Actitudes positivas,
desigualdades persistentes: Análisis crítico de las percepciones de género
sobre Inteligencia Artificial en educación superior peruana
Julio Postigo-Zumarán. Universidad de San Martín
de Porres. Perú.
León Maximiliano Cutipa-Murga. Universidad
Tecnológica del Perú. Perú.
Jorge Polanco-Argüelles. Universidad
Tecnológica del Perú. Perú.
Julio César Huamani-Cahua. Universidad de San
Martín de Porres. Perú.
Received: 2026-03-07; Revised: 2026-03-12; Accepted: 2026-04-27; Published: 2026-09-01
How
to cite:
Postigo-Zumarán, J., Cutipa-Murga,
L.M., Polanco-Argüelles, J., Huamani-Cahua, J.C. (2026). Positive attitudes,
persistent inequalities: Critical analysis of gender perceptions about
Artificial Intelligence in Peruvian higher education [Actitudes positivas,
desigualdades persistentes: Análisis crítico de las percepciones de género
sobre Inteligencia Artificial en educación superior peruana]. Pixel-Bit,
Revista de Medios y Educación, 77, Art. 6. https://doi.org/10.12795/pixelbit.120494
ABSTRACT
The growing integration
of artificial intelligence in higher education requires understanding student
attitudes toward this technology, particularly in Latin American contexts where
research is scarce and measurement instruments lack cultural validation. This
study analyzed gender differences in attitudes toward AI among 1,912 university
students from Arequipa, Peru, using the AIAS-4 scale adapted to Spanish.
Exploratory (n=1,136) and confirmatory (n=776) factor analyses confirmed a unidimensional
structure with adequate psychometric properties (α=.884;
CFI=.991; RMSEA=.076). Results revealed that female students present
significantly more positive attitudes toward AI than males (M=3.15 vs. M=2.97;
p<.001; d=0.204), contradicting patterns reported in European and Asian
contexts. A descending trend in positive attitudes as the academic year
progresses was also observed. These findings reveal a critical paradox: favorable
attitudes coexist with structural female underrepresentation in the
technological field. Implications for designing gender-sensitive educational
policies that promote equitable participation in artificial intelligence are
discussed.
RESUMEN
La creciente
integración de la inteligencia artificial en la educación superior requiere
comprender las actitudes del estudiantado hacia esta tecnología, especialmente
en contextos latinoamericanos donde la investigación es escasa y los instrumentos
de medición carecen de validación cultural. Este estudio analizó las
diferencias de género en las actitudes hacia la IA en 1,912 estudiantes
universitarios de Arequipa, Perú, utilizando la escala AIAS-4 adaptada al
español. Se aplicaron análisis factorial exploratorio (n=1,136) y confirmatorio
(n=776), confirmando una estructura unidimensional con propiedades
psicométricas adecuadas (α=.884; CFI=.991; RMSEA=.076). Los resultados revelaron
que las estudiantes mujeres presentan actitudes significativamente más
positivas hacia la IA que los hombres (M=3.15 vs. M=2.97; p<.001; d=0.204),
contradiciendo patrones reportados en contextos europeos y asiáticos. Se
observó además una tendencia descendente en actitudes positivas conforme avanza
el año académico. Estos hallazgos evidencian una paradoja crítica: actitudes
favorables coexisten con subrepresentación estructural femenina en el campo tecnológico.
Se discuten las implicaciones para el diseño de políticas educativas con
perspectiva de género que promuevan participación equitativa en inteligencia
artificial.
KEYWORDS · PALABRAS
CLAVES
Inteligencia
artificial; Educación superior; Género; Tecnología educativa; Alfabetización
digital · Artificial intelligence; Higher
education; Gender; Educational technology; Digital literacy.
1.
Introduction
Artificial intelligence
(AI) has established itself as a fundamental agent of transformation in various
dimensions of contemporary life, including higher education, where it is being
progressively integrated into teaching, learning, and academic management (Bond
et al., 2024). In Latin America, the use of AI in higher education advances
amid opportunities for pedagogical innovation and structural challenges related
to technological infrastructure, teacher training, and differences in
accessibility (Fernández-Miranda et al., 2024; Salas-Pilco &
Yang, 2022; Heredia Pérez et al., 2025; Aviles-Valenzuela et al.,
2025). Regarding
the Peruvian context, recent research shows that scientific production on AI in
higher education represents only 2% of total publications worldwide, reflecting
a notable asymmetry compared to countries like China and the United States (Contreras
& Olaya, 2025; Tapullima Mori et al., 2024), and persistent infrastructural gaps that thwart the
democratization of access to these technologies (Contreras
& Olaya, 2025).
The most recent and
prominent subfield in AI is Generative Artificial Intelligence, which has
experienced a significant upswing with the development of tools such as ChatGPT,
Gemini, and Copilot, which are capable of creating text and images and
generating code through machine learning algorithms (Jovanović
& Campbell, 2022; Chiu, 2024). Generative AI offers undeniable opportunities for
the personalization of learning, knowledge management, and the stimulation of
creativity (Jovanović & Campbell, 2022; Peres
et al., 2023). On the other hand, there have emerged ethical and
academic questions that affect the originality of the work, the improvement of critical
thinking, and the protection of academic integrity (Bittle
& El-Gayar, 2025; García-López & Trujillo-Liñán, 2025;
Kofinas et al., 2025).
A key aspect in the
implementation of these emerging technologies is the analysis of students'
attitudes towards AI, which involve cognitive, affective and behavioral
dispositions that predispose the acceptance, commitment and critical
appreciation of such tools (Katsantonis
& Katsantonis, 2024; Sultana et al., 2025). The international literature on gender differences
in attitudes toward AI presents heterogeneous and, at times, contradictory findings.
While some studies do not find significant differences between men and women (Purificato
et al., 2023; Hajam & Gahir, 2024), others document that men show more positive
attitudes, greater frequency of use, and less technological anxiety regarding
AI (Stöhr
et al., 2024; Russo
et al., 2025; Ofosu-Ampong,
2024). This trend has been mainly
documented in European and Asian contexts, where women tend to perceive AI
tools as more complex and less useful for their academic needs (Beig
& Qasim, 2023).
Recent research also suggests that
AI anxiety acts as a mediating variable in this relationship: women are the
ones who report higher levels of this construct, which could explain part of
the observed attitudinal differences (Russo et al., 2025). Concerning the
specific use of generative tools, men use platforms like ChatGPT more
frequently and for a broader range of academic tasks (Stöhr et al., 2024;
Møgelvang et al., 2024). This body of evidence, constructed almost entirely outside
Latin America, leaves a relevant question open: whether these patterns reflect
something universal or whether they are conditioned by cultural, economic, and
institutional factors that vary according to context.
In this context,
empirical evidence on gender differences in attitudes toward AI in Latin
America, and especially in Peru, is scarce and fragmented, which makes it
impossible to recognize regional patterns and evidences the significant gap in the
specialized literature. Most studies have observed generalities about the
implementation of technologies, neglecting the gender dimension in the critical
appropriation of AI (Frumin
et al., 2026; Valdivieso & González, 2025;
Ancheta-Arrabal et al., 2021). This knowledge gap is worrying because gender
inequalities in science, technology, engineering, and mathematics (STEM)
continue in the region, given that only one in three scientists is a woman and
at higher academic levels, that proportion continues to decline (UNESCO,
2024).
This study is presented
as a response to the knowledge gap mentioned before. One of its aims is to
contribute to the analysis of gender differences in attitudes toward artificial
intelligence in university students in Arequipa, Peru, using the AIAS-4 scale (Grassini,
2023) in its version adapted for the Latin American
context. The research falls under the field of Emerging Technologies and
Critical Artificial Intelligence, particularly on artificial intelligence and
the reproduction of structural inequalities in education.
This critical perspective
becomes especially relevant because individual attitudes toward technology do
not operate in a social vacuum, but are situated within institutional,
cultural, and economic frameworks which shape access, training, and career
opportunities (Cai
et al., 2017; Li et al., 2025). Therefore, even if students have positive attitudes
toward AI, such dispositions do not necessarily lead to equitable engagement in
the design, development, and governance of AI systems. This explains the need
to critically question the relationship between favorable attitudes and
persistent differences, avoiding the uncritical interpretation that
psychologically positive dispositions are assumed to be sufficient to overcome
structural gender barriers in a given technological field.
The main objective of
this study is to analyze gender differences in attitudes toward artificial intelligence
in university students in Arequipa, Peru. As secondary objectives, differences
are examined according to the field of study and academic year, which makes it
possible to situate the central finding within the specific academic conditions
of the sample and evaluate to what extent the attitudes toward AI vary
depending on students' formative paths.
2. Method
2.1. Research design
This research adopted a
descriptive-comparative quantitative approach with a cross-sectional
non-experimental design (Ato et al., 2013). Data collection occurred at a single point in time during
the 2025 academic year. It is worth noting that the cross-sectional design
allows establishing differences between groups at a given time and examining
associations between variables but does not enable inferences about
directionality or causality (Ato et al., 2013).
Given that the
instrument used (Grassini's AIAS-4, 2023) was originally developed in a
Norwegian context with a 10-point scale and considering that it required linguistic
and psychometric adaptation to the Latin American context, content validation
and exploratory factor analysis procedures were incorporated as preliminary
steps to the comparative analysis. These validation procedures do not
constitute the main objective of the study but guarantee the reliability of the
findings on gender differences reported in it.
2.2. Participants
The target population
consisted of undergraduate university students enrolled in higher education institutions
in Arequipa, Peru's second most populated city and the main university center
in the south of the country. The sample was obtained from a private university
in the city, selected for its accessibility and for offering academic programs
in the three areas of knowledge considered in the study. Data were collected
directly in classrooms during regular class hours; participation was voluntary
in all cases. The initial sample comprised 1,945 students; after excluding 33
cases due to unspecified gender or incomplete questionnaires, the final
analytical sample consisted of N = 1,912 participants.
The gender distribution
was relatively balanced, with 905 men (47.3%) and 1,007 women (52.7%).
Regarding the areas of study, the sample was distributed as follows: Area 1 (n
= 373, 19.5%), Area 2 (n = 997, 52.1%), and Area 3 (n = 542, 28.4%). Concerning
the distribution by academic year, there was a larger concentration in the intermediate
and advanced years: first year (n = 200, 10.5%), second year (n = 240, 12.5%),
third year (n = 293, 15.3%), fourth year (n = 342, 17.9%), fifth year (n = 607,
31.7%), and sixth year (n = 230, 12.0%). Ages ranged from 18 to 28 years (M =
21.76, SD = 2.23).
The inclusion criteria
were: (a) being enrolled at the university in the academic year 2025, (b)
voluntarily agreeing to participate in the study by providing informed consent,
and (c) having fully completed the questionnaire. Participants with incomplete
responses, duplicate responses, or invalid response patterns were excluded.
Purposive sampling was
adopted for operational viability and thematic relevance: it guaranteed access
to groups with diverse academic profiles - by area and year of study - in an
institution that concentrates representative programs of the private university
system in Arequipa. This design does not allow us to assert statistical representativeness
with respect to the entire university population of Arequipa, but it provides a
broad and heterogeneous sample that is suitable for the comparative objectives
of the study.
2.3. Instrument
To measure attitudes
toward artificial intelligence, the AIAS-4 scale was used. It was developed and
validated by Grassini (2023) with 1,946 Norwegian participants. The scale
consists of four items that evaluate general attitudes toward AI from a unidimensional
perspective. Its selection was based on three criteria: the instrument's
concision, which allows its administration in classroom contexts without
disrupting the class dynamic; the replicability of its unidimensional structure
in different cultural contexts, which facilitates comparisons between studies;
and the fact that its items do not presuppose technical familiarity with AI
systems, which makes it suitable for populations with heterogeneous educational
backgrounds. For this study, the five-item version was applied with the
objective of empirically evaluating the behavior of the additional item in the
Peruvian context. The instrument was adapted through translation-back
translation into Spanish and adjustment of the original scale —of 10 points— to
a five-point Likert format (1 = Strongly disagree to 5 = Strongly agree),
following common practices in Latin American educational research (Lloret-Segura
et al., 2014).
2.4. Procedure
The study was approved
by the institution's academic authorities and followed the ethical principles
of the Declaration of Helsinki. All participants received information about the
study's objectives, the voluntary nature of their participation, data confidentiality,
and their right to withdraw at any time. The data collection and analysis
processes were carried out in three sequential phases.
Phase 1: Content adaptation
and validation. Five expert judges in education, psychology, and educational
technology evaluated each item in three dimensions (clarity, coherence, and
relevance). Aiken's V coefficient was calculated to quantify the degree of
agreement, considering V values ≥ .70 as acceptable (Aiken,
1980; Merino-Soto,
2023). The results showed values above .72 in
all dimensions, confirming the validity of the content of the adapted
instrument.
Phase 2: Application and
construct validation. The questionnaire was administered in person during
regular class times, with an average time of five minutes. For the validation
of the construct, the analytical sample (N = 1,912) was randomly divided into
two independent subsamples: the first (n = 1,136; 59.4%) was used in
Exploratory Factor Analysis (EFA) and the second (n = 776; 40.6%) in
Confirmatory Factor Analysis (CFA), according to the two-stage validation procedure
recommended for instruments of new adaptation (Lloret-Segura et al., 2014). The
exploratory factor analysis (EFA), executed in Jamovi 2.3, identified a
unidimensional structure that explained 64% of the total variance. The item 5,
formulated in reverse order concerning the perception of AI as a threat, showed
a moderate negative factor loading (−.45), a behavior that replicates the
one reported in the original validation by Grassini (2023), and for this
reason, it was excluded from the composite score. The resulting four-item scale
showed adequate reliability indicators (Cronbach's α =
.884; McDonald's ω = .886). The confirmatory factor analysis (CFA),
applied to the second subsample with the MLR estimator, corroborated the
unidimensional structure; the indices are reported in section 3.1.3.
Phase 3: Analysis of
gender differences. Once the psychometric properties of the instrument were confirmed,
comparative analysis was carried out, which constitutes the main objective of
the study.
2.5. Data analysis
Statistical analyses
were performed using Jamovi version 2.3 (The Jamovi Project, 2024). Descriptive
statistics were calculated to depict attitudes toward AI in the total sample
and by subgroups. To evaluate gender differences, the main objective of this
study, Student's t-test for independent samples, was applied, after verifying
the homogeneity of variances using Levene's test. The effect size was
calculated using Cohen's d, interpreted according to conventional criteria:
small (d ≈ 0.20), medium (d ≈ 0.50), and large (d ≈ 0.80) (Cohen,
2013). Additionally, differences according to area of knowledge
and academic year were explored using one-way analysis of variance (ANOVA),
complemented by Tukey post hoc tests when the F statistic was significant.
Prior to the
comparative analysis, the required statistical assumptions were verified. The
independence of the observations was ensured by the study design: each
participant answered the questionnaire only once and individually, with no
possibility of influence when answering the questions. Regarding normality, although
the Kolmogorov-Smirnov tests were statistically significant - an expected
result with samples larger than 1,000 cases, where these tests detect trivial
deviations – the visual inspection of the histograms and Q-Q plots showed
approximately symmetrical distributions in both groups. Given the sample size
is large (n > 300 per group), the central limit theorem guarantees that the
sample means approximate to a normal distribution, consequently justifying the
use of parametric tests (Field, 2018; Pallant, 2020).
Figure 1 summarizes the
research model, identifying the dependent variable, the independent variable,
the covariates controlled in the multivariate analysis, and the subsamples used
in the psychometric validation phases.
Figure
1
Research
model
Note. Solid arrows indicate analyzed
relationships; dashed arrows indicate controlled variables or validation
procedures. EFA = Exploratory Factor Analysis; CFA = Confirmatory Factor
Analysis; α = Cronbach's alpha; η²p =
partial eta squared.
With the aim of
examining whether the gender difference in attitudes toward AI persists when
controlling the effect of available academic variables, an analysis of
covariance (ANCOVA) was carried out with the total score as the dependent
variable, gender and field of study as between-subjects factors, and academic
year as a continuous covariate. Type II sum of squares was used to obtain the independent
effects of each predictor, which permits the evaluation of each variable's
contribution while controlling the others
3. Results
The results are
presented in five sections: psychometric properties of the adapted instrument,
gender differences in attitudes towards AI (central objective), differences by
area of knowledge and academic year (complementary analyses), multivariate
analysis (ANCOVA), and synthesis of the main findings.
3.1. Psychometric
properties of the adapted instrument
3.1.1. Content
validity
Content validity
analysis using expert judgment revealed adequate Aiken's V coefficient values for
the five items evaluated. All items achieved values above the
established criterion (V ≥ .70), confirming adequate clarity, coherence,
and relevance. Item 5 (reverse formulation regarding the threat of AI)
presented the lowest values (V = .75) but was considered
acceptable to continue with the exploratory factor analysis (see Table 1).
Table 1
Aiken's V coefficients by item and dimension
(n = 5 expert judges)
|
Item |
Clarity |
Coherence |
Relevance |
Mean V |
|
1 |
.88 |
.92 |
.90 |
.90 |
|
2 |
.92 |
.88 |
.92 |
.91 |
|
3 |
.84 |
.90 |
.88 |
.87 |
|
4 |
.90 |
.92 |
.94 |
.92 |
|
5 |
.76 |
.72 |
.78 |
.75 |
Note. All values exceed the
recommended cutoff point (V ≥ .70). Item 5 corresponds to "AI is a
threat to humans" (reverse formulation).
3.1.2. Exploratory
Factor Analysis
Prior to Exploratory
Factor Analysis (EFA), the factorability assumptions were verified. Bartlett's test
of sphericity was statistically significant (χ² =
3847.65, df = 10, p < .001), and the Kaiser-Meyer-Olkin (KMO) measure of
sampling adequacy reached a value of .819. The principal axis extraction method
with oblimin rotation identified a single factor that explained 64.16% of the
total variance, significantly higher than the second factor (13.64%),
confirming the unidimensional structure. The first four items had factor
loadings greater than .80, while item 5 showed a moderate negative factor
loading (−.453) and low communality (.205), replicating the behavior
reported in the original validation by Grassini (2023), and was therefore
excluded from the final score.
Table 2 confirms the
suitability of the correlation matrix for factor analysis. The KMO value of
.819 indicates a satisfactory partial correlation between the items, exceeding
the minimum recommended threshold of .70 (Kaiser, 1974). Bartlett's test was significant,
which discards the hypothesis of an identity matrix and validates the appropriateness
of the analysis.
Table 2
Matrix of polychoric
correlations between items
|
|
Item
1 |
Item
2 |
Item
3 |
Item
4 |
Item
5 |
|
Item
1 |
1.000 |
— |
— |
— |
— |
|
Item
2 |
.724** |
1.000 |
— |
— |
— |
|
Item
3 |
.685** |
.698** |
1.000 |
— |
— |
|
Item
4 |
.708** |
.739** |
.712** |
1.000 |
— |
|
Item
5 |
−.342** |
−.368** |
−.329** |
−.385** |
1.000 |
Note. **p
< .001. Item 5 shows negative correlations due to its inverse formulation.
Table 3 presents the
factor loadings and communalities, where item 5 showed inverse behavior consistent
with the original validation (Grassini, 2023).
Table 3
Eigenvalues and explained
variance in the EFA
|
Factor |
Eigenvalue |
%
Variance |
%
Cumulative |
|
1 |
3.208 |
64.16% |
64.16% |
|
2 |
0.682 |
13.64% |
77.80% |
|
3 |
0.421 |
8.42% |
86.22% |
|
4 |
0.368 |
7.36% |
93.58% |
|
5 |
0.321 |
6.42% |
100.00% |
Note. The first factor explains
more than four times the variance of the second factor, confirming the
one-dimensional structure.
Table 4 details the explained
variance, confirming that the first factor predominates significantly over the
remaining ones.
Table 4
Factor loadings and communalities of the
EFA
|
Item |
Content |
Loading |
h² |
Uniqueness |
|
1 |
AI
will improve my life |
.856 |
.733 |
.267 |
|
2 |
AI
will improve my work |
.879 |
.773 |
.227 |
|
3 |
I
will use AI technology in the future |
.831 |
.691 |
.309 |
|
4 |
AI
is positive for mankind |
.862 |
.743 |
.257 |
|
5 |
AI
is a threat to human beings |
−.453 |
.205 |
.795 |
Note.
Extraction method: Principal axes. h² = communality; Uniqueness = 1 − h².
Factor loadings > .70 are considered excellent (Hair et al., 2019).
3.1.3.
Confirmatory Factor Analysis
Confirmatory Factor
Analysis (CFA) was performed, using the MLR estimator on 40.6% of the random
sample (n = 776). The indices of the four-item unidimensional model were
outstanding; the CFI = .991 and the TLI = .973 exceed the .95 threshold (Hu
& Bentler, 1999), The SRMR of .018 is below the criterion of .08, and
the RMSEA of .076 reached the acceptable fit zone, considering that the degrees
of freedom are reduced (Kenny
et al., 2015). All standardized factor loadings were significant (p
< .001) and greater than .79.
The χ²/df
ratio was high (6.42), an expected result given the large sample size, which
makes this statistic sensitive to trivial deviations. The CFI, TLI, and SRMR
indices, which are more robust to this effect, confirm the adequate fit of the
model (see Table 5).
Table 5
Goodness-of-fit indices
of the CFA model
|
Index |
Obtained
value |
Recommended
value |
Interpretation |
|
χ² |
12.84 |
— |
— |
|
gl |
2 |
— |
— |
|
p |
.002 |
>
.05 |
Significant |
|
χ²/gl |
6.42 |
<
3.00 |
Elevated
(sensitive to N) |
|
RMSEA |
.076 |
<
.08 |
Acceptable |
|
IC
90% RMSEA |
[.038,
.120] |
límite
sup. < .10 |
Acceptable |
|
CFI |
.991 |
>
.95 |
Excellent |
|
TLI |
.973 |
>
.95 |
Excellent |
|
SRMR |
.018 |
<
.08 |
Excellent |
|
PCLOSE |
.070 |
>
.05 |
Marginal |
Note. df = degrees of freedom;
RMSEA = root mean square error of approximation; 90% CI = 90% confidence
interval; CFI = comparative fit index; TLI = Tucker-Lewis index; SRMR =
standardized root mean square residual; PCLOSE = test of tight fit of RMSEA.
Criteria according to Hu & Bentler (1999), Browne & Cudeck (1992) and
Kline (2015).
Table 6 presents the
standardized CFA loads, all significant (p < .001), converging with the EFA
results.
Table 6
Standardized factor loadings of the CFA
model
|
Item |
Content |
λ |
SE |
z |
p |
R² |
|
1 |
AI
will improve my life |
.842 |
.019 |
44.32 |
<
.001 |
.709 |
|
2 |
AI
will improve my work |
.867 |
.017 |
50.99 |
<
.001 |
.752 |
|
3 |
I
will use AI in the future |
.798 |
.022 |
36.27 |
<
.001 |
.637 |
|
4 |
AI
is positive for mankind |
.851 |
.018 |
47.28 |
<
.001 |
.724 |
Note. λ
= standardized factor loading; SE = standard error; z = z-statistic; R² =
variance explained by the latent factor.
3.1.4. Reliability
The reliability of the instrument
was evaluated using Cronbach's alpha and McDonald's ω. Both
coefficients exceeded the recommended criterion of .70, reaching values of
.884 and .886, respectively. The Average Variance Extracted (AVE = .711)
exceeded the threshold of .50, demonstrating convergent validity (Fornell
& Larcker, 1981). The corrected item-total correlations ranged from
.726 to .798, with α ranging from .837 to .865 if any item is removed,
confirming the adequate internal consistency of the instrument (see Table 7).
Table 7
Reliability indicators of the adapted
AIAS-4 scale
|
Indicator |
Value |
Interpretation |
|
α
Cronbach |
.884 |
Good |
|
α
ordinal |
.891 |
Good |
|
ω
McDonald |
.886 |
Good |
|
Average Variance Extracted (AVE) |
.711 |
Adequate
(> .50) |
Note.
The ordinal α is calculated using polychoric correlations, appropriate for
ordinal items. The AVE represents the average proportion of variance explained
by the factor.
3.2. Gender differences
in attitudes towards artificial intelligence
Once the psychometric
properties of the instrument were confirmed, the analysis of gender
differences, the central objective of this study, was carried out. The groups
were relatively balanced in size, with a slight majority of women (52.7%). The average
age was similar in both groups, with no statistically significant differences
(t = 0.83, p = .407).
The results show that
women obtained higher average scores (M = 3.15, SD = 0.81) compared to men (M =
2.97, SD = 0.97), indicating significantly more positive attitudes toward AI in
the female group. Men showed greater dispersion in their responses (SD = 0.97
vs. 0.81), suggesting greater heterogeneity in attitudes within this group (see
Table 9).
Table 8
Descriptive characteristics
of the sample by gender
|
Variable |
Men |
Women |
Total |
|
Age M (SD) |
21.68
(2.19) |
21.83
(2.26) |
21.76
(2.23) |
|
Area
of knowledge |
|||
|
Area
1 n (%) |
129
(14.3%) |
244
(24.2%) |
373
(19.5%) |
|
Area
2 n (%) |
579
(64.0%) |
418
(41.5%) |
997
(52.1%) |
|
Area
3 n (%) |
197
(21.8%) |
345
(34.3%) |
542
(28.4%) |
|
Academic
year |
|||
|
Year
1-2 n (%) |
193
(21.3%) |
247
(24.5%) |
440
(23.0%) |
|
Year
3-4 n (%) |
295
(32.6%) |
340
(33.8%) |
635
(33.2%) |
|
Year
5-6 n (%) |
417
(46.1%) |
420
(41.7%) |
837
(43.8%) |
Note.
M = Mean; SD = Standard deviation.
Levene's test indicated
heterogeneity of variances between groups (F = 15.24, p < .001), so the
statistics adjusted for unequal variances (Welch's correction) are reported.
The results revealed statistically significant differences (t = −4.48, df
= 1714.89, p < .001), with a small-to-medium effect size according to
Cohen's criterion (d = 0.204), with a 95% confidence interval: [−0.26, −0.10]
(see Table 10). These results indicate that female students have significantly
more positive attitudes toward artificial intelligence than male students,
contradicting the patterns reported in the international literature where men
typically show more favorable attitudes toward emerging technologies.
Table 9
Descriptive statistics of attitudes
towards AI by gender
|
Género |
N |
M |
DE |
Mín |
Máx |
IC
95% |
|
Men |
905 |
2.97 |
0.97 |
1.00 |
5.00 |
[2.91,
3.04] |
|
Women |
1,007 |
3.15 |
0.81 |
1.00 |
5.00 |
[3.10,
3.20] |
|
Total |
1,912 |
3.07 |
0.89 |
1.00 |
5.00 |
[3.03,
3.11] |
Note.
M = Mean; SD = Standard Deviation; Min = Minimum; Max = Maximum; 95% CI = 95%
Confidence Interval.
Table 10 reports the
inferential statistics, with Welch's correction for heterogeneity of variances
(F = 15.24, p < .001).
Table 10
Student's t-test for gender differences in
attitudes towards AI
|
Variable |
Levene |
T |
df |
p |
Mean Diff. |
|
Attitudes towards AI |
F =
15.24*** |
−4.48 |
1714.89 |
< .001 |
−0.18 |
|
95% CI of the difference: [−0.26, −0.10] |
|||||
Note. df = degrees of freedom;
Diff. = Difference; Cohen's d = effect size; 95% CI = 95% confidence interval.
***p < .001.
3.3. Differences by
area of knowledge
The analysis of
variance (ANOVA) revealed statistically significant differences between the
areas (F (2, 1909) = 4.28, p = .014, η² = .004). However, the effect size was very small (η² <
.01), suggesting that the knowledge area explains less than 1% of the
variability in attitudes toward AI. Tukey's post hoc tests indicated that significant
differences were found only between Area 2 and Area 3 (p = .012).
3.4. Differences by
academic year
The analysis of
variance (ANOVA) confirmed statistically significant differences between
academic years (F (5, 1906) = 4.92, p < .001, η² =
.013). Tukey's post hoc tests revealed that students from fifth and sixth year
presented significantly fewer positive attitudes than first through fourth year
students (p < .05), suggesting a descending trend in positive attitudes as academic
training progresses.
Taken together, results
position gender as the most consistent predictor of attitudes toward AI in this
sample, opposite to what has been reported in international literature. The
field of study and academic year contribute with additional explanations, though
of negligible practical magnitude. The fact that attitudes are relatively
homogeneous across disciplines suggests that the perception of AI depends less on
the field of study than on other factors that lie beyond the scope of this
study.
3.5. Multivariate analysis: ANCOVA
To assess whether
gender differences persisted when simultaneously controlling available academic
variables, an ANCOVA model was estimated with the total AIAS-4 score as the
dependent variable, gender and field of study as between-subjects factors, and
academic year as a continuous covariate. The results confirmed that the effect
of gender persisted after adjustment for both variables: F(1, 1905) = 15.67, p
< .001, η²p = .008, with adjusted means of M = 3.13 for women
and M = 2.96 for men. Academic year showed a significant independent effect
(F(1, 1905) = 5.18, p = .023, η²p = .003). The area of knowledge,
however, was not significant when controlling for academic year (F(2, 1905) =
0.51, p = .603, η²p < .001), suggesting that the differences between
areas observed in the previous univariate analysis could be attributable, at least
in part, to the unequal distribution of students by academic year across
programs. The model explains 1.5% of the total variance (R² = .015) (see Table
11).
Table 11
ANCOVA Results:
Attitudes towards AI by gender, controlling area of knowledge and
academic year
|
Source |
SS |
df |
F |
p |
η²p |
|
Gender |
12.26 |
1 |
15.67 |
< .001 |
0.008 |
|
Area of knowledge |
0.79 |
2 |
0.51 |
0.603 |
< .001 |
|
Academic year |
4.05 |
1 |
5.18 |
0.023 |
0.003 |
|
Residual |
1490.91 |
1905 |
— |
— |
— |
Note. SS = Type II sum of squares; df = degrees of
freedom; η²p = partial eta
squared. Adjusted mean for women: M = 3.13; adjusted mean for men: M = 2.96.
Model R² = .015.
4. Discussion and
conclusions
This study analyzed
gender differences in attitudes toward AI among 1,912 university students in
Arequipa, Peru. Women scored significantly higher than men on the adapted
AIAS-4 scale (M = 3.15 vs. M = 2.97; p < .001; d = 0.204), a difference that
persisted even after controlling the field of study and academic year (F(1,
1905) = 15.67; p < .001; η²p = .008). Further analyses revealed a descending
trend in positive attitudes as students progressed through their academic
training. The following section discusses the implications of these findings.
4.1. The paradox of
positive attitudes and persistent underrepresentation
Before interpreting the
direction of the finding, it is convenient to situate its magnitude. The
difference between groups is statistically robust - replicated in both the bivariate
and multivariate analyses - but the effect size is small (d = 0.20; η²p =
.008 in the ANCOVA). This means that the gender of the participants explains
less than 1% of the variance in attitudes toward AI, which reflects
considerable individual heterogeneity within each group. In practical terms,
the distributions for men and women overlap considerably: the 0.18-point
difference on a five-point scale does not characterize women as a homogeneous group
in favor of AI or men as a homogeneously skeptical one. What the data show is a
statistically distinguishable central tendency, not a categorical difference
between gender
A small effect is not
necessarily an irrelevant effect: in large samples, such as this one, modest
differences can be statistically significant and, if consistent, theoretically
informative (Ellis, 2010; Ferguson, 2009). The practical relevance of this finding
does not lie in the magnitude of the attitude gap, but in its direction: the
fact that women show equal or more favorable attitudes toward AI than men
refutes the hypothesis that the underrepresentation of women in the technology
field is due to a motivational or attitudinal deficit. That is the study's
substantive contribution.
The finding acquires
even more significance when it is contrasted with the documented
underrepresentation of women in the development, research, and governance of artificial
intelligence systems. Globally, only one out of three scientists is a woman, a
proportion that decreases even further at the higher levels of scientific
careers and in the hierarchies of the academic and technological world (UNESCO,
2024). In the specific context of artificial intelligence,
women represent a minority in development teams, scientific publications of
high impact, and leadership positions in technology companies (Carvajal
et al., 2025).
This paradox demands an
interpretation that transcends individual psychological explanations and that
recognizes the role of social, economic, and institutional structures in the
configuration of educational and professional trajectories. Positive attitudes,
while necessary, are not a sufficient condition to guarantee women's equitable
participation in technological fields. As critical literature on gender and
technology points out, the "digital gender gap" is not explained by attitudinal
or motivational deficits in women, but by systemic barriers that operate at
multiple levels: educational, professional, cultural, and symbolic (Sáinz
et al., 2016; Leavy,
2018; Peláez-Sánchez
et al., 2023; Singh et al., 2025).
4.2. Divergences with
international literature and contextual specificities
The findings of this
study contrast with previous research carried out in European and Asian
contexts, where it is typically reported that men have more positive attitudes
toward AI than women. For example, Møgelvang
et al., (2024) documented in Norway that men use generative AI chatbots
more frequently and show more interest in these tools than their female
counterparts. Similarly, Russo
et al., (2025) found in Italy that women report higher levels of
anxiety regarding AI, which is associated with less positive attitudes toward
technology. Both findings point in the opposite direction to what was observed
in the present research, where women from Arequipa obtain significantly higher
scores on the AIAS-4 scale. This reversal of the pattern is not insignificant: it
suggests that the conditions under which men and women interact with AI differ
substantially across contexts, and that extrapolating European findings to the
Latin American case without critical analysis leads to distorted
interpretations.
These discrepancies
between geographical contexts suggest that attitudes toward technology do not
follow a universal pattern, but they are mediated by the cultural, educational,
and socioeconomic conditions related to each region. In the Latin American case,
and particularly in Peru, a plausible hypothesis - though not verifiable with
the current design - is that female university students value artificial
intelligence as a low-cost resource with the potential to compensate for
inequalities in the access to academic tools. By offering functionalities
previously reserved for contexts with greater educational capital, Generative
AI can generate perceptions different from those documented in European or Asian
environments with different starting points. Establishing whether this truly
occurs would, however, require studies that incorporate direct measures of the
reasons for use and the meanings attributed to these tools.
4.3. Structural
explanations of the underrepresentation of women in AI
The paradox identified
in this study - favorable attitudes without equitable participation—demands an
analysis of the structural barriers that operate beyond individual dispositions.
The literature identifies at least five interrelated mechanisms that perpetuate
the underrepresentation of women in technological fields: (1) gender
stereotypes in STEM that associate technical competence with masculine
attributes (Tellhed
et al., 2023); (2) the scarcity of female role models in visible
positions in the field of AI (World
Economic Forum, 2023); (3) differences in technical self-efficacy resulting
from gender socialization (Yau
& Cheng, 2012); (4) hiring practices that reproduce gender biases in
the technology sector (Alam,
2022); and (5) AI systems themselves can perpetuate and
amplify existing gender biases through algorithms trained on biased data (Li
et al., 2023).
4.4. Implications for
education and gender equality policies
The findings of this
study have a direct implication for the design of educational policies: if
women already exhibit favorable attitudes toward AI, interventions should not
be aimed at modifying these attitudes but rather at removing the obstacles that
prevent them from translating into concrete career paths. This shifts the focus
from raising awareness - which assumes an attitudinal deficit that the data
does not confirm - to the transformation of structural conditions.
In practical terms,
universities can take action on at least two fronts. The first is curricular:
incorporating AI training with a gender perspective not as additional content
but as a cross-cutting theme that critically examines who designs the systems,
what biases they introduce, and who is excluded from their governance. The
second is institutional: creating conditions that reduce the gap between
attitude and participation through mentorship programs with female role models
in the technology field, equitable access to computing resources, and explicit
recognition of women's work in AI within academic spaces. None of these
measures operates in isolation; their effectiveness depends on their
coordinated implementation and systematic monitoring of results.
4.5. Additional
findings and limitations
Further analyses
revealed that the area of knowledge explains a minimal proportion
of the variance in attitudes toward AI (η² =
.004), suggesting that perceptions of this technology transcend specific
academic disciplines. The downward trend in positive attitudes as academic
years increase (η² = .013) deserves attention and can be interpreted as
an effect of critical maturity or disillusionment with the gap between
technological promises and the realities of the Peruvian labor market.
The present study has
four limitations that should be considered when interpreting the results. The
most significant is its cross-sectional design. This type of design allows the
identification of differences between groups at a specific point in time, but
it does not establish causal relationships or allow tracking on how attitudes evolve
over time. Consequently, the interpretations proposed in the discussion
regarding the factors that could explain gender differences are plausible but
unverified. Answering that question requires longitudinal or experimental
designs, which this study cannot provide.
A second limitation
concerns the comparative approach. Bivariate and multivariate analyses detect
differences between groups defined by gender, field of study, and academic year,
but they do not capture what happens within each group or the interactions
between variables. The fact that gender explains less than 1% of the total
variance is revealing: the unmeasured factors - technological self-efficacy,
previous digital access history, experiences of discrimination in academic
spaces - have considerably more weight than the included demographic variables.
Incorporating these factors as mediators or moderators remains an agenda for
future research.
The instrument represents
a third limitation. The AIAS-4 measures general attitudes toward AI without
distinguishing between more specific dimensions, such as awareness of
algorithmic biases or desire to participate professionally in the field. An
instrument that distinguishes between these dimensions could produce a profile
of gender differences, different from the one reported here. Finally, the lack
of qualitative data leaves unanswered questions about the meaning that participants
attribute to their own answers, which would have allowed us to interpret the
observed quantitative patterns in a deeper way.
To sum up, this study
documents that Peruvian female university students exhibit more positive
attitudes toward AI than their fellow male students, a finding that contrasts
with predominant patterns in the international literature and deserves
analytical attention precisely for that reason. The paradox is significant: favorable
dispositions toward technology coexist with a persistent underrepresentation of
women in the field. This indicates that attitudes, although necessary, are
insufficient to transform career paths when structural conditions are not in
place. Universities, educational policy organizations, and the technology
sector face a concrete question: if women already value AI, what prevents them
from participating on equal terms in its design, development, and governance?
Answering this question requires longitudinal research, but also educational
policy decisions that cannot be indefinitely subordinated to the accumulation
of empirical evidence.
Authors‘ contributions
Conceptualization: Author1 and Author4;
methodology: Author1 and Author2; formal analysis: Author1 and Author3;
investigation: Author2, Author3 and Author4; writing—original draft: Author1;
writing—review and editing: Author1, Author2, Author3 and Author4; supervision:
Author1.
Funding
This research did not receive specific external
funding.
Ethical concerns
The study was approved by the institution's
academic authorities and adhered to the ethical principles of the Declaration
of Helsinki. All participants provided voluntary informed consent.
Conflict of interest
The authors declare that there is no conflict
of interest.
Artificial intelligence
usage statement
During the preparation of this manuscript, the
authors used generative artificial intelligence tools (Claude, Anthropic) to assist
in improving writing style and textual coherence. All content was reviewed,
verified, and validated by the authors, who assume full responsibility for the
scientific integrity of the study.
Data availability statement
The data supporting the
findings of this study are available upon reasonable request to the
corresponding author.
References
Aiken,
L. R. (1980). Content Validity and Reliability of Single Items or Questionnaires.
Educational and Psychological Measurement, 40(4), 955-959. https://doi.org/10.1177/001316448004000419
Alam, A. (2022). Psychological,
Sociocultural, and Biological Elucidations for Gender Gap in STEM Education: A
Call for Translation of Research into Evidence-Based Interventions. 95-107.
https://doi.org/10.2991/ahsseh.k.220105.012
Ancheta-Arrabal, A., Pulido-Montes,
C., & Carvajal-Mardones, V. (2021). Gender
Digital Divide and Education in Latin America: A Literature Review. Education Sciences, 11(12). https://doi.org/10.3390/educsci11120804
Ato, M., López-García,
J. J., & Benavente, A. (2013). Un sistema de clasificación de los diseños
de investigación en psicología. Anales de Psicología / Annals of Psychology,
29(3), 1038-1059.
https://doi.org/10.6018/analesps.29.3.178511
Aviles-Valenzuela, A.,
Acosta-Barreno, K., Espinel-Obregoso, F. P., & Carrasco, A. S. E. (2025). Trends
and Analysis of Artificial Intelligence Research in Latin America (2013–2023). Open
Information Science, 9(1). https://doi.org/10.1515/opis-2025-0023
Beig, S., & Qasim, S. H. (2023). Assessing
Students’ Attitude Towards Artificial Intelligence with Respect to Gender and
Use of Computer and Mobile Devices. IJFMR - International Journal For
Multidisciplinary Research, 5(3). https://doi.org/10.36948/ijfmr.2023.v05i03.4130
Bittle, K., & El-Gayar, O. (2025).
Generative AI and Academic Integrity in Higher Education: A
Systematic Review and Research Agenda. Information, 16(4). https://doi.org/10.3390/info16040296
Bond, M., Khosravi, H., De Laat, M., Bergdahl,
N., Negrea, V., Oxley, E., Pham, P., Chong, S. W., & Siemens, G. (2024). A
meta systematic review of artificial intelligence in higher education: A call
for increased ethics, collaboration, and rigour. International Journal of
Educational Technology in Higher Education, 21(1), 4. https://doi.org/10.1186/s41239-023-00436-z
Browne, M. W., & Cudeck, R. (1992). Alternative ways of assessing model fit. Sociological Methods & Research, 21(2), 230–258. https://doi.org/10.1177/0049124192021002005
Cai, Z., Fan, X., & Du, J. (2017). Gender
and attitudes toward technology use: A meta-analysis. Computers &
Education, 105, 1-13. https://doi.org/10.1016/j.compedu.2016.11.003
Carvajal, D., Franco, C., & Isaksson, S.
(2025). Will Artificial Intelligence Get in the Way of Achieving Gender
Equality? Discussion Paper Series in Economics, Discussion Paper Series in Economics,
Article 3/2024. https://ideas.repec.org//p/hhs/nhheco/2024_003.html
Chiu, T. K. F. (2024). Future research
recommendations for transforming higher education with generative AI. Computers
and Education: Artificial Intelligence, 6, 100197.
https://doi.org/10.1016/j.caeai.2023.100197
Cohen, J. (2013). Statistical Power Analysis
for the Behavioral Sciences (2.a ed.). Routledge. https://doi.org/10.4324/9780203771587
Contreras, F. C., &
Olaya, J. (2025). La inteligencia artificial en la educación superior peruana:
Tendencias y desafíos. Revista Tribunal, 5(13), 39-61.
https://doi.org/10.59659/revistatribunal.v5i13.254
Ellis, P. D. (2010). The
essential guide to effect sizes: Statistical power, meta-analysis, and the
interpretation of research results. Cambridge University Press. https://doi.org/10.1017/CBO9780511761676
Fernández-Miranda, M., Román-Acosta,
D., Jurado-Rosas, A. A., Limón-Dominguez, D., Torres-Fernández, C.,
Fernández-Miranda, M., Román-Acosta, D., Jurado-Rosas, A. A., Limón-Dominguez,
D., & Torres-Fernández, C. (2024). Artificial
Intelligence in Latin American Universities: Emerging Challenges. Computación
y Sistemas, 28(2), 435-450. https://doi.org/10.13053/cys-28-2-4822
Ferguson, C. J. (2009). An effect size primer:
A guide for clinicians and researchers. Professional Psychology: Research
and Practice, 40(5), 532–538. https://doi.org/10.1037/a0015808
Field, A. (2018). Discovering statistics
using IBM SPSS statistics (5th ed.). SAGE.
Fornell, C., & Larcker, D. F. (1981).
Evaluating structural equation models with unobservable variables and
measurement error. Journal of Marketing Research, 18(1), 39-50. https://doi.org/10.2307/3151312
Frumin, I., Vorochkov, A., Kiryushina, M., Platonova, D., & Terentiev, E. (2026). Mapping the Generative AI Research in Higher Education: 2022–2024 Insights. Higher Education Quarterly, 80(1), e70075. https://doi.org/10.1111/hequ.70075
García-López, I. M., &
Trujillo-Liñán, L. (2025). Ethical and regulatory
challenges of Generative AI in education: A systematic review. Frontiers in
Education, 10. https://doi.org/10.3389/feduc.2025.1565938
Grassini, S. (2023). Development and validation
of the AI attitude scale (AIAS-4): A brief measure of general attitude toward artificial
intelligence. Frontiers in Psychology, 14. https://doi.org/10.3389/fpsyg.2023.1191628
Hair, J. F., Black, W. C., Babin, B. J., &
Anderson, R. E. (2019). Multivariate data analysis (8th ed.). Cengage.
Hajam, K. B., & Gahir, S. (2024). Unveiling
the Attitudes of University Students Toward Artificial Intelligence. Journal
of Educational Technology Systems, 52(3), 335-345. https://doi.org/10.1177/00472395231225920
Heredia Pérez, G., Benavides Galvez, J. B.,
Rojas Campos, E., & Sanchez Bustamante, E. (2025). Inteligencia artificial y pedagogía: Retos para la
educación superior en el Perú. https://doi.org/10.5281/zenodo.16997148
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1-55. https://doi.org/10.1080/10705519909540118
Jovanović, M., & Campbell, M. (2022). Generative
Artificial Intelligence: Trends and Prospects. Computer, 55(10), 107-112. https://doi.org/10.1109/MC.2022.3192720
Kaiser, H. F. (1974). An index of factorial
simplicity. Psychometrika, 39(1), 31–36. https://doi.org/10.1007/BF02291575
Katsantonis, A.,
& Katsantonis, I. G. (2024). University Students’ Attitudes
toward Artificial Intelligence: An Exploratory Study of the Cognitive,
Emotional, and Behavioural Dimensions of AI Attitudes. Education Sciences,
14(9). https://doi.org/10.3390/educsci14090988
Kenny, D. A., Kaniskan, B., & McCoach, D.
B. (2015). The Performance of RMSEA in Models With Small Degrees of Freedom. Sociological
Methods & Research, 44(3), 486-507. https://doi.org/10.1177/0049124114543236
Kline, R. B. (2015). Principles and practice
of structural equation modeling (4th ed.). Guilford Press.
Kofinas, A. K., Tsay, C. H.-H., & Pike, D.
(2025). The impact of generative AI on academic integrity of authentic
assessments within a higher education context. British Journal of
Educational Technology, 56(6), 2522-2549. https://doi.org/10.1111/bjet.13585
Leavy, S. (2018). Gender bias in artificial
intelligence: The need for diversity and gender theory in machine learning. Proceedings
of the 1st International Workshop on Gender Equality in Software Engineering,
GE ’18, 14-16. https://doi.org/10.1145/3195570.3195580
Li, B., Qi, P., Liu, B.,
Di, S., Liu, J., Pei, J., Yi, J., & Zhou, B. (2023). Trustworthy
AI: From Principles to Practices. ACM Comput. Surv., 55(9),
177:1-177:46. https://doi.org/10.1145/3555803
Li, B., Tan, Y. L., Wang, C., & Lowell, V. (2025). Two years of innovation: A systematic review of empirical generative AI research in language learning and teaching. Computers and Education: Artificial Intelligence, 9, 100445. https://doi.org/10.1016/j.caeai.2025.100445
Lloret-Segura, S.,
Ferreres-Traver, A., Hernández-Baeza, A., & Tomás-Marco, I. (2014). El
Análisis Factorial Exploratorio de los Ítems: Una guía práctica, revisada y
actualizada. Anales de Psicología, 30(3), 1151-1169. https://doi.org/10.6018/analesps.30.3.199361
Merino-Soto, C. (2023). Coeficientes V de Aiken: Diferencias en los juicios de validez de contenido. MHSalud, 20(1), 1-14.
Møgelvang, A., Bjelland, C., Grassini, S.,
& Ludvigsen, K. (2024). Gender Differences in the Use of Generative
Artificial Intelligence Chatbots in Higher Education: Characteristics and Consequences.
Education Sciences, 14(12). https://doi.org/10.3390/educsci14121363
Ofosu-Ampong, K. (2024). Artificial
intelligence research: A review on dominant themes, methods, frameworks and
future research directions. Telematics and Informatics Reports, 14,
100127. https://doi.org/10.1016/j.teler.2024.100127
Pallant, J. (2020). SPSS survival manual (7th ed.). McGraw-Hill.
Peláez-Sánchez, I. C., George Reyes, C. E.,
& Glasserman-Morales, L. D. (2023). Gender digital divide in education 4.0:
A systematic literature review of factors and strategies for inclusion. Future
in Educational Research, 1(2), 129-146. https://doi.org/10.1002/fer3.16
Peres, R., Schreier, M., Schweidel, D., &
Sorescu, A. (2023). On ChatGPT and beyond: How generative artificial
intelligence may affect research, teaching, and practice. International
Journal of Research in Marketing, 40(2), 269-275. https://doi.org/10.1016/j.ijresmar.2023.03.001
Purificato, E., Lorenzo,
F., Fallucchi, F., & De Luca, E. W. (2023). The
Use of Responsible Artificial Intelligence Techniques in the Context of Loan
Approval Processes. International Journal of Human–Computer Interaction,
39(7), 1543-1562. https://doi.org/10.1080/10447318.2022.2081284
Russo, C., Romano, L., Clemente, D., Iacovone,
L., Gladwin, T. E., & Panno, A. (2025). Gender differences in artificial
intelligence: The role of artificial intelligence anxiety. Frontiers in
Psychology, 16. https://doi.org/10.3389/fpsyg.2025.1559457
Sáinz, M., Meneses, J., López, B.-S., &
Fàbregues, S. (2016). Gender Stereotypes and Attitudes Towards Information and
Communication Technology Professionals in a Sample of Spanish Secondary
Students. Sex Roles, 74(3), 154-168. https://doi.org/10.1007/s11199-014-0424-2
Salas-Pilco, S. Z., & Yang, Y. (2022). Artificial
intelligence applications in Latin American higher education: A systematic
review. International Journal of Educational Technology in Higher Education,
19(1), 21. https://doi.org/10.1186/s41239-022-00326-w
Singh, S., Rahul, K., Paliwal, M., Wani, I. A.,
& Suri, S. (2025). Gendering the digital divide: A systematic review
of women’s digital inclusion challenges and emerging research directions. Digital
Transformation and Society, 4(4), 503-531. https://doi.org/10.1108/DTS-04-2025-0083
Stöhr, C., Ou, A. W., & Malmström, H.
(2024). Perceptions and usage of AI chatbots among students in higher education
across genders, academic levels and fields of study. Computers and
Education: Artificial Intelligence, 7, 100259. https://doi.org/10.1016/j.caeai.2024.100259
Sultana, A., Abdul Latheef, N., Siby, N., &
Ahmad, Z. (2025). Exploring Students’ Attitudes Toward Artificial Intelligence
(AI): Psychometric Validation of AI-Attitude Scale. Sage Open, 15(4),
21582440251378375. https://doi.org/10.1177/21582440251378375
Tapullima Mori, C., Mamani Benito, O. J., Turpo Chaparro, J., Olivas Ugarte, L. O., & Carranza Esteban, R. F. (2024). Inteligencia artificial en la educación universitaria: Revisión bibliométrica en Scopus y Web of Science. Revista Electrónica Educare, 28(Extra 1), 1.
Tellhed, U., Björklund, F., & Kallio
Strand, K. (2023). Tech-Savvy Men and Caring Women: Middle School Students’
Gender Stereotypes Predict Interest in Tech-Education. Sex Roles, 88(7),
307-325. https://doi.org/10.1007/s11199-023-01353-1
The jamovi project. (2024). jamovi (Version
2.3) [Computer software]. https://www.jamovi.org
UNESCO Biblioteca Digital. (2024). UNESCO
Call to Action: Closing the gender gap in science. bit.ly/3PoZ2zy
Valdivieso, T., & González, O. (2025).
Generative AI Tools in Salvadoran Higher Education: Balancing Equity, Ethics, and
Knowledge Management in the Global South. Education Sciences, 15(2).
https://doi.org/10.3390/educsci15020214
World Economic Forum. (2023). The global risks report 2023 (18th ed). World Economic Forum. bit.ly/4rg9Ux2
Yau, H. K., & Cheng, A. L. F. (2012).
Gender Difference of Confidence in Using Technology for Learning. Journal of
Technology Studies, 38(2). https://doi.org/10.21061/jots.v38i2.a.2