Research Article
Ethnoeducation, Afro-descendant youth leadership, and artificial intelligence literacy: an exploratory study with educational actors from Esmeraldas, Ecuador
Synthesis: Medranda Morales and García Iturralde (2026) examine a culturally situated account of AI literacy with 90 educational, youth and community actors in Esmeraldas, Ecuador, then explain the survey with a three-participant focus group of young Afro-Ecuadorians. Every dimension scored high on a five-point scale, with Afro-descendant identity the highest and critical AI literacy the lowest — and the instrument's identity subscale was its least reliable, so the ranking is softer than the means suggest. The focus group is where the equity argument turns concrete: participants described occasional, task-bounded use and image tools that altered their skin tone, hair texture or facial features — automated bias the survey recorded only as agreement that AI "can reproduce stereotypes."
Key Findings
- Afro-descendant identity was the highest-rated dimension, with pride near the top of the scale on almost every response.
- Critical AI literacy was the lowest-rated dimension, and AI's support for learning the weakest single item.
- Reliability was uneven: youth leadership measured consistently, Afro-descendant identity poorly, despite strong overall reliability.
- All dimensions moved together — every pairing was positively associated, strongest between ethnoeducation and educommunication.
- Image tools altered participants' racialized features. One participant said a tool "changed my skin tone and did not recognize my features" and called it "a form of racism, even if it is digital."
- Reported use was narrow, unlike the survey consensus. The three young participants used AI for logos, notes and study; one said he does not use it much day to day.
Six dimensions, one instrument
The questionnaire was built from the conceptual framework and the research objective, grouping items into six dimensions: Afro-descendant identity; self-esteem and cultural dignity; ethnoeducation; Afro-descendant youth leadership; educommunication; and critical AI literacy. Indicators fixed the wording: identity pride, elders' memory, and territory for identity; discrimination tied to hair, skin color, or ways of speaking for cultural dignity; orality, music, spirituality, and community participation for ethnoeducation; voice, listening, and message production for educommunication; and source and bias review for critical AI literacy, which the authors place inside media and information literacy. Items were five-point Likert statements (1 = lowest agreement, 5 = highest) plus open-ended questions, 30 scale items in total. Missing responses were not imputed; the pride statement, addressed to Afro-descendant respondents, carries a valid n of 68.
What the focus group added
The qualitative phase was built to interrogate the survey, not to illustrate it. One virtual focus group met on May 20, 2026 with three young Afro-Ecuadorian participants from Esmeraldas, selected purposively — YP1, a young man, and YP2 and YP3, young women, all of university age. Manual thematic analysis produced five themes: instrumental uses of AI; AI as support but not a substitute for thinking; racialized representation and symbolic harm; identity reaffirmation; and ethical, community-based training. The mixed-methods joint display marks two of its five rows as divergent. Where the questionnaire recorded near-unanimous endorsement, participants described practice that was occasional and sometimes abandoned; identity appeared not as neutral affirmation but as assertion made defensively, after tools had failed them.
Reading the reliability pattern
Internal consistency split the instrument in two. Youth leadership (α = .836), educommunication (α = .819), and ethnoeducation (α = .785) held together. Critical AI literacy sat approximately at the conventional .70 reference point (α = .701). Self-esteem and cultural dignity fell below it (α = .645), and Afro-descendant identity was low (α = .471) on 66 complete cases. The overall α = .911, computed on 62 complete cases out of 90, does not override those subscales. The authors attribute the identity result to heterogeneous item content and to the conditional wording and reduced denominator of the pride statement, and they read alpha as score consistency rather than construct validity. No expert review or factor analysis was run, so it is an measurement starting point, not a validated scale.
What the correlations do and do not show
Because responses clustered at the ceiling, the authors used Spearman's rank-order coefficient rather than Pearson's. All 15 pairwise associations were positive and significant at p < .001, ranging from ρ = .433 to ρ = .680. The strongest joined ethnoeducation to educommunication (ρ = .680), youth leadership to educommunication (ρ = .674), and ethnoeducation to youth leadership (ρ = .669); critical AI literacy was associated with educommunication (ρ = .655). Every coefficient describes co-variation inside a purposive sample of 90, not mediation or prediction. Given the narrow 4.60-to-4.72 band, the authors treat the pattern as a disposition to endorse normatively desirable statements — a self-report consensus that may partly produce the correlations themselves.
What this means for practice
- Instructors. Teach AI as a question before a tool: have students interrogate who produced a source, on what data, from whose perspective, and keep authorship with them.
- Program designers. Sequence training as the focus-group participants asked — identity and cultural belonging first, then basic concepts, privacy, bias review, and responsible use — and build it as culturally relevant pedagogy.
- Administrators. Fund educator preparation in ethnoeducation and critical AI use as one professional-development line, not two, because teachers are asked to hold both at once.
- Researchers. Treat the six-dimension instrument as provisional: revalidate the identity and self-esteem subscales, add items that spread responses, and replicate before comparing groups.
Limitations
- The sample is purposive and non-probabilistic (n = 90), with educators at 62.2%; several subgroups were too small for inferential comparison.
- Location is concentrated in one province: 87.8% in Esmeraldas city and 6.7% in San Lorenzo, a combined 94.4%, with 5.6% in Santo Domingo.
- A single focus group of three young people documents experience but cannot establish prevalence.
- Reliability is weakest where the identity argument is strongest: α = .471 for Afro-descendant identity and α = .645 for self-esteem and cultural dignity.
- All measures are self-report; dimension means span only 4.60 to 4.72 with 88.9% to 95.6% of scores at or above 4; and the design is cross-sectional.
Citation
Medranda Morales, N. J., & García Iturralde, R. G. (2026). Ethnoeducation, Afro-descendant youth leadership, and artificial intelligence literacy: an exploratory study with educational actors from Esmeraldas, Ecuador. Frontiers in Education, 11, 1910214.