Ecuador’s Universities Embrace Generative AI While Leaving Students Without Rules

Generative artificial intelligence has swept into university classrooms across Latin America faster than almost any technology before it, and a new study from Ecuador offers one of the most detailed portraits yet of what that means for a developing country’s higher education system. Researchers from the Universidad Técnica Particular de Loja and the Pontificia Universidad Católica del Ecuador examined how social science lecturers and students perceive, use and worry about AI tools, and what their universities are actually doing to govern them. The findings, published in Frontiers in Education, reveal a striking paradox: students and teachers have adopted AI enthusiastically, yet almost no institution has given them clear rules for doing so.

The research team, led by Abel Suing, combined three methods in a way that is rare in this field. They conducted semi-structured interviews with 14 lecturers from Ecuadorian, Spanish, Chilean and Peruvian universities, plus 8 students across five degree programmes. They surveyed 200 social science students, 58.2 percent of them women with an average age of 24, using a questionnaire built around technology acceptance theory and UNESCO’s AI competency framework. Finally, they performed a content analysis of seven sets of AI regulations from Ecuadorian universities, checking each document against the competency dimensions identified by UNESCO-IESALC in 2025. This triangulation allowed the researchers to compare what people say with what institutions have actually written down.

The survey results paint a picture of pragmatic adoption. Students reported a moderate-to-high level of AI use, with a mean score of 3.68 on a five-point scale, deploying the tools to draft assignments, summarise information, understand complex topics and save time. Crucially, the statistical analysis showed that this use was driven by perceived usefulness and extrinsic motivation rather than by ease of use. When the researchers controlled for how easy students found the tools, the correlations between use and usefulness (a partial Spearman coefficient of 0.516) and between use and motivation (0.546) remained strong and significant. In other words, Ecuadorian students are not using AI because it is simple; they are using it because it works for them.

Perhaps the most revealing finding concerns ethics. Students expressed moderate-to-high concern about plagiarism, dependency, unreliable information and the effect of AI on their learning, yet these concerns showed no statistical relationship with how much they actually used the technology. The correlation between use and ethical concern was a negligible 0.072, statistically indistinguishable from zero. Students, it seems, operate in a double bind: they recognise the risks and keep using the tools anyway. Recognising danger, the authors conclude, is not by itself a mechanism of self-regulation, which is precisely why institutional frameworks matter.

That institutional framework is largely missing. Students rated their demand for AI regulation at 4.00 out of 5, close to full agreement, but rated the clarity of regulation at their own universities at just 3.21, a statistically significant gap of 0.80 points. The documentary evidence confirmed the perception: of the seven regulatory instruments the researchers found across Ecuadorian universities, only one, at Universidad Internacional SEK, is formally designated a policy. The rest are guidelines, protocols or work guides with varying approval levels, and the design of discipline-specific AI systems for students went entirely unmentioned in every document.

The content analysis also showed what Ecuadorian universities prioritise when they do regulate. Ethics in the use of AI occupied the first position for both lecturers and students across the seven instruments, followed by conceptual understanding of the technology and permitted applications. At the opposite end, professional development for teachers and AI for research received almost no attention, and the design of discipline-specific AI systems scored zero for students. The authors characterise this as a regulatory phase rather than a competency-development phase: institutions are writing rules about control and ethics while neglecting the skills that would let their communities use AI well.

The interviews add human texture to these numbers. Lecturers described using AI to create teaching materials, plan classes, gamify introductory lessons, manage databases and complete literature reviews in hours rather than weeks. One political scientist noted that AI can propose a research question and then sweep validated platforms for reliable sources, producing summaries of the relevant works. Yet the same lecturers admitted they feel unprepared, with training courses that rarely reach the level of applying AI to a specific subject. Older staff were described as particularly reluctant, and the researchers heard of a ‘moment of disorientation’ between optimistic early adopters and technophobic colleagues who discourage students from using AI at all.

Material constraints loom large in a country where, the lecturers reported, neither internet nor electricity reaches every location. Campus connectivity can be so poor that instructors revert to traditional lesson planning, and at some public universities students pool money to pay for software licences because they lack credit cards or institutional access. Private universities tend to move faster in enabling digital environments, while public institutions face bureaucratic obstacles and resource limits. These gaps in connectivity, licensing and training, the authors argue, are not merely logistical inconveniences but matters of educational justice that must be built into any AI policy.

Students told a parallel story of self-taught adoption. Many described using AI daily, asking it to find research sources, plan lessons, generate flashcards in NotebookLM or practise languages without tiring. Some benefited from institutional licences such as Scopus AI; others hit paywalls after exhausting free limits. Training was minimal to nonexistent, with several students reporting they had never had a class on AI beyond superficial naming of tools. They also voiced sophisticated worries: difficulty telling whether the model is hallucinating, fear of dependency, and the observation that lecturers themselves can no longer distinguish original work from AI-assisted production. One future teacher admitted she would not be able to tell if a student were telling her the truth.

The deepest anxiety, shared by lecturers and students alike, concerns critical thinking, the epistemological core of the social sciences. Interviewees warned that intensive AI use can produce ‘inert thinking’, superficial learning and copy-paste habits that erode argumentative skills, and that the social sciences without critical attitude become merely ‘canned knowledge’. Yet the same participants showed how AI can strengthen critical thinking when used deliberately: comparing perspectives, analysing bias in generated responses, generating classroom debates, or helping students with reading difficulties convert text to audio and organise their ideas. The distinction that emerges, consistent with recent systematic reviews, is that AI’s effect on thinking depends on the type of use rather than exposure alone.

The study’s conclusions carry weight beyond Ecuador. The researchers argue that governance of AI in higher education requires an evaluative redesign of assessment, incorporating oral defences, prompt traceability and analysis of the questions students pose to AI, together with competencies differentiated by academic role and by discipline. As long as accreditation bodies merely check that a policy document exists, they warn, universities will keep producing instruments of symbolic visibility rather than investing in genuine training. Ecuador’s national AI strategy, published in January 2026, and accreditation indicators requiring evidence of appropriate AI use provide external pressure, but the translation into university practice remains embryonic. For a generation of students already using these tools every day, the message of the research is clear: the demand for rules is coming from the users themselves, and the institutions are running behind.

Subject of Research: Perceptions, practices and governance of generative AI in university social science teaching in Ecuador

Article Title: Perceptions, practices and governance of generative AI in university social science teaching in Ecuador

Article References: Suing, A., Marcelino Romero Gutiérrez, J., Galarza-Ligña, V. N., & Coronado-Otavalo, X. (2026). Perceptions, practices and governance of generative AI in university social science teaching in Ecuador. Frontiers in Education, 11, Article 1948805. https://doi.org/10.3389/feduc.2026.1948805

Image Credits: AI Generated

DOI: 10.3389/feduc.2026.1948805

Keywords: generative AI, higher education, Ecuador, social sciences, AI governance, AI ethics, critical thinking, UNESCO-IESALC, technology acceptance, university policy, digital divide, AI literacy