The study, led by Julio Cabero-Almenara, Julio Barroso-Osuna, and Antonio Palacios-Rodríguez of the University of Seville, together with María Isabel Loaiza Aguirre of the Private Technical University of Loja in Ecuador, took a deliberately granular approach. Rather than treating attitudes toward artificial intelligence as a single, undifferentiated feeling, the researchers drew on the classic ABC model of attitudes, which decomposes any attitude into cognitive, affective, and behavioral components. The cognitive component captures beliefs and conceptions about the technology, the affective component captures emotions of acceptance or rejection, and the behavioral component captures intentions and dispositions to act. To this triad the team added a fourth, less commonly measured dimension: the ethical component, covering students’ sensitivity to issues such as algorithmic bias, privacy, and the responsible handling of personal data.

The sample comprised 570 students enrolled in master’s programs related to educational technology at universities in Ecuador, Spain, and the Dominican Republic. Ecuador contributed 188 participants, or 33 percent of the sample; Spain contributed 110, or 19.3 percent; and the Dominican Republic contributed 272, or 47.7 percent. Women made up nearly two-thirds of the participants, and more than half were under thirty years old. Importantly, 44 percent of the students were also working as teachers at some level of the education system, a detail that would prove statistically decisive. The participants rated their own technical mastery of artificial intelligence and their knowledge of its didactic use on a scale from one to ten, and their self-assessments clustered in the moderate range, with slightly higher confidence in pedagogical application than in technical operation.

The measurement instrument was an adaptation of the SATAI, the Student Attitude Toward Artificial Intelligence scale originally developed by Suh and Ahn and later adapted for university contexts, expanded with nine newly constructed items probing ethical attitudes. The final questionnaire contained 32 items across the four components, each rated on a seven-point Likert scale. Reliability was exceptional: the total instrument achieved a Cronbach’s alpha of 0.96, a figure well above the 0.90 threshold conventionally regarded as indicating excellent internal consistency. The researchers also computed McDonald’s omega, and the dimension-level values confirmed that the instrument measured its four components coherently. Because the data were not normally distributed, the team relied on non-parametric statistics, applying the Mann-Whitney U test for two-group comparisons, the Kruskal-Wallis test for comparisons across three or more groups, and Spearman rank correlations to map the relationships among components.

The headline finding is that attitudes toward artificial intelligence among these students are broadly positive. On the seven-point scale, average scores ran roughly two points above the scale’s midpoint. The cognitive and behavioral components scored highest, indicating strong beliefs in the value of the technology and a firm willingness to engage with it, followed closely by the affective-emotional component, which signals genuine emotional acceptance rather than grudging tolerance. Behavioral intention, the students’ stated readiness to incorporate artificial intelligence into their academic and professional lives, was similarly robust. Yet the ethical component stood out for the wrong reasons: it received the lowest scores of any dimension. The authors interpret this as evidence of limited knowledge and training regarding the ethical challenges of artificial intelligence, including biases embedded in training models that can shape the responses users receive, and the privacy implications of how personal data are collected and processed.

The correlational analysis added technical depth to this picture. All four components correlated positively and significantly with one another, but the strength of those relationships varied in revealing ways. The strongest association, at a Spearman coefficient of .774, linked the cognitive and behavioral components, suggesting that what students believe about artificial intelligence strongly shapes what they intend to do with it. The affective-emotional component correlated at .723 with behavioral intention, confirming that emotional comfort with the technology is a key driver of adoption. Most striking was the ethical component, which showed the weakest correlations in the model, reaching only .424 with behavioral intention. In other words, ethical awareness appears to be only loosely coupled to the decisions students make about using artificial intelligence, a disconnection that could allow enthusiastic adoption to proceed without ethical guardrails. Notably, in the Spanish subsample the ethical component’s relationships with other dimensions weakened further, in some cases to statistical insignificance.

The group comparisons told an equally nuanced story. Gender produced no significant differences in any component or in overall attitudes, a result consistent with a substantial body of international research but at odds with studies reporting male advantage. Country of study, by contrast, mattered: Ecuadorian students scored highest on most components, Dominican students led specifically on the ethical dimension, and Spanish students posted the lowest averages across the board, which the authors suggest may reflect a more moderate or critical stance toward artificial intelligence in education. Field of study also shaped attitudes significantly. Students from Arts and Humanities and from Engineering and Architecture recorded the highest overall scores, science students excelled on the cognitive component, and social science and law students led on intention to use. Health sciences students ranked lowest on every dimension analyzed, a gap the authors flag as urgent given the coming transformation of AI-enhanced healthcare environments.

Two personal variables proved especially powerful. Students who were themselves practicing teachers held significantly more favorable attitudes toward artificial intelligence than their non-teaching peers, with mean differences of roughly fifty points across all components, suggesting that hands-on professional experience makes the technology feel useful rather than threatening. Even more dramatic was the effect of self-assessed knowledge: students who rated their technical and didactic mastery of artificial intelligence at eight, nine, or ten out of ten scored more than one hundred points higher than those at the lower levels, across every component and in their intention to use the technology. The authors caution, however, that artificial intelligence often functions as a black box whose internal workings are opaque even to frequent users, so self-reported confidence may not always track genuine understanding, particularly of how the information these systems generate is actually produced.

The study arrives amid a broader regional reckoning. Recent reviews describe artificial intelligence as a disruptive agent in the digital transformation of Latin American universities, driving new forms of teaching, evaluation, and academic management, while flagging persistent obstacles: the digital divide, inadequate infrastructure, insufficient teacher training, and a shortage of long-term research. A World Bank report cited in the study found that while 65 percent of teachers consider artificial intelligence an opportunity, more than 80 percent say their institutions lack clear guidelines for its use. Flagship institutions such as the National Autonomous University of Mexico, the Monterrey Institute of Technology, and the Pontifical Catholic University of Chile have launched pilot programs, and UNESCO and the Development Bank of Latin America and the Caribbean have convened ministerial summits on AI ethics in the region.

The authors are candid about the limitations of their work. The sampling was one of convenience rather than probability, the instrument was a self-report questionnaire vulnerable to inflated self-perception, and the sample was drawn exclusively from educational technology programs, a population likely more tech-friendly than average. They frame the study as exploratory and call for replication with probabilistic samples, more diverse disciplines, and performance-based instruments that ask students to solve real cases rather than rate their own confidence. Still, the practical prescription is clear: universities should integrate digital ethics literacy programs and adopt clear institutional guidelines so that the region’s evident enthusiasm for artificial intelligence is matched by the critical judgment needed to use it responsibly. The students, the data suggest, are ready; the ethical framework must now catch up.

Subject of Research: Multidimensional attitudes of Latin American university students toward artificial intelligence in higher education

Article Title: Artificial Intelligence in higher education: cognitive, emotional, behavioral and ethical attitudes of Latin American students

Article References: Cabero-Almenara, J., Barroso-Osuna, J., Loaiza Aguirre, M. I., & Palacios-Rodríguez, A. (2026). Artificial Intelligence in higher education: cognitive, emotional, behavioral and ethical attitudes of Latin American students. Journal of New Approaches in Educational Research, 15(1), Article 13. https://doi.org/10.1007/s44322-026-00063-2

Image Credits: AI Generated

DOI: 10.1007/s44322-026-00063-2

Keywords: artificial intelligence, higher education, Latin America, student attitudes, digital ethics, SATAI scale, educational technology, behavioral intention, AI literacy, university students, ethics training, survey research