are coded separately. Overall, media type was the only moderator with a statistically significant omnibus interaction test. The remaining moderators showed numerical variation in subgroup point estimates, but their interaction tests were not statistically significant. These findings should be treated as exploratory because several subgroups were small, confidence intervals were sometimes wide, and some coding categories combined methodologically different study characteristics.

The funnel plot is presented in Figure 4. Most studies were concentrated around standardized mean differences between approximately 0 and 2, with standard errors generally below 0.40. However, several studies were positioned outside the central region of the funnel, including one study with a standardized mean difference close to 4.7 and several studies with relatively large standard errors. The distribution was therefore not fully symmetrical. Visual asymmetry may indicate potential small-study effects, selective reporting, or differences in the methodological characteristics of smaller and larger studies. Nevertheless, the funnel plot alone does not provide sufficient evidence of publication bias. The observed pattern may also reflect substantial between-study heterogeneity, variation in research design, differences in outcome measurement, or differences in intervention and participant characteristics. Because the meta-analysis showed high heterogeneity, the funnel plot should be interpreted cautiously. A formal statistical test of funnel-plot asymmetry, such as Egger’s regression test, would be required to determine whether the asymmetry was statistically detectable. In the absence of a formal test result, the current evidence suggests possible asymmetry but does not confirm publication bias.

The random-effects meta-analysis of 31 study-level estimates produced a pooled standardized mean difference of g = 1.17, 95% CI [0.88, 1.46], indicating an average difference in favour of the intervention conditions. However, the effect estimates varied considerably across studies, ranging from approximately −0.03 to 4.74. The high heterogeneity, I² = 91.5%, indicates that the pooled effect should be interpreted as the mean of a broad distribution of study effects rather than as a uniform effect across educational contexts.

The moderator analyses did not identify statistically detectable differences according to educational level, outcome type, instructional model, science subject, cultural context, or intervention duration. Media type was the only moderator with a statistically significant omnibus interaction test. Nevertheless, this finding should be regarded as exploratory because several media categories contained few studies and some subgroup estimates had wide confidence intervals. The funnel plot showed possible asymmetry, but this pattern cannot be attributed exclusively to publication bias. Substantial heterogeneity and differences in study design, sample size, outcome measurement, and intervention characteristics may also have contributed to the observed distribution. Accordingly, the results indicate a positive average association between the examined interventions and educational outcomes, while the magnitude and consistency of the effects varied across the included studies (Gulnara et al., 2025).

The authors would like to express their gratitude to the Indonesian Endowment Fund for Education and the Center for Higher Education Funding and Assessment for their financial and administrative support through the Indonesian Education Scholarship, Doctoral Scholarship Program for Indonesian Lecturers. The authors also acknowledge the Ministry of Higher Education, Science, and Technology of the Republic of Indonesia for supporting doctoral research and academic development. In addition, appreciation is extended to the institutions that provided access to bibliographic databases and analytical tools essential for this study.

•  Ahanonye UA, Otulaja F, Risenga I, et al.: The Representation of Indigenous Knowledge in the South African Further Education and Training Phase Life Sciences Curriculum. Afr. J. Res. Math., Sci. Technol. Educ. 2026; 1–14. Publisher Full Text
•  Ahmmad M, Shahzad K, Iqbal A, et al.: Trap of Social Media Algorithms: A Systematic Review of Research on Filter Bubbles, Echo Chambers, and Their Impact on Youth. Societies. 2025; 15(11). Publisher Full Text
•  Aikenhead GS, Jegede OJ: Cross-cultural science education: A cognitive explanation of a cultural phenomenon. J. Res. Sci. Teach. 1999; 36(3): 269–287. Publisher Full Text
•  Ali LU, Suranto S, Indrowati M: Analysis of Primary Science Teachers’ Lesson Plans on Their Awareness of Local Content in Lombok, Indonesia. Munaddhomah: Jurnal Manajemen Pendidikan Islam. 2026; 7(1): 216–232. Publisher Full Text
•  Amreta MY: Development of Numeracy Assessment Based on Ethnomathematics in the Merdeka Curriculum Phase B. Syaikhuna: Jurnal Pendidikan dan Pranata Islam. Cholil Bangkalan: Sekolah Tinggi Agama Islam Syaichona Moh; 2024; 15(1): 12–24. Publisher Full Text
•  Anggreini: Implementation of Ethno-STEM based inquiry learning model assisted by LKPD to train creative thinking skills learners in the Merdeka curriculum learning. Journal of Physics: Conference Series. 2024; 2900(1). Publisher Full Text
•  Arztmann M: Effects of games in STEM education: a meta-analysis on the moderating role of student background characteristics. In Studies in Science Education. 2023; 59(1): 109–145. Publisher Full Text
•  Asamoah P, Zokpe D, Boateng R, et al.: Domain knowledge, ethical acumen, and query capabilities (DEQ): a framework for generative AI use in education and knowledge work. Cogent Education. 2024; 11(1). Publisher Full Text
•  Boxerman JZ, Sangueza CR, Nelson-barber S: Culturally Grounded STEM Education: Three Cases of Broadening Participation Among Indigenous Islanders.2026; 1–28.
•  Budniok S, Bakermans-Kranenburg M, Bosmans G, et al.: Epistemic trust as a window into the allostatic function of oxytocin: A synthesis of rodent and human studies. Neurosci. Biobehav. Rev. 2026; 181: 106500. PubMed Abstract | Publisher Full Text
•  Ding M: Can public service motivation increase work engagement?—A meta-analysis across cultures. In Frontiers in Psychology. 2023; 13. Publisher Full Text
•  Ee LS, Ibrahim NH, Surif J, et al.: Project-based environmental module for indigenous students in Malaysia. International Journal of Evaluation and Research in Education. 2025; 14(3): 2042–2050. Publisher Full Text
•  Erlangga SY, Sarwanto, Harlita: Unpacking the cognitive and ethical pathways of generative AI tools in higher education: a PLS-SEM study of learning performance mediation and moderation effects. Research in Learning Technology, 34(SE-Original Research Articles). 2026; 34. Publisher Full Text
•  Fan L, Wang Y, Wan M, et al.: Exploring the Embeddedness of Cultural and Tourism Integration in Traditional Villages for Common Prosperity: A Case Study of the “Tai Guan Ren” Festival in Huanggang Dong Village, Qiandongnan, Guizhou. China. Tropical Geography. 2025; 45(9): 1605–1616. Publisher Full Text
•  Festiyed, Ismail IA, Lufri L, et al.: Culturally responsive science education through ethnoscience: A meta-analysis of impact on junior high school students. Multidisciplinary Reviews. 2025; 9(1): e2026004. Publisher Full Text
•  Fianti F, Neratania A: Developing Physics Teaching Materials Based on Differentiated Merdeka Curriculum Using an Ethnoscience-Integrated Contextual Approach. Jurnal Inovasi Pendidikan IPA. Universitas Negeri Yogyakarta; 2024; 10(2): 160–174. Publisher Full Text
•  Fitria D: Enhancing 21st-Century Skills through Blended Problem-Based Learning with Ethnoscience Integration: A Mixed-Methods Study in Indonesian Junior High Schools. International Journal of Learning Teaching and Educational Research. 2025; 24(1): 464–480. Publisher Full Text
•  Fitriyah IJ, Erlangga SY, Muqoyyidin AW: Evaluating trust in education: a systematic review of the ethical challenges of deepfake technologies relevant to education. Discov. Educ. 2026a. Publisher Full Text
•  Fitriyah IJ, Saputro S, Muqoyyidin AW, et al.: Supervisor support as a determinant of mental well-being and academic resilience among doctoral students: a qualitative study in Indonesian higher education. Discov. Psychol. 2026b; 6. Publisher Full Text
•  Fitriyah IJ, Erlangga SY, Muqoyyidin AW, et al.: Exploring the role of moderators in the effectiveness of digital games for stem education: a systematic review and meta-analysis. Discov. Educ. 2025; 4(1): 297. Publisher Full Text
•  Gulnara G, Ramankulov S, Madina M, et al.: STEM-based approaches to soft skills development: a synthesis of meta-analytic findings and empirical evidence. Frontiers in Education. 2025; 10. Publisher Full Text
•  Gunnoo L, Subadar UA, Fauzel S: The mediating effect of attitude on adoption of mobile payment services in small island developing states. The case of Mauritius. Social Sciences & Humanities Open. 2023; 8(1): 100724. Publisher Full Text
•  Hsin CT: Implementing a Project-Based Learning Module in Urban and Indigenous Areas to Promote Young Children’s Scientific Practices. Res. Sci. Educ. 2023; 53(1): 37–57. Publisher Full Text
•  Hsu P-L: Regulating Emotions through Cogenerative Dialogues to Sustain Student Engagement in Science Internships: A Case Study. Sustainability (Switzerland). 2022; 14(10). Publisher Full Text
•  IntHout J, Ioannidis JPA, Rovers MM, et al.: Plea for routinely presenting prediction intervals in meta-analysis. BMJ Open. 2016; 6(7): e010247. PubMed Abstract | Publisher Full Text | Free Full Text
•  Juuso Á: Between a rock and hard place: discussing Sámi issues in contemporary Finland – a conversation with Ánneristen Juuso. Decolonising Social Work in Finland. Policy Press; 2024; 35–48. Publisher Full Text
•  Kuncoro KS, Fitriasari P, Erlangga SY, et al.: Artificial Intelligence and Intelligent Tutoring Systems in Mathematics Education: A Bibliometric Analysis (2001?2025) [version 1; peer review: 3 approved with reservations]. F1000Res. 2026; 15(107). Publisher Full Text
•  Latip A, Kadarohman A: Local and indigenous knowledge (LIK) in science learning: A systematic literature review. Journal of Turkish Science Education. 2024; 21(4): 651–667. Publisher Full Text
•  Le Marchand M, Ben Rais Lasram F, Araignous E, et al.: Towards an ecosystem approach to a simulated floating wind farm combined with climate change in the Bay of Biscay (France). Reg. Stud. Mar. Sci. 2025; 87: 104218. Publisher Full Text
•  Mallibhat K: Scaffolding Models in a Project-Based Learning Course-A Case Study. Journal of Engineering Education Transformations. 2024; 37(Special Issue 2): 796–803. Publisher Full Text
•  Maringa JN, Imre G, György M: Understanding the history and operations of Al Shabaab in Kenya through the lens of Kant’s categorical imperative of moral philosophy. Social Sciences & Humanities Open. 2025; 11: 101510. Publisher Full Text
•  Marsuki MF, Erlangga SY, Ali LU, et al.: Supporting Data and Reporting Materials for "Educational Effects of Ethnoscience, Indigenous Knowledge, and Local Wisdom in Science and STEM Education" (Version 1.0). [Dataset]. Zenodo. 2026. Publisher Full Text
•  Mdleleni S, Ngcoza KM: Using Topic-Specific Pedagogical Content Knowledge to Study Indigenous Knowledge Integration in Grade 11 Physical Science Lessons: Teachers’ Perspectives. Afr. J. Res. Math., Sci. Technol. Educ. 2025; 29(1): 128–141. Publisher Full Text
•  Mlotshwa N, Tsakeni M: Challenges Experienced by Teachers when Integrating Indigenous Knowledge Systems in Natural Sciences Practical Work. Afr. J. Res. Math., Sci. Technol. Educ. 2024; 28(2): 300–313. Publisher Full Text
•  Monk S: Using Arts-Based Inquiry and Affective Learning to Teach Indigenous Studies in a First-Year Preservice Teacher Education Course. J. Transform. Educ. 2023; 21(2): 303–321. Publisher Full Text
•  Muliadi A, Rokhmat J, Sukarso AA: Ethnoscience integrated project-based learning model for enhancing students’ creative thinking skills and cultural literacy: Expert perspectives and preliminary testing. International Journal of Innovative Research and Scientific Studies. 2025; 8(3): 3676–3685. Publisher Full Text
•  Mund SI: Caring for Pacific salmon: Reconsidering salmon-human relationships. Anthropol. Today. 2025; 41(3): 4–6. Publisher Full Text
•  Ogegbo AA, Ramnarain U: A Systematic Review of Pedagogical Practices for Integrating Indigenous Knowledge Systems in Science Teaching. Afr. J. Res. Math., Sci. Technol. Educ. 2024; 28(3): 343–361. Publisher Full Text
•  Ogunniyi MB: Teachers’ Stances and Practical Arguments Regarding a Science-Indigenous Knowledge Curriculum: Part 1. Int. J. Sci. Educ. 2007; 29(8): 963–986. Publisher Full Text
•  Oladipupo PO, Olawale-Jimoh AR, Orioye AA, et al.: Indigenous practices and their influence on senior secondary biology students’ academic performance and self-concept. Discov. Educ. 2025; 4(1). Publisher Full Text
•  Page MJ, Mckenzie JE, Bossuyt PM, et al.: The PRISMA 2020 statement: an updated guideline for reporting systematic reviews Systematic reviews and Meta-Analyses. 2021; 372. . PubMed Abstract | Publisher Full Text | Free Full Text
•  Peize W, Sangiamvibool A: Visual Anthropology and Cultural Integrity: Cultivating Long-Term Field Relations. Cultura. International Journal of Philosophy of Culture and Axiology. 2025; 22(2): 1–14. Reference Source
•  Prayogi S, Ahzan S, Rokhmat J, et al.: Dynamic blend of ethnoscience and inquiry in a digital learning platform (e-learning) for empowering future science educators’ critical thinking. Journal of Education and E-Learning Research. 2023; 10(4): 819–828. Publisher Full Text
•  Prihatni Y, Erlangga SY, Saryanto S, et al.: Measurement and Structural Modelling of Epistemic Regulation Under Algorithmic Ambiguity in AI-Mediated Science Learning [version 1; peer review: 2 approved with reservations]. F1000Res. 2026; 15(804). Publisher Full Text
•  Putra ZH, Hermita N, Alim JA, et al.: The role of project-based ethno-didactics of mathematics learning toward pre-service elementary teachers’ perspectives, mathematical, and didactic knowledge. Frontiers in Education. 2025; 10. Publisher Full Text
•  Rahmawati R, Marmoah S, Roslan RM: Ethnoscience Materials to Build Scientific Literacy. Seminars in Medical Writing and Education. 2025; 4: 747. Publisher Full Text
•  Rathore S, Raj Y, Debnath P, et al.: Ethnopharmacology, phytochemistry, agrotechnology, and conservation of Inula racemosa Hook f.–A critically endangered medicinal plant of the western Himalaya. Journal of ethnopharmacology. 2022. Reference Source
•  Ribeiro I, Lopes H, Remoaldo PC, et al.: Contribution of nature-based solutions (NbS) to resilience - A review of European projects between 2014 and 2023. Environ. Dev. 2026; 57: 101367. Publisher Full Text
•  Sakata N, Bremner N, Bórquez-Morales LS: The Nueva Escuela Mexicana reform in Mexico: Promises and tensions in revaluing the people’s Mexico. Int. J. Educ. Dev. 2025; 116: 103295. Publisher Full Text
•  Snively G, Corsiglia J: Discovering indigenous science: Implications for science education. Sci. Educ. 2001; 85(1): 6–34. Publisher Full Text
•  Sotero MC, Alves ÂGC, Arandas JKG, et al.: Local and scientific knowledge in the school context: characterization and content of published works. J. Ethnobiol. Ethnomed. 2020; 16(1): 23. PubMed Abstract | Publisher Full Text | Free Full Text
•  Stewart GT: Smoke and Mirrors: Indigenous Knowledge in the School Curriculum. Curriculum Studies Worldwide. Springer International Publishing; 2021; 41–56. Publisher Full Text
•  Tolbert S, Hipkins R, Cowie B, et al.: Epistemic agency, Indigenous knowledge, and the school science curriculum: reflections from Aotearoa New Zealand. Int. J. Sci. Educ. 2025; 47(15–16): 1986–2002. Publisher Full Text
•  Varah A, Purvis A, de Palma A : Biodiversity response to local-scale conservation interventions: A systematic review and meta-analysis protocol. Ecological Solutions and Evidence. 2025; 6(3). Publisher Full Text
•  Wahyudi W: Hybrid Ethno-Project Based Learning Integrated With Virtual Assistive Technology to Enhance Students’ Critical Thinking in Fundamental Physics Course. TEM Journal. 2023; 12(4): 2006–2012. Publisher Full Text
•  Winingsih P, Kuswanto H, Saputro H, et al.: Analysis of understanding of physics concepts through problem solving units review in free fall motion materials. Revista Mexicana de Física E. 2023; 20. Publisher Full Text
•  Wood E, Hedges H: Curriculum in early childhood education: Global policy discourses and country-level responses in Aotearoa New Zealand and England. The Curriculum Journal. Wiley; 2024. Publisher Full Text
•  Zidny R, Sjöström J, Eilks I: A Multi-Perspective Reflection on How Indigenous Knowledge and Related Ideas Can Improve Science Education for Sustainability. Sci. & Educ. 2020; 29(1): 145–185. Publisher Full Text
•  Zigon J: A Critical Phenomenological-Hermeneutics of Us. Zeitschrift Fur Ethnologie. 2023; 148(1): 127–130. Reference Source
•  Zohar A, Ben-Ari G: Teachers’ knowledge and professional development for metacognitive instruction in the context of higher order thinking. Metacogn. Learn. 2022; 17(3): 855–895. Publisher Full Text