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

1 Introduction

The advent of generative artificial intelligence (AI) in education is altering teaching and learning at every level. Efforts are consequently under way to increase students' and teachers' willingness to use it appropriately, although the risk of academic fraud is also acknowledged, and warnings have been raised about a mechanisation of work that obstructs critical thinking. In response to these risks, and specifically for higher education, the UNESCO International Institute for Higher Education in Latin America and the Caribbean identified, in 2025, the challenges and barriers facing AI competency frameworks in higher education, for both students and lecturers (Valentini and Blancas, 2025). Its aim is to guide the responsible and creative use of AI, albeit in an environment in which most institutions have yet to define roadmaps for action.

Among the fields of knowledge, the social sciences (CCSS) are essential for identifying how AI is being integrated, given the relevance of people's relationships and behaviour in the face of this new technology, particularly in countries such as Ecuador, where local dynamics are conditioned by underdevelopment, weak governability and high crime rates. World Bank governance indicators (2006) attest to the institutional crisis, and the Ecuadorian State's weak capacities in the face of systemic corruption are also well documented (Transparency International, 2025). Against this backdrop, universities appear to have responded only minimally to the emerging phenomenon of AI. In 2025, Ecuador had 45 specialised AI programmes; however, “the inclusion of technical aspects in university programmes in the social sciences and humanities is an incipient process, with few permanent initiatives identified to date” (Albornoz, 2025, p. 52).

A further consideration in this Andean country is that it was only in January 2026 that the Ministry of Telecommunications and the Information Society issue the “Strategy for Fostering the Development and Ethical and Responsible Use of Artificial Intelligence in Ecuador” (Registro Oficial, 2026), which identifies as a priority line of action the need to “integrate AI content into educational processes from basic through to higher education”. This priority is consistent with UNESCO's aims and reaffirms the need to expand knowledge of AI for everyday life through education. Accreditation requirements point in the same direction: in Ecuador, indicators 10 and 11 of the “external evaluation model for accreditation purposes to ensure the quality of universities and polytechnic schools” require evidence of the appropriate use of AI mechanisms or tools (CACES, 2023).

Within this context, it is pertinent to examine how AI is being received in university-level social science teaching in Ecuador, through the voices of those at the centre of this transition: students and their lecturers. The experiences of this Andean country serve as a point of reference for other nations with similar economic, political and cultural contexts. A new qualitative and quantitative study will help to contrast the findings of previous analyses regarding predispositions towards incorporating AI into higher education in this and other countries. Previous studies have concluded, for instance, that university students in Ecuador, Peru, Mexico and the Dominican Republic display positive attitudes towards AI in education, recognising its potential to personalise learning and improve the quality of education (Cabero-Almenara et al., 2026; Ríos et al., 2024).

Recent research on AI in higher education, both regionally and internationally, tends to adopt approaches centred on technological adoption and operational efficiency (TAM/UTAUT models) within STEM fields, focusing on uses (how, for what purpose) rather than on the implications for knowledge and thought (why, for whom), thereby isolating the ethical and pedagogical dimensions and media literacy. Moreover, such studies overwhelmingly focus on contexts within the so-called Global North, characterised by robust telecommunications infrastructure and evolving regulatory frameworks. A gap therefore exists in our understanding of what occurs in universities where structural gaps in access, teacher training and entrenched pedagogical traditions persist, a gap that is particularly critical within the social sciences, disciplines whose epistemological core lies in critical interpretation and the analysis of complex phenomena.

This study addresses this gap and explores how teachers and students of social sciences in Ecuador perceive and manage the challenges posed by AI. The research questions are: How do university lecturers and students in the social sciences in Ecuador perceive and use AI? What factors motivate and influence the use of AI in the teaching of social sciences at universities in Ecuador? Are there policies regarding the adoption of and literacy in AI within university educational practices in Ecuador? The objectives are: 1) to determine the perceptions of lecturers and students regarding the integration of AI into the teaching of social sciences at universities in Ecuador in relation to international competence frameworks; 2) to identify the factors that motivate and influence the use of AI in the university teaching of social sciences in Ecuador; 3) to identify university policies on generative AI at universities in Ecuador, comparing them with international competence frameworks.

2 Theoretical framework

The use of AI at university is explained through technology-focused approaches such as the Technology Acceptance Model (TAM), the Unified Theory of Acceptance and Use of Technology (UTAUT) and Cognitive Evaluation Theory, which measure perceived usefulness, performance expectancy and perceived ease of use, and which together form a cognition-attitude-behaviour causal chain (Liu et al., 2024; Strzelecki, 2024). What these models leave unresolved, however, is the implications, relevance and effective deployment of AI in the development of competencies within the teaching and learning process.

The need to move beyond quantitative perception indicators is particularly evident within the social sciences, since AI has implications for the humanities and their critical perspective. Unlike in STEM fields, a fundamental question arises: what does it mean to learn social sciences when an algorithm could construct a state of the art without any human perception or opinion? AI could thus reconfigure what is understood as knowledge and confront higher education with the challenge of moving towards a culture of procedural transparency, in which originality comes to mean the capacity to question and engage in critical dialogue (Valentini and Blancas, 2025).

Yet these changes are unfolding in a setting where many teachers lack sufficient knowledge to integrate AI effectively into their pedagogical practices (Alé et al., 2025; Sartor-Harada and Ulloa-Guerra, 2026), and who are concerned about “the possible substitution of traditional roles […] and the lack of formal training in the use of these tools” (González and González, 2024, p. 79), as well as about “ethical use, professional displacement and the erosion of academic autonomy” (Buele and Llerena-Aguirre, 2025, p. 8). Nevertheless, these teachers recognise the importance of training to use AI and of addressing issues such as privacy, algorithmic bias and the authenticity of student work (Miranda et al., 2024).

Meanwhile, students “value the importance and relevance of AI tools” (Espinoza-Bravo et al., 2025, p. 15), and perceive a need for training in digital ethics and clear institutional frameworks for their responsible use (Saltos et al., 2025). Specifically, university students on social science degree programmes demonstrate basic knowledge of AI tools, but their advanced technical competencies remain scarce (Duque-Rengel and Puertas-Hidalgo, 2026; Grájeda et al., 2024). Both lecturers and students alike highlight the urgent need for institutional policies and training to integrate AI appropriately and to ensure equitable access to it (Ramírez-Montoya et al., 2024), which “demonstrates the urgency of updating study programmes and developing continuous training strategies, for both lecturers and students” (Cepeda et al., 2025, p. 163).

This global phenomenon has its own particularities in the Andean countries of Ecuador, Colombia and Peru, which share a common culture and history. AI is perceived as a driver of educational innovation, particularly with regard to personalised learning and the optimisation of teaching (Cabero-Almenara et al., 2025; Rodríguez et al., 2025), albeit with uneven progress in its integration into higher education. In Ecuador, “the adoption of AI within the social sciences is neither automatic nor homogeneous, but is instead conditioned by cultural, institutional and epistemological factors” (Loor et al., 2025, p. 6), and, as in other regions, “the majority of lecturers agree on the need to generate institutional regulations for the use of AI” (Mejía and Gómez, 2024).

Ecuadorian university students display a complex relationship with AI, “characterised by moderate dependency and ambivalent attitudes” (Arce et al., 2025, p. 10); they use it to improve their writing, albeit with a fear of becoming dependent on it (Vilma et al., 2025), and “although the use of AI is widespread among Ecuadorian university students, its direct impact on academic performance remains inconclusive” (Alcívar-Junco, 2026, p. 8). In general terms, there is a “coexistence of high technological acceptance with limited institutional structuring for its pedagogical use” (Asencio et al., 2026).

To clarify the use and integration of AI in higher education in Ecuador, a review of recent studies was conducted. As of May 2026, 13 scientific publications were identified that share similarities in their analysis of perceptions or uses of AI amongst lecturers and students (Table 1). Most of these studies sought to identify levels of knowledge, perceived benefits, barriers to adoption and ethical challenges, with the exception of Espinoza-Bravo et al. (2025) and Sumba and Segarra (2026), who added theoretical dimensions based on models of technology acceptance (AIDUA) or digital competences. In addition, Mora Torres and Plaza Alarcon (2026) expands on this by integrating official secondary data (INEC's ENEMDU-TIC bulletins, UNESCO's RAM Report, and the Digital Transformation Policy 2025–2030) with evidence published between 2024 and 2025.

Table 1. Studies on the incidence of generative AI in Ecuadorian higher education.

Source: authors’ own elaboration.

Jadán and Arostegui (2023) link AI with educommunication. Mora Torres and Plaza Alarcon (2026) presents a macro and comparative perspective, distinct from the predominantly perceptual approach. The predominant methodological approach is quantitative or mixed methods. Ten out of 13 studies use surveys as their main instrument, with the exception of Buele and Llerena-Aguirre (2025), who conduct a narrative review, and Mora Torres and Plaza Alarcon (2026), who analyses secondary sources and official data. The sample sizes range from 49 to 3,500 participants, and are generally non-probabilistic or convenience samples. Although interviews are used in four cases, they play a secondary role (Arias et al., 2024; Cepeda et al., 2025; Hernández, 2025).

The findings converge on the following points: 1) high adoption of, and familiarity with, AI, with lecturers and students alike being familiar with and using AI (especially ChatGPT); 2) a positive perception, albeit accompanied by ethical concerns regarding academic integrity and the displacement of the teaching role; 3) deficits in training and capacity-building; 4) insufficient infrastructure, particularly within public and grant-aided institutions (Hernández, 2025; Jara, 2024); and 5) a gap between perception and practice (Yllescas et al., 2025). The conclusions likewise converge on the need for institutional policies and regulation, continuous teacher training and ethical openness. None of these studies, however, addresses the social sciences and humanities specifically. It should be stressed that the implications of AI use in communication, philosophy, social work or education are qualitatively distinct from those in engineering or medicine. Nor do they examine institutional policies or comparative regulation across Ecuadorian universities.

The foregoing underpins the identified need for a comprehensive educational approach encompassing the various dimensions of AI, in order to prepare students for the challenges and opportunities presented by this emerging field (Verduga-Alcívar et al., 2026); such an approach entails a form of media literacy that is embodied in the AI competency framework for higher education identified by the UNESCO International Institute for Higher Education in Latin America and the Caribbean (UNESCO-IESALC, 2025) and in the conclusions of Zhao et al. (2026), which call for a “human-centred mindset” and ethical understanding (Miao, 2024). “One of the most pressing challenges is to ensure that AI literacy extends beyond STEM disciplines […] to all fields, including the arts and humanities” (Valentini and Blancas, 2025, p. 20).

3 Methodology

The methodology is mixed, combining qualitative and quantitative approaches of a descriptive and explanatory scope (Hernández et al., 2018). The instruments employed were semi-structured interviews, surveys and content analysis. The study was approved by the Research Ethics Committee for Human Subjects of the Universidad Técnica Particular de Loja, Ecuador (ref. 2026-04-INT-EO-RM-02); participants' informed consent was recorded, and protocols for the anonymous handling of data were duly observed.

The interview questionnaires for lecturers and students were designed on the basis of instruments used in recent research and validated through expert peer review. The interviews sought to determine the factors that motivate and constrain the use of AI in university-level social science teaching in Ecuador. The resulting testimonies were subsequently examined within the AI competency framework identified in “The Challenges of AI in Higher Education and Institutional Responses: Is there room for competency frameworks?” (UNESCO-IESALC, 2025). The sources underpinning the design of the lecturer interviews are drawn from the research of Baytas and Ruediger (2025), Pérez-Montesdeoca et al. (2026), Shi (2025) and Ututallum (2025). The student interview questionnaire was constructed on the basis of the research of Summers et al. (2024) and Zastudil et al. (2023).

Table 2 presents the profiles of the lecturers interviewed. These comprised 14 individuals: 10 men and 4 women, with a mean age of 41 years. The interviews were conducted between 11 April 2026 and 18 May 2026. The Ecuadorian institutions represented are: Universidad Técnica Particular de Loja, Universidad Técnica de Machala, Universidad Estatal Península de Santa Elena, Universidad Nacional de Loja, Universidad Politécnica Estatal del Carchi, Pontificia Universidad Católica del Ecuador, Universidad Técnica de Cotopaxi and Universidad San Francisco de Quito. International institutions represented are Universidad de Málaga, Universidad de Extremadura, Universidad Católica del Norte (Chile) and Universidad de Lima (Peru). The analysis was conducted in order to identify emerging categories and recurring patterns within the testimonies.

Table 2. Profiles of the lecturers interviewed.

Source: authors' own elaboration.

Table 3 shows the profiles of the students interviewed. It covers 8 participants: 4 women and 4 men. Average age: 25 years. Period of the study: between 18 May and 12 June 2026. The degree programmes represented are Communication, Language Teaching, Law, Primary Education and a Master's in Market Research. The interviewees' academic levels range from the fourth to the seventh semester of their undergraduate studies, with one postgraduate student. The participants are from the Universidad Técnica Particular de Loja (UTPL), the Universidad Técnica del Norte (UTN), the Pontificia Universidad Católica del Ecuador, Ibarra Campus (PUCESI), the Pontificia Universidad Católica del Ecuador, Santo Domingo Campus (PUCESE) and the Universidad Internacional de La Rioja (UNIR). Two interviewees were undertaking exchange semesters abroad at the time of the interviews: at Alberto Hurtado University in Chile and the Catholic University of Córdoba in Argentina.

Table 3. Profiles of the students interviewed.

Source: authors' own elaboration.

The second instrument was a non-probabilistic survey directed at social science students at Ecuadorian universities. The questionnaire was developed based on the research of Jiménez-Perona et al. (2025), who assess the impact, pedagogical implications and attitudes towards generative AI within university educational processes, and of Yılmaz et al. (2024), in order to gauge actual intention to use. It was also aligned with the dimensions of the AI competency frameworks identified by UNESCO-IESALC (2025). The questionnaire items were presented on a Likert scale, ranging from 1 (disagree) to 5 (strongly agree). Prior to distribution, a pilot test was conducted and the instrument was validated through expert peer judgement.

The survey was conducted digitally, using Google Forms in a snowball sampling approach. Institutional digital channels, academic networks and student messaging groups were utilised. The first screen of the questionnaire included the informed consent form, guaranteeing the anonymity of responses, voluntary participation, and the strictly academic use of the data, in accordance with ethical standards. Statistical analysis was carried out using SPSS v.25 software. The sample comprised 116 women (58.2 %) and 84 men (41.8 %), with an average age of 24 years (SD = 6.38; median = 22; range: 18–48). The majority of respondents were undergraduates (88.5 %), compared with 12 % of postgraduates. The universities with the highest representation were PUCE-SD (44 cases), PUCE-SI (40), UNEMI (32), UPEC (29), and ECOTEC (11) = 78 % of the sample. The remainder comprises 14 additional institutions, though these are concentrated in universities in Quito and Guayaquil. The questionnaire allows the information to be organised into dimensions and constructs (Table 4), which are used to derive indices.

Table 4. Analytical dimensions.

Source: authors' own elaboration.

The analysis is inferential and correlational across dimensions; Spearman's Rho correlation coefficient was employed (owing to non-normality confirmed via the Shapiro–Wilk test across all variables), together with such other tests as were necessary to assess the relationship between dimensions/constructs. The hypotheses are as follows: The third research instrument, content analysis, served to examine the institutional policy documents identified through a search of three complementary sources, conducted between 15 and 30 June 2026: 1) a web search using the keywords “regulation on the use of generative AI in universities”, “protocol for the use of AI in Ecuadorian universities” and “regulatory framework for the use of AI in Ecuadorian universities”, with a cut-off date of 20 June 2026; 2) an exploration of the institutional websites of Ecuadorian universities in search of published regulations on the use of AI; and 3) a review of the institutional repositories available on the platform of Ecuador's Higher Education Council. The competencies applied in the analysis are those identified by the UNESCO International Institute for Higher Education in Latin America and the Caribbean (UNESCO-IESALC, 2025); five dimensions are proposed for teaching staff: 1) human-centred approach, 2) AI ethics, 3) foundations and applications, 4) AI pedagogy and 5) application to professional teacher development. For students, four dimensions are proposed: 1) human-centred approach, 2) AI ethics, 3) applications and techniques associated with its use, and 4) design of discipline-specific AI systems. Zhao et al. (2026) validate the UNESCO proposal through its application in higher education institutions.

• H1. There are significant differences among all the dimensions of the survey.
• H2. The gender variable shows that women display greater awareness across all dimensions of the survey.
• H3. Students use AI despite their ethical concerns.
• H4. Perceived regulatory clarity (is it clearly regulated?) is significantly lower than the perceived need for regulation (should it be regulated?).
• H5. Perceived usefulness and motivation positively predict AI use, controlling for ease of use.
• H6. Demand for regulation is positively associated with perceived risk and with level of use.

The combination of three instruments—semi-structured interviews, a non-probabilistic survey and content analysis—represents an important methodological contribution to this field. Of the thirteen Ecuadorian studies identified in Table 1, most rely on a single instrument, almost always a survey (Buele and Llerena-Aguirre, 2025). This limits the possibility of comparing what actors say with what the documentary evidence on institutional governance shows.

As far as the authors have been able to verify, this study is the first to apply the UNESCO-IESALC (2025) AI competency framework to the social sciences in Ecuador in an empirical and comparative manner. It stands out for distinguishing between the competencies expected of lecturers and those expected of students, and for being the first study in Ecuador to subject that framework to a systematic verification against seven real institutional instruments, rather than relying solely on respondents' perceptions of their existence.

The lecturers and students interviewed were selected through a convenience sampling approach aimed at ensuring diversity across institutions (public and private, national and international universities), fields of study and levels of education, rather than pursuing statistical representativeness. The final number of participants (14 lecturers and 8 students) was determined once no new themes emerged from the interviews. The institutional documents examined through content analysis were selected according to a criterion of thematic relevance: instruments with a general regulatory approach to the use of AI in teaching and learning processes were included, whereas those confined to specific guidelines for the preparation of academic articles were excluded.

The three sources of evidence — interviews, the survey and content analysis — were integrated analytically. The interview findings help to interpret and contextualise the statistical associations identified in the survey. In addition, the content analysis contributes to characterising the regulatory gap perceived by lecturers and students. It is important to note that the statements regarding the effects of AI on critical thinking are grounded in the self-reported perceptions of lecturers and students. The scientific contribution of this study lies in showing the gap between the institutional discourse on regulation and everyday teaching practice, where regulation is expressed as the development of specific competencies. This distinction is not captured by technology-acceptance models such as TAM or UTAUT, which focus on users' individual behaviour.

With regard to the categories applied in the analysis, Chee et al. (2025) propose a framework of AI literacy competencies differentiated by user profile and argue for distinguishing the competencies expected of lecturers from those expected of students. Chiu et al. (2025), for their part, validated a teacher AI competence self-efficacy scale (TAICS), confirming that training constitutes a measurable dimension, as is likewise identified here in the interviews. With respect to institutional governance, it was considered pertinent to examine AI policy on the basis of institutional regulations, as in Tsao (2025), in order to contrast (normative) discourse with the practice declared by the actors involved. Furthermore, the assessment of AI's effects on critical-thinking skills was undertaken on the basis of the conceptual framework proposed by Helal et al. (2025). Finally, the review by Apata et al. (2026) of AI's contributions to higher education in relation to Sustainable Development Goals 4 and 10 supports the inclusion of equity of access as an analytical category.

4 Results

4.1 Results of the lecturer interviews

The integration of AI within the social sciences corresponds to technological change, shifts in the labour market, and new international integration frameworks, as is the case with UNESCO, which “sets out protocols for countries [and raises] unique ethical and social challenges, such as issues relating to equity, transparency, privacy and, above all, accountability” (E10). Since the launch of ChatGPT at the end of 2022, its presence in university classrooms has become frequent, albeit not widespread, since “very few of us lecturers use AI and its tools in class” (E01). Nevertheless, the permanence of this innovation in education is accepted: “AI is here to stay. That is simply how it is, and we are seeing it across the most diverse fields, in trade, in services, in everything” (E05).

It is likewise a technological revolution that imposes unavoidable methodological transformations on “multi-screen” and “sensorially overstimulated” students (E14). This transition is taking place amid tension, since “the university has been heavily criticised for its inability to connect with people's everyday lives” (E08). While hundreds of young people use AI tools, universities are still attempting to “jump on the bandwagon and provide training solely as an instrument”, without addressing deeper political or regulatory dimensions (E06). To this must be added the ethical, social, transparency, privacy and accountability challenges involved (E10), together with the risk of training “extremely mediocre professionals” lacking critical awareness (E11).

The fundamental motivation for adopting AI lies in its capacity to systematise pedagogical tasks and optimise processes, such as the personalisation of needs through prompts (E01), the “creation of materials, planning, assessment work […] and even the creation of soundscapes, sound bases and phonographies” (E02); but equally in research, for “managing databases and conducting literature reviews, bibliometric studies, within a matter of hours”, tasks that previously required far more time (E04). The methodological process is streamlined by “proposing a research question, whereupon these intelligences begin to explore validated platforms containing reliable information and subsequently present us with summaries of the works” (E04), or by “carrying out an initial sweep of the content of your thesis […] which allows for sound data analysis and also for communicating the results” (E10). It is likewise used to “gamify classes at the more introductory levels” (E03).

Most lecturers at Ecuadorian universities are not “adequately prepared to use these tools [although] we take courses on AI use, we do not apply it to higher education teaching and, worse still, to a specific subject” (E07). This situation is not confined to Ecuador, since “most teacher training programmes do not include practical training on the use of AI with a pedagogical, ethical and disciplinary focus” (E13). “Often there is neither the training nor the time to explore this in depth. What is missing is integrating these tools from the planning stage onwards” (E12). To this must be added that “older lecturers are usually those with a greater fear of using technology” (E09). “There is a great deal of staff over 50 or 60 years of age who are today highly reluctant to use these tools” (E07), together with the inflexibility of “more vertical, less horizontal organisations and institutions” that are less open to change (E08).

Material gaps nevertheless persist. “There is a digital divide in this context, in terms of technological infrastructure and connectivity” (E13). It is not uncommon for connectivity problems to arise on campus: “The internet signal on campus was very poor, which forced a return to traditional lesson planning” (E03); more generally, across the country “neither the internet nor electricity reaches every location” (E07), not to mention the existence of economic barriers; for example, at some public universities there are students who “have to pool funds together and pay for” software “licences” (E02). Within this scenario, regulatory gaps emerge. One case reported was as follows: “A circular appeared saying that we can effectively do nothing, because it has not yet been regulated within our institutions” (E06), although “many countries have no regulations or laws on AI [but] we ask universities for regulations, and sometimes they do not even exist at the national level” (E09).

In light of this, only a small proportion of institutions are attempting to develop policies in this regard (E04), or to introduce oral defences of examinations, where “unfortunately 99% choose not to defend” (E03). Expectations also persist regarding the lack of an intercultural approach, since many tools are designed according to an “Anglo-Saxon, individualist logic, or one centred on dominant cultural models” (E12). On the other hand, AI can be a great ally for educational inclusion. “Students with reading and writing difficulties can use AI assistants to convert text into audio, organise their ideas before writing, and [teachers can] provide immediate feedback, tailored to each student's pace, thereby improving the experience of those with specific educational needs” (E12).

With AI, “it can definitely develop critical thinking, but not out of thin air, not on its own” (E10). Critical thinking can be encouraged through “comparing different perspectives, analysing bias in generated responses, or generating in-class debates” (E13), or through exercises aimed at “identifying challenges according to the author, seeking common ground, and seeking out shortcomings” (E05). In other cases, AI is used “with examples that are easy for students to understand, in order to grasp a theory that carries a complex argumentative structure behind it” (E07), to merge perspectives into complex narratives, something that “was previously unthinkable” without costly production (E02), or to generate storytelling and personalised case banks (E10). The overarching goal is to develop “autonomous, empathetic individuals, capable of understanding, researching and discerning” (E08).

An “intensive use of AI makes us weaker in terms of critical capacity, in terms of creativity” (E09), which can lead students to “copy and paste”, thereby undermining their argumentative skills (E01) — that is, superficial learning — and a “deprofessionalisation of academic activity” among lecturers (E04). This appears to represent a displacement of the capacity for reflection. It generates “inert thinking […] what is happening in higher education is that we have a human being… a zombie, one that clouds, and ultimately silences, the capacity for critical thought” (E01). Generative AI acts as an “extremely dangerous factor in producing extremely mediocre professionals” and in shaping an instrumental student “who thinks that university is a kind of school for adults, where what is most valued is the lecturer who is best able to synthesise information”, such that “the social sciences without a critical attitude, without critical thinking, are nothing; they are canned knowledge that is merely reproduced, merely repeated” (E11).

The indiscriminate use of these technologies may prevent students from “recognising their” individual “skills” (E5), such as oral expression, leadership, writing or logical thinking. This loss of protagonism is illustrated by the observation that, in classrooms, “we find young people reading from a teleprompter in order to deliver a narrative” (E14). In specific subjects such as Literature, “the enjoyment of reading a 200- or 300-page book, and savouring that reading experience, is being lost” (E03). Dangers also arise from the absence of an “ethical compass” in algorithmic design, given that the system may provide “prejudiced information, which may be discriminatory, and which may feed into and threaten human rights themselves” (E10).

AI is thus, gradually, becoming an element in the process of curricular integration. “Today we have moved from containing it to integrating it throughout the entire educational process” (E01), albeit on an experimental basis, while remaining aware that “the tools change all the time” (E14). Moreover, “it is one thing to consider that they should be implemented, and quite another to know how. Very few people are able to use AI tools applied to education […] the way classes are taught is highly exploratory” (E07). This scenario generates a “moment of disorientation” between “lecturers who use AI, are optimistic about its use, and encourage students to employ it” and technophobic lecturers who “not only do not use AI themselves, but also encourage students not to use it” (E09).

Private universities tend to move faster in enabling digital environments and usage policies, whereas the public sector faces bureaucratic obstacles (E06). In one case, “two years ago we introduced a policy for the use and experimentation of AI as part of our course syllabi […] we do not prohibit its use in the classroom, nor experimentation among our students” (E14). At another institution, “training courses for staff are constantly encouraged, particularly in managing the use of AI to create educational resources, to integrate AI into assessment, for example, and to conduct more personalised assessments” (E03). Practical advantages arising from private infrastructure are also highlighted, with reference to the use of virtual tools such as “the notebook, which does it quickly, in five minutes […] enabling interaction with students through video, audio and rather more contemporary slide presentations” (E01).

Within public-sector institutions, adoption tends to be limited by resource constraints. Institutional use tends to remain superficial, largely confined to “queries, preparing classes using AI”, which “is not, fundamentally, an adequately efficient use of AI” (E07). In general, the level of competency among teaching staff varies widely, ranging from those familiar with ChatGPT to others who feel overwhelmed by it (E12). Despite efforts to redesign curricula with an emphasis on AI (E02) or to offer ongoing training, many classrooms lack connectivity (E03).

4.2 Results of the student interviews

Students indicate that the adoption of AI occurs progressively, particularly in the final years of their degree programmes, driven by the need to optimise time in preparing theoretical frameworks for their theses, searching for literature, and creating didactic outlines for pre-professional practice. Communication and pedagogy degree programmes show earlier adoption, with a stronger ethical orientation, whereas in other disciplines, such as law or basic education, use tends to be more instrumental and exploratory in nature. As one participant put it: “I'm not good at finding research sources. Rather than going to a source myself, I ask AI to help me find sources based on a very specific topic” (EE2). AI “allows me to develop my ideas better, according to what I want to do, and it also lets me innovate, giving me more time to carry out other activities” (EE8).

For another student, the initial motivation was mediated by her lecturers, in order to organise ideas and create content. In another case, the motivation stemmed from a practical need, as pedagogical support. “I use it all the time, especially for creating and planning classes, asking for prompts on ChatGPT to act as an expert teacher, generating a well-structured lesson” (EE2). AI is used on a daily basis, having become a permanent resource, “all day, every day”, for both academic and personal activities, such as resolving queries or simplifying tasks. “I use it very frequently, especially now, for very simple things, like explaining a topic in simple terms or helping me summarise information. Availability is extensive: it is available 24 h a day, every day” (EE4). There are, however, more sporadic uses: “I use AI about 30% of the time” (EE5). “It is still very low; I haven't felt motivated to use it for anything beyond assignments” (EE1).

Among the positive factors, students mention access to institutional licences. “Within the platforms, I have seen great progress in trying to incorporate it. For example, we now have many applications included that already benefit us — that is, they are free for us as students, such as Scopus AI” (EE1). Within social science or communication degree programmes, its strategic use is encouraged so that students do not fall behind in the face of future labour market demands.

Some lecturers act as motivators for the use of AI. “They have strongly promoted the use of AI […] but ethically” (EE1). Certain lecturers introduce new platforms into classes and offer appropriate criteria for their use (EE7). AI is also used in tutoring, such as for practising languages, since it “never tires of repeating things, so that's something I've tried to make use of” (EE2). Along the same lines, AI is used to compare teaching methodologies and design didactic tools: “In NotebookLM I can enter certain information, and it allows me to create presentations and flashcards that help facilitate understanding of the subject” (EE6).

A few students indicate that they cannot access paid versions, meaning that free versions constitute a barrier. “Sometimes you exceed the limit that the AI allows — that's the first limit, since it stops being free and starts asking for payment” (EE7). “What limits me somewhat is the financial side and how to acquire and pay for subscriptions […] I tried to pay with a debit card, and I couldn't. It only asks for a credit card, which I don't currently have activated” (EE8). According to one interviewee, no institutional licences are available for students at her university (EE2).

Systematic training in AI is minimal. Students agree that they receive little training in the use of AI tools. “Honestly, I haven't seen any classes on AI or anything of the sort” (EE1). “I haven't received this kind of training regarding digital tools. I have received something, but very superficial, nothing in-depth at all” (EE4). “We haven't had training specifically aimed at specialising in this. Yes, in terms of getting to know these tools, but not in training us on how to use them correctly” (EE6).

Students acknowledge receiving basic guidance on prompting within an ICT subject, but note that this instruction was insufficient: “I feel it would be somewhat better to try to explore in greater depth what these tools actually are, rather than merely naming them” (EE3). At the same time, a restrictive attitude is perceived among some lecturers: “Lecturers often used to tell us that artificial intelligence wasn't very good, that we shouldn't rely on it too much, and that it was better for us to do the work ourselves by searching for bibliographic sources in the library” (EE4). Attitudinal limitations also arise, specifically distrust regarding the veracity of the information provided. “My main difficulty is not being able to tell whether AI is hallucinating or not, because I can easily believe what it tells me, but there's no evidence to show me where that information came from” (EE5). “Another limitation is that we become more dependent on AI, as though we expect it to solve everything for us” (EE4).

AI generates incorrect or incomplete information. “Not being able to tell whether AI is hallucinating or not, because I can easily believe what it tells me” (EE5). “We cannot trust artificial intelligence 100% […] sometimes the information also tends to be erroneous, tends to be incomplete” (EE4). There is a fear of falling into inaction and submitting copied work. EE2 acknowledges that copying and pasting generated text is a common practice among classmates and that “it shows quite clearly” in presentations and written work. “Sometimes I used to think AI was something negative, because the only motivation for using it seemed to be to copy or draft something and submit it, or as if it made your brain lazy” (EE1). “We become more dependent on artificial intelligence, as though we expect it to solve everything for us” (EE4). “The information it gives can be given to several other classmates as well — in other words, it's not authentic work, but rather it starts to produce repetitive information. And that then falls into cases where they say the work has been duplicated” (EE7).

There is evidence of teaching staff's inability to distinguish original production from AI-assisted production, together with a lack of balance between AI and independent thought. “At this point we can no longer tell what is genuine and what isnx't. It would be difficult for everyone, even for me as a future lecturer — I wouldn't be able to tell if my student were telling me the truth” (EE5). “Sometimes we lack that balance between using technology, using these intelligences, and developing our own ideas. So, we end up expecting artificial intelligence to solve all our work for us” (EE4).

The environments in which AI is taught in Ecuador vary considerably. Some universities are actively integrating AI through platforms and courses, whereas in others adoption is slower or more informal. EE1 observes that, at her university, “within the platforms, I have seen great progress in trying to incorporate it”. EE2 mentions a conference at UTN where “AI dominated proceedings”, and that, in one class, her lecturer “asked for the use of AI”. EE6, however, comparing her experience in Ecuador and Chile, notes that in Chile “lecturers don't mention artificial intelligence tools, and we don't use it for classes either, not even for assignments”.

UNIR (distance-learning mode) reports having chatbots integrated into its platform. “We even have a chat, with ChatGPT, through which we can complete assignments and take part in forums, since it is entirely online in any case” (EE8). At UTN, AI has taken over the scientific research agenda (as in the ATC conference mentioned above), with lecturers now even requiring logical frameworks and advanced prompting within ICT subjects. In addition, at the PUCE campuses (Ibarra and Santo Domingo), contrasting environments can be observed. While some tutors guide the use of AI during the thesis stages for in-depth searches, other lecturers show outright rejection, grounded in a fear of academic fraud, and prefer traditional library-based methodologies.

Universities are called upon to provide more training, so that they stop prohibiting AI and instead offer formal courses on how to use it under ethical principles and with guidelines on academic honesty. “I feel it would be better to try to explore in greater depth what these tools are, rather than merely naming the type they are” (EE3). “If AI is already the new era in the world […] trying to prohibit it harms us in terms of the future” (EE1). There is a consensus regarding the need to learn to construct better prompts and to verify information. “One should know quite a lot in order to be able to work with AI; it comes down to crafting good prompts” (EE7).

By contrasting lecturers' and students' perceptions and practices with the AI competency framework for higher education identified by UNESCO-IESALC (2025); Zhao et al., 2026, points of convergence and divergence can be identified. Among the former are the following: The differences are as follows:

• Demand for training Both lecturers and students acknowledge the need for greater training. Lecturers specify in detail the type of training they require, whereas students request formal spaces that most of their universities have yet to offer systematically.
• Absence of a regulatory framework Neither students nor lecturers find, within their institutions, a clear and consolidated regulatory framework for the use of AI. This absence generates uncertainty among students and frustration among lecturers.
• Ethical risks The dimension of “AI ethics” and a “human-centred mindset” is the most highly valued by both groups. Concern regarding cognitive dependency and the loss of soft skills (writing, public speaking, deep reading) constitutes the greatest common denominator. Plagiarism is likewise mentioned by both groups. Awareness of hallucinations is greater among lecturers, although it also emerges among some students.
• Widespread use Both lecturers and students use AI on a daily basis, although the degree of reflection on such use varies. It is generally accepted that AI “is here to stay” and that prohibiting it would prove counterproductive.

• Uses Lecturers use AI for administrative management and to improve pedagogical efficiency but are reluctant to integrate it into the classroom. It can assist critical thinking and problem-solving. They express concern about ethical use and bias mitigation. Students use AI on a daily basis to study or structure their theses. It functions as a didactic and interactive resource, as well as a support tool for learning.
• Training For lecturers, this represents a continuous, self-taught learning process that adapts to AI's new functionalities. Students indicate a need for training integrated into the curriculum. AI “should already form part of education as an additional subject”. There is demand for formal, systematic digital literacy.
• Impacts Lecturers' perspective concerns technological inequality between public and private education, the urban-rural divide, the risk of deprofessionalisation, and the transformation of pedagogical relationships. According to students, the impacts are diffuse and individual in nature.
• Professional development Continuous training in AI is a necessity that affects lecturers' day-to-day practice. Many feel alone in this process of updating their skills. For students, mastering AI is a skill for the future labour market, although its urgency is not perceived as such.
• Teaching role Teachers admit to feeling overwhelmed by the speed of change and by the lack of clear institutional guidelines on how to sanction or validate the use of AI. Students indicate that many lecturers “are very closed-minded” or display “fear and a lack of knowledge”.
• Pedagogical design This is evident when lecturers design learning activities that incorporate AI strategically, whereas students make use of lecturers' proposals in order to use AI reactively, in response to specific needs.
• Attitude towards AI Lecturer: “It is not intelligence, it is patterns.” “A third actor within the educational triad.” Technical and conceptual understanding. They incorporate notions regarding algorithms, bias, differences between types of AI and associated risks. Students: AI is a useful application that provides functional knowledge (which tools to use, and for what purpose).
• Equity Among lecturers, there is a sense of inequality in students' prior preparation, particularly among those from rural areas or with more limited access to technology during their basic education. Students perceive inequality in access to paid tools or advanced functionalities.

4.3 Survey results

All six constructs display internal consistency (α > 0.80); the items within each dimension coherently measure the underlying dimension. However, the “Regulation and Governance” construct has an α of 0.203, which is unacceptable. This constitutes evidence that the two items conceptually measure opposing directions: one assesses the normative demand for regulation, while the other assesses the perceived clarity of existing regulation. Construct means range between 3.40 (Feasibility) and 3.68 (AI Use), indicating a tendency towards moderate agreement across all dimensions. No construct records a particularly high or particularly low mean; students do not display extreme positions with regard to AI.

The correlational analysis using Spearman's ρ reveals associations consistent with the theoretical model of technology acceptance.

“AI Use” is positively and significantly associated with “Usefulness” (ρ = 0.571), “Motivation” (ρ = 0.601) and “Feasibility” (ρ = 0.348), all with p < 0.001. This demonstrates that students who perceive greater usefulness, motivation and ease of use tend to report greater use of AI.

“Usefulness” and “Motivation” are correlated (ρ = 0.730): students who perceive AI as useful for their academic performance are also more motivated to use it.

“Ethical Concern” does not correlate with any of the acceptance constructs (Use, Usefulness, Motivation, Feasibility). Ethical concerns apply independently of perceptions towards AI.

“Regulation” is weakly but significantly associated with “AI Use” (ρ = 0.292), “Feasibility” (ρ = 0.222), “Usefulness” (ρ = 0.265), “Motivation” (ρ = 0.211) and “Ethical Concern” (ρ = 0.240). Demand for regulation is present both among intensive users and among those who perceive greater risks.

The first hypothesis is not confirmed. Friedman's test: χ² = 8.321, df = 5, p = 0.139. The mean scores of the six constructs do not differ to a statistically significant degree. Students display relatively homogeneous scores across all dimensions, revealing a balanced profile within the sample. In the post-hoc analysis with Bonferroni correction (α = 0.0033), the comparison between “AI Use” and “Feasibility” of use proved significant (p = 0.001), indicating that students report using AI with a frequency/intensity slightly higher than that which they perceive to be easy to use.

The second hypothesis, that “women display greater awareness across all dimensions”, is not confirmed. Men score significantly higher on “use”, “perceived ease”, “usefulness” and “motivation”. Women score significantly higher only on “ethical concerns” and risk. No difference is found in relation to regulation. Men position themselves as more intensive users, with greater confidence in the tools, whereas women display greater critical sensitivity towards ethical risks (plagiarism, dependency, uncertainty).

The third hypothesis, “students use AI despite their ethical concerns”, is confirmed. Wilcoxon test: W = 7193.5, p = 0.380. Spearman correlation, Use ↔ Concern: ρ = 0.072, p = 0.313. Although the test of mean differences is not significant (Use = 3.68 vs. Concern = 3.56), the key finding is the absence of any correlation between “Use” and “Ethical Concern”. Both dimensions are thus statistically independent: knowing a student's level of ethical concern does not allow one to predict his or her level of use, and vice versa. This suggests that students maintain a double-bind stance: they actively use AI in their studies (writing, summarising, comprehension, time optimisation) while simultaneously expressing moderate-to-high concerns regarding plagiarism, dependency, trust in the information provided, and effects on learning. There is no inhibitory effect whereby ethical concerns reduce use.

The fourth hypothesis, that perceived regulatory clarity is significantly lower than the perceived need for regulation, is confirmed. Paired Wilcoxon test: W = 1189.0, p < 0.001. Mean difference = 0.80 points (d ≈ 0.69). A perceived regulatory gap emerges within the Ecuadorian university context. Students consider that the use of AI in higher education should be regulated (mean = 4.00, close to agreement), yet simultaneously indicate that it is not clearly regulated within their universities (mean = 3.21). A legitimate demand for regulatory frameworks is thus perceived. Students operate in the absence of such frameworks, using AI without clear guidance on limits or academic integrity policies.

The fifth hypothesis, that “perceived usefulness and motivation positively predict AI use, controlling for ease of use”, is confirmed. Spearman partial correlations, controlling for “Feasibility”: “Usefulness” → “Use”: ρ = 0.516, p < 0.001; “Motivation” → “Use”: ρ = 0.546, p < 0.001. Both “Usefulness” and “Motivation” maintain positive, strong and significant predictions of AI use. This validates a technology acceptance model in which instrumental usefulness (improved performance, efficiency, comprehension) and extrinsic motivation (time saved, professional utility, contextual relevance) act as autonomous determinants of usage behaviour. Ease of use explains very little of the usefulness-motivation → use relationship. Students do not use AI primarily because it is easy; they use it because they perceive it as useful and motivating.

The sixth hypothesis, “demand for regulation is positively associated with perceived risk and with level of use”, is confirmed. Regulation ↔ Perceived risk: ρ = 0.240, p < 0.001. Regulation ↔ Level of use: ρ = 0.292, p < 0.001. Demand for regulation does not arise solely from students who reject AI; it is positively associated with both perceived risk and level of use. Those who use AI most are also those who most demand clear rules. At the same time, those who perceive greater ethical risks likewise demand more regulation. Regulation is thus not viewed as a restriction imposed by non-users of the technology, but rather as a necessary form of ordering.

4.4 Content analysis results

Seven university documents were identified that contain regulations on the use of AI directed towards the teaching and learning process, for both lecturers and students (Table 5). Instruments with general regulatory approaches were selected; those consisting of guidelines oriented specifically towards the preparation of academic articles were excluded.

Table 5. Regulations on the use of AI at Ecuadorian universities.

Source: authors' own elaboration.

These were then compared against the UNESCO-IESALC (2025) framework; the results are presented in Table 6. The few Ecuadorian universities that have policies for the integration and use of AI address the dimensions of ethics and control more thoroughly, whereas those relating to competency development appear to be neglected; Ecuadorian university regulation is therefore in a regulatory phase, rather than a phase of AI competency development. It is observed that the hierarchy of competencies addressed is not the same for lecturers as for students, although in both cases AI ethics occupies the first position.

Table 6. AI dimensions for lecturers and students.

Dimensions: A) Human-centred approach; B) Ethics in the use of AI; C) Technical applications associated with the use of AI; D) AI pedagogy; E) Application to professional teacher development; F) AI for research; G) Design of discipline-specific AI systems. Compliance: Full = 1; Partial = 0.5; None = 0. Source: authors' own elaboration.

In second place is conceptual understanding of the technology; students are primarily guided towards permitted uses and conditions of application, which is consistent with the procedural nature that predominates in the provisions directed at the student body. At the opposite end of the scale, the least well-addressed competencies also partially coincide between the two groups. Professional development and AI for research record almost no mention regarding teaching staff, while the design of discipline-specific AI systems records no mention at all with regard to students. In both groups, these correspond to the most specialised dimensions within their respective frameworks.

5 Discussion

The results show that AI is, in practice, being integrated into social science (CCSS) teaching in Ecuador, although this integration is not underpinned by a broad and consolidated institutional design. This scenario is consistent with the warning issued by Valentini and Blancas (2025) regarding the absence of guidance for the responsible use of AI across most higher education institutions in the region and is also in line with what Zhao et al. (2026) identify as “inconsistent institutional guidance” in the adoption of AI. The Ecuadorian case confirms this gap on three fronts: the experiences of lecturers and students, the perceptions quantified through the survey, and the content of the institutional policies and provisions analysed. This evidence shows that the incorporation of AI advances mainly through the initiative of its users, while systematic training, the redesign of assessment and institutional regulation develop more slowly. This cross-reading of the results is one of the main contributions of this study, since it supports the argument that university AI governance and guidelines must articulate competencies differentiated by academic role and by the particularities of each discipline (UNESCO-IESALC, 2025). This differentiation finds further support in the AI literacy competency framework proposed by Chee et al. (2025), which documents significant variation according to user profile and reinforces the need not to treat lecturers and students as a homogeneous group.

The lecturers interviewed agree that the incorporation of AI is irreversible, that the training they have received is insufficient, and that it is necessary to foster critical thinking. These findings are consistent with Alé et al. (2025) and Sartor-Harada and Ulloa-Guerra (2026), who note that effectively integrating AI requires technical, pedagogical and scientific competencies. This is consistent with the teacher AI competence self-efficacy scale (TAICS) (Chiu et al., 2025), which confirms training and self-perceived competence as determining factors in the effective integration of AI into teaching practice.

Beyond the technology-acceptance factors examined by the TAM/UTAUT models used in STEM degree programmes (Liu et al., 2024; Strzelecki, 2024), the accounts from lecturers reveal a crisis in modes of assessment in social science education, and point to the need to incorporate oral defence, prompt traceability, or the analysis of the quality of questions posed to AI. Lecturers also point to the need to distinguish between instrumental and reflective uses of AI, a concern that aligns with the training urgency identified by Ramírez-Montoya et al. (2024). This pedagogical warning also connects with the demand for institutional regulations noted by Mejía and Gómez (2024). However, the institutional response should not be confined to the technical learning of tools, nor should it translate solely into restrictions, since prohibition-centred regulation could reduce pedagogical initiative without resolving the underlying problems linked to ethical integrity, critical thinking and the quality of learning.

The students' accounts align with Duque-Rengel and Puertas-Hidalgo (2026) and Grájeda et al. (2024) in that social science students possess functional knowledge for using AI tools, yet they still show limitations in critically evaluating AI-generated responses, recognising hallucinations or applying the tools in specialised ways. Participants value the usefulness of these tools, as noted by Espinoza-Bravo et al. (2025) and Saltos et al. (2025); however, their experiences show that a positive assessment does not necessarily equate to adequate mastery. The results also qualify the claims made by Asencio et al. (2026), since institutional integration is not homogeneous: it varies across universities, campuses, degree programmes and even among lecturers within the same institution. Students' experience may therefore depend more on individual initiative than on a shared institutional strategy.

Another relevant finding is the gap between the need for regulation and perceived regulatory clarity (Hypothesis 4). This result does not emerge in isolation from the survey; rather, it coincides with lecturers' and students' accounts of the absence or inconsistency of institutional guidance, as well as with the heterogeneity observed across the seven regulatory instruments analysed. The convergence of the three sources thus supports the existence of a gap between usage practices and the development of institutional guidance. Likewise, the associations of perceived usefulness and motivation with use remain significant when controlling for feasibility (Hypothesis 5). This result qualifies the centrality that TAM/UTAUT models assign to perceived ease of use (Liu et al., 2024; Strzelecki, 2024) and suggests that, in this sample, perceived usefulness and extrinsic motivation are closely related to use. The pattern observed by gender was not examined in the Ecuadorian studies reviewed and should therefore be regarded as an exploratory finding requiring confirmation in future research (Hypothesis 2 not confirmed).

The coexistence of frequent use and ethical concern, as reported by students in the survey, reveals a tension that deserves particular attention. Students may recognise the risks of plagiarism, dependency, loss of learning or lack of reliability while, at the same time, continuing to use AI for the time it saves and the support it offers in completing academic tasks. This ambivalence is consistent with Arce et al. (2025), who describe a relationship characterised by moderate dependency and ambivalent attitudes, and with Vilma et al. (2025), who identify the use of AI to improve writing alongside a fear of becoming dependent on it. The specific contribution of this study is to show that ethical concern is not significantly associated with the reported level of use. This suggests that recognising risks does not, on its own, constitute a sufficient mechanism of self-regulation. Accordingly, the demand for ethics training and clear institutional frameworks identified by Saltos et al. (2025) should translate into concrete competencies for verifying information, cross-checking sources, declaring the use of AI, and distinguishing between technological assistance and the substitution of intellectual work.

The lecturer interviews show that this tension is particularly acute in the social sciences, where training involves not only obtaining information or producing well-structured texts but also interpreting complex phenomena, contrasting perspectives and constructing one's own arguments. Participants warn that indiscriminate delegation of tasks to AI may weaken processes such as deep reading, reflection and taking a position; these concerns are echoed in the student interviews, where participants recognise risks of dependency, loss of authenticity and difficulty in verifying the information generated. These findings align with Valentini and Blancas (2025), who argue that AI literacy must extend beyond STEM disciplines to consider the particularities of the arts and humanities, as well as with the human-centred perspective proposed by Miao (2024). Nevertheless, lecturers' accounts also show that AI can foster critical thinking when used to compare explanations, identify bias, formulate questions and debate its responses.

In the social sciences, therefore, the analysis should not be limited to accepting or prohibiting AI but should instead focus on distinguishing between uses that support disciplinary reasoning and those that substitute for intellectual processes that are essential to students' training. This gains further relevance in light of the systematic review and conceptual framework proposed by Helal et al. (2025), who assess AI's effects on critical thinking and caution that these depend on the type of use rather than on exposure to the technology alone.

The analysis of university policies indicates that the AI ethics dimension occupies the first position for both lecturers and students, while professional development and the design of discipline-specific AI systems go unmentioned; this confirms the warning issued by Valentini and Blancas (2025) regarding the reduction of AI literacy to its ethical and control dimension. This is consistent with Zhao et al. (2026) and Miao (2024): by only partially addressing the “human-centred mindset”, priority is given to risk mitigation over capacity development. Only one of the seven instruments is formally designated a “policy” (Universidad Internacional SEK), while the rest adopt a variety of names and levels of approval. This heterogeneity shows that the institutionalisation of AI guidelines follows different trajectories across the universities analysed. Moreover, Tsao (2025) shows how institutional discourses and the practices declared by students and lecturers do not always converge towards the same normative direction.

The recent “Strategy for Fostering the Development and Ethical and Responsible Use of Artificial Intelligence in Ecuador” (Registro Oficial, 2026) provides a national framework, although its translation into university-level instruments remains at an early stage. The CACES (2023) accreditation indicator introduces an external requirement regarding the appropriate use of AI; compliance with it, however, does not on its own guarantee the comprehensive development of the competencies proposed by Verduga-Alcívar et al. (2026).

The study thus shows that the adoption of generative AI in university-level social science teaching cannot be understood solely on the basis of individual acceptance of the technology. Integrating the results of the three instruments reveals a complex relationship between practical usefulness, insufficient training and institutional governance. While lecturers and students incorporate AI to meet immediate needs, universities still concentrate their efforts on ethical regulation and address professional development, research and discipline-specific competencies to only a limited extent. In the Ecuadorian context, this governance gap overlaps with the limitations in connectivity, access to licences and teacher training identified in the interviews and described by Hernández (2025) and Jara (2024). The contribution of this study lies not merely in recording a regulatory lag, but in showing that technological adoption unfolds within a scenario shaped by institutional capacities and inequalities of access that must be taken into account when designing policies for the social sciences.

The findings contribute evidence on AI governance in Ecuadorian social-science higher education, an area in which the literature has so far been concentrated on other regions and disciplines (Liu et al., 2024; Strzelecki, 2024). By triangulating interviews, survey data and content analysis, the study offers a reproducible pathway for examining the coherence between institutional discourse, actors' practice and formal regulation, as in Tsao (2025). Finally, the study contributes to the discussion on equity of access to AI in higher education (Apata et al., 2026) by showing that gaps in connectivity, licensing and teacher training constitute a matter of educational justice that must be explicitly incorporated into the design of university AI policies.

6 Conclusions

All three research objectives were achieved. The first, determining lecturers' and students' perceptions of AI integration, was met through interviews that placed participants at a foundational level of competency, with greater development evident in the human-centred and ethical approach than in technical and design competencies. The second objective, establishing the factors that motivate and constrain AI use, was achieved through the qualitative and quantitative identification of perceived usefulness and motivation as determinants of use, ahead of ease of use, together with the identification of constraints such as connectivity, access to licences and teacher training, as reported in the interviews. The third objective, identifying university policies on AI, was achieved through a review of the seven Ecuadorian university regulations, which display uneven attention to the competency dimensions defined by UNESCO-IESALC (2025); within these, ethics and control predominate over the dimensions of professional development and research.

Of the six hypotheses, four were confirmed and two were not. Hypothesis 1 (significant differences among the six survey constructs) and Hypothesis 2 (greater awareness among women across all dimensions) were not confirmed. It was confirmed that students use AI despite their ethical concerns (Hypothesis 3), that a gap exists between the demand for regulation and perceived regulatory clarity (Hypothesis 4), that usefulness and motivation predict use behaviour (Hypothesis 5), and that demand for regulation is associated with both perceived risk and level of use (Hypothesis 6).

For lecturers and students of social sciences in Ecuador, the factors that motivate use are essentially instrumental, while those that constrain it are structural. The use of AI among social science students is moderate-to-high (mean = 3.68) and is governed by perceived usefulness and extrinsic motivation, rather than by ease of use. There is an institutional regulatory deficit (demand for regulation = 4.00, compared with perceived clarity = 3.21), which makes governance a priority. These findings correspond to the surveyed and interviewed sample and should not be interpreted as representative of the entire population of social science university students in Ecuador.

In light of the hypotheses put forward in this research, it is concluded that students' acceptance of AI does not imply an absence of ethical questioning or opposition to regulation; indeed, there are indications of a profile centred on pragmatic acceptance of AI accompanied by high expectations of governance. This shows that, in the Ecuadorian case, the institutional management of AI in higher education follows the inertia of organic use in the classroom, regardless of any formal regulation. Regarding gender, there is no substantial difference in students' use and perception of AI; nevertheless, women show a slightly greater critical awareness of the ethical behaviour of AI and its implications, whereas men use AI more intensively and with greater trust.

It should be noted that both male and female university students are indeed aware of the risks of dependency, trust and effects on learning that AI may bring about; however, ethical dilemmas surrounding AI do not limit students' use of it, and students also remain largely uninformed about the regulation and limitations governing its use. It is likewise significant to note that higher education students do not use AI because they find it easy, but because they consider it useful and inspiring. Ecuadorian universities have shown very limited attention to building a regulatory framework for AI with implications for lecturers and students alike. There is evidence that certain higher education institutions have regulatory procedures in place that reflect a symbiosis between ethical aspects and control over the use of AI. These statements should be read as a characterisation of the participating sample's profile, rather than as a generalisation valid for the whole of Ecuador's higher education students and institutions.

This has direct practical implications for universities and for sectoral public policy: as long as the institutions responsible for accrediting university quality in Ecuador limit themselves to assessing the existence of a document in the form of a policy, without differentiating competencies by discipline and role, as established by UNESCO-IESALC (2025), it is foreseeable that higher education institutions will continue to produce instruments of symbolic visibility rather than investing in a genuinely comprehensive policy with a real impact on training lecturers and students to use AI ethically.

The results and conclusions presented should be interpreted considering certain limitations. The survey sample is non-probabilistic, having been obtained through snowball sampling, and 78% of it is drawn from five universities; consequently, these results should not be regarded as representative of all social science students in Ecuador, but rather as evidence pertaining specifically to the participating institutions. The “regulation” construct displayed unacceptable reliability (α = 0.203). Finally, Hypothesis 5 does not incorporate sociodemographic variables that could act as confounding factors. Looking ahead, this study lays the groundwork for comparing Ecuador's results with those of other Andean countries, both in terms of perceived relationships and the institutional frameworks of private and public universities.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. During the preparation of this manuscript, the authors drew on Sonnet 5, from the Claude family (Anthropic), a generative artificial intelligence tool, to categorise and organise the interviewees' accounts and to compile a table of the AI regulations of Ecuadorian universities. Based on the AI's output, the content was reviewed and edited in line with the researchers' interpretation. The authors take full responsibility for the text. The tool used was freely accessible at the time the manuscript was drafted. The instructions provided consisted of a generic prompt aimed at grouping the interview testimonies into emerging thematic categories and systematising, in a comparative table, the information from the seven institutional regulations identified, without incorporating the model's own interpretations or conclusions. The findings produced by the tool were fully corroborated by the research team. The interview data processed using this tool had previously been anonymised. The informed consent given by participants explicitly included the option for the anonymised data to be processed through AI tools or services for systematisation and study purposes.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/feduc.2026.1948805/full#supplementary-material

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Bruce Maxim, University of Michigan, United States

Reviewed by: Gianmarco Bonavolontà, University of Cagliari, Italy

Analia Veronica Losada, Universidad de Flores, Argentina