RIAF Journal ISSN: 2953-6693 Vol – 4 No. 2, July 2026
153
Articles
Jorge Andrés Bayona Martínez
Universidad Estatal de Milagro, Ecuador
jbayonam@unemi.edu.ec
ORCID https://orcid.org/0009-0005-9332-5764
Edwin Esneider Andrade Perdomo
Universidad estatal de Milagro, Ecuador
eandradep7@unemi.edu.ec
ORCID https://orcid.org/0009-0001-1200-9593
Steven Arturo Torres Burgos
Universidad estatal de Milagro, Ecuador
storresb5@unemi.edu.ec
ORCID https://orcid.org/0000-0001-9299-3254
Abstract: Introduction: Physical inactivity and
emotional well-being difficulties are growing
problems in the school population, whereas
artificial-intelligence-based technologies offer new
pedagogical possibilities. Objective: to assess the
effect of a physical education program mediated by
pedagogical possibilities. Objective: to assess the
effect of a physical education program mediated by
artificial intelligence on the physical fitness and
emotional well-being of basic-education students.
Materials and methods: a quasi-experimental
design with two groups and pre- and post-
measurements was used over twelve weeks. Two
hundred and forty primary and secondary students
participated and were assigned to an experimental
group and a control group. Cardiorespiratory
endurance was assessed with a twelve-minute run
test, muscular endurance with elbow push-ups in
thirty seconds, speed over fifty meters, and an
emotional well-being index. Analyses of
covariance, mixed analyses of variance, effect sizes
with confidence intervals and power analysis were
applied. Results: the experimental group showed
significantly greater improvements than the
control group in all variables, with significant
group-by-time interactions and moderate-to-large
effect sizes. Gains in physical fitness were
positively associated with improvements in
emotional well-being. Conclusions: the results,
derived from simulated data for methodological
purposes, suggest that artificial intelligence, used as
a pedagogical tool, could simultaneously favor
physical fitness and emotional well-being, although
studies with empirical data are required to confirm
these trends.
Keywords: Artificial intelligence; physical fitness;
emotional well-being; basic education; educational
technology.
Artificial intelligence as a pedagogical tool for physical fitness and emotional
well-being in basic education
Jorge Andrés Bayona Martínez
1
; Edwin Esneider Andrade Perdomo
2
& Steven Arturo Torres Burgos
3
RIAF. International Journal of Physical Activity
Universidad de Guayaquil, Ecuador
Frequency: Semi-annual
Vol. 4, N° 2, 2026
revista.riaf@ug.edu.ec
Received: June 5th, 2026
Approved: July 6
th
, 2026
Published: July 25
th
, 2026
URL: https://revistas.ug.edu.ec/index.php/riaf
DOI: https://doi.org/10.53591/riaf.v4i2.3372
The authors publishing in RIAF acknowledge and accept the
following conditions: The authors retain copyright over their work
and grant RIAF the right of first publication, under the Creative
Commons Attribution-NonCommercial-NoDerivatives 4.0
International License, which allows third parties to share the work
provided its author and first publication in this journal are indicated.
The authors retain copyright and guarantee RIAF the right to
publish the work through any channels it deems appropriate.
Authors are free to share, copy, distribute, execute, and publicly
communicate the published version of the work in RIAF, provided
they acknowledge its publication in this journal. Authors are
authorized to disseminate their work electronically once it has been
accepted for publication.
RIAF Journal ISSN: 2953-6693 Vol – 4 No. 2, July 2026
154
Introduction
Regular physical activity constitutes a recognized determinant of health during childhood and
adolescence. The World Health Organization recommends that children and adolescents aged five to
seventeen accumulate at least sixty minutes of daily moderate-to-vigorous physical activity and
incorporate muscle and bone-strengthening exercises at least three days per week (Bull et al., 2020).
Despite these recommendations, most of the school-age population worldwide does not meet the
minimum levels of physical activity, and the prevalence of inactivity tends to increase with age and is
higher among girls (Marques et al., 2020). The most recent international reviews confirm that
approximately eighty percent of adolescents are insufficiently active and that school, social, and digital
environments are crucial for reversing this trend (van Sluijs et al., 2021). The global landscape of physical
activity indicators reinforces this concern, evidencing persistently low grades in most countries (Aubert
et al., 2022).
Physical fitness, understood as the ability to perform motor tasks efficiently, integrates components
such as cardiorespiratory endurance, muscular strength-endurance, and speed, and has historically been
linked to better health indicators and, to a lesser extent, to the academic performance of schoolchildren
(Xu et al., 2023). Alongside the physical dimension, emotional well-being has gained relevance on the
educational agenda, given that adolescence is a sensitive period for the emergence of symptoms of
anxiety, stress, and low mood. Available evidence indicates that physical activity interventions produce
small to moderate improvements in various mental health indicators of children and adolescents, with
especially notable effects on stress and social competence (Fu et al., 2025). School-based controlled
studies have documented reductions in depressive symptoms and increases in life satisfaction following
multicomponent physical activity programs (Ahmed et al., 2023).
In this context, artificial intelligence has emerged as a set of technologies capable of personalizing
teaching, automating learning assessment, and providing adaptive feedback. In the field of physical
education, its usefulness has been proposed for individualizing classes, providing knowledge, evaluating
students, and guiding accompaniment processes (Lee & Lee, 2021). Several recent developments show
its application in teaching assessment, the design of motor simulation environments, and the
improvement of the overall quality of school health (Li et al., 2024; Zhang et al., 2022). Complementarily,
wearable devices with automated analytics have shown favorable, albeit heterogeneous, effects on the
daily physical activity of the school population (Casado-Robles et al., 2022; Chen et al., 2025), and active
video games have evidenced benefits on physical fitness, enjoyment, and psychological well-being in
educational contexts (Marsigliante et al., 2024; Rosi et al., 2025).
In the socio-emotional dimension, digital technologies and artificial intelligence-supported
approaches have been incorporated into social and emotional learning programs, whose efficacy on
competencies, attitudes, and school climate has robust meta-analytic support (Cipriano et al., 2023).
Likewise, the potential of conversational systems and emotional analytics applications to support the
mental health of children and adolescents has been described, always under principles of human
supervision, equity, and data protection (Dritsona et al., 2025). However, their adoption requires the
development of competencies and artificial intelligence literacy in both students and teachers, as well as
the consideration of ethical implications (Farrelly & Baker, 2023; Ng et al., 2024).
RIAF Journal ISSN: 2953-6693 Vol – 4 No. 2, July 2026
155
Despite growing interest, a gap persists in the
state of knowledge: most of the literature
addresses separately the effects of artificial
intelligence on motor learning or on emotional
well-being, without integrating both dimensions
into a single design oriented towards basic
education. Furthermore, available studies are
concentrated in higher education and high-
income contexts, and rarely articulate physical
fitness tests with emotional well-being measures
within a single pedagogical proposal. This
limitation constrains the understanding of
whether an artificial intelligence-mediated
intervention can simultaneously and coherently
favor the physical development and emotional
balance of students.
Consequently, the present work aims to
evaluate the effect of a physical education
program mediated by artificial intelligence on the
physical fitness and emotional well-being of basic
education students, and to investigate the
relationship between physical and emotional
gains. The relevance of this proposal lies in
providing a replicable methodological model that
articulates the objective assessment of physical
fitness with the measurement of emotional well-
being within the framework of educational
technological innovation.
Materials and methods
Study Design
A quasi-experimental design with two non-
equivalent groups—experimental and control—
and repeated measurements at two time points,
before and after the intervention, was adopted,
with a duration of twelve weeks. This approach
was selected for its feasibility in real school
contexts, where individual random assignment is
unfeasible and students are grouped into pre-
existing classrooms; therefore, assignment to
conditions was carried out by clusters (complete
classrooms), preserving the natural organization
of the class group and reducing contamination
between conditions. The unit of analysis was the
student, and the factorial design was mixed,
combining a between-subjects factor (group:
experimental vs. control) and a within-subjects
factor (time: pre- vs. post-measurement). To
minimize biases, the physical tests and the well-
being scale were administered by evaluators blind
to the participants' condition (blinded
assessment), the volume and frequency of
physical education classes were kept constant in
both groups, and contextual variables such as
session schedules and teacher characteristics were
controlled as much as possible. The planning and
reporting of the study conformed to
recommendations for non-randomized
investigations.
Se adoptó un diseño cuasiexperimental con
dos grupos no equivalentes —experimental y
control— y mediciones repetidas en dos
momentos, antes y después de la intervención,
con una duración de doce semanas. Esta
aproximación se seleccionó por su viabilidad en
contextos escolares reales, en los que la
asignación
Contexto, población y muestra
The reference population consisted of basic
education students from the municipality of San
Gil in the Department of Santander, Colombia.
The study was conducted in educational
institutions in that city over a twelve-week period
corresponding to a block of the 2025-2026 school
year. The sample included 240 students, evenly
distributed between an experimental group (n =
120) and a control group (n = 120), with equal
representation from grades 6 and 7 (n = 120) and
grades 8 and 9 (n = 120)—60 students per grade
level in each group—and a balanced distribution
by sex. Group formation sought initial
equivalence in educational level, age, and sex
composition, a condition that was empirically
verified and reported in the results section.
RIAF Journal ISSN: 2953-6693 Vol – 4 No. 2, July 2026
156
Inclusion criteria:
students enrolled in the
considered basic education levels, with student
assent and informed consent from the
responsible adult, and without medical
contraindication for the practice of physical
activity.
Exclusion criteria:
presence of injuries or
medical conditions that prevented the
performance of physical tests, participation in less
than eighty percent of the sessions, or absence
from any of the measurements. Sampling is
described as non-probabilistic convenience
sampling, with group assignment by clusters
(classrooms) to reduce contamination between
conditions.
Artificial Intelligence-Mediated
Intervention
The grupo experimental group developed the
physical education program supported by four
artificial intelligence-based technological
components, while the control group continued
with conventional physical education classes of
equal frequency and volume. The components
were: (1) physical activity monitoring applications
with adaptive plans and automated performance
feedback, consistent with evidence on wearable
devices in the school environment (Casado-
Robles et al., 2022; Chen et al., 2025); (2) an
educational conversational assistant oriented
towards psychoeducation and emotional self-
regulation, framed within social and emotional
learning programs (Cipriano et al., 2023; Dritsona
et al., 2025); (3) automated motor gesture analysis
using computer vision for technique correction;
and (4) active video games to promote enjoyment
and adherence to physical activity (Marsigliante et
al., 2024; Rosi et al., 2025). The pedagogical
design was grounded in the artificial intelligence
application framework in physical education
proposed by Lee and Lee (2021) and addressed
teaching competencies and artificial intelligence
literacy (Ng et al., 2024).
The application of the program was carried out
by the physical education teachers, previously
trained in the use of the tools and provided with
protocols and instructions for each component, to
ensure implementation homogeneity. Intervention
fidelity was controlled through usage logs of the
applications and attendance records, including in
the analysis only those students with participation
equal to or greater than eighty percent of the
sessions. The control group developed
conventional physical education classes with the
same class load and same curricular content, so that
the only systematic difference between conditions
was the artificial intelligence-based technological
mediation.
Variables and instruments
Cardiorespiratory endurance. The twelve-minute
run test (Cooper test) was used, recording the
distance covered in meters. From this, maximum
oxygen consumption was estimated using the
Cooper equation: maximal oxygen consumption =
(distance in meters − 504.9) / 44.73. The validity
and reliability of this test in preadolescents and
adolescents have been the subject of systematic
review, so its estimates should be interpreted with
caution (Martínez-Lemos et al., 2024); as a well-
validated field alternative, the twenty-meter shuttle
run test is also recognized (Mayorga-Vega et al.,
2015). In its application, the test is conducted on a
marked track and consists of covering the greatest
possible distance in twelve minutes at a self-
regulated pace; the total distance achieved allows
estimation of maximal oxygen consumption, a
reference indicator of aerobic capacity.
Upper body muscular endurance. The elbow
push-up test in thirty seconds was applied, counting
the number of correct repetitions. The test
quantifies upper body muscular endurance based on
the maximum number of correct repetitions
performed in thirty seconds, according to a
standardized technical criterion for movement
amplitude and cadence.
RIAF Journal ISSN: 2953-6693 Vol – 4 No. 2, July 2026
157
Speed. The time, in seconds, taken to cover
fifty meters in a straight line was measured, where
lower values indicate better performance. Time
was recorded using a stopwatch on a fifty-meter
straight-line course with a standing start, retaining
the best of two attempts; being a speed test,
shorter times express better performance.
Emotional well-being. The World Health
Organization Well-Being Index, composed of
five items and transformed to a scale from 0 to
100, was used, where higher scores reflect better
well-being; this is a brief self-report instrument
with adequate psychometric properties (Topp et
al., 2015). The index comprises five positively
connoted statements referring to the last two
weeks, rated on an ordinal scale; the raw score is
transformed to a range of 0 to 100, where higher
values reflect a better state of well-being, and
constitutes a brief, sensitive, and easily
administered measure.
Procedure
The study was organized into three phases. In
the baseline phase (week zero), the four
measurements—Cooper test, elbow push-ups,
fifty-meter speed, and emotional well-being—
were administered under standardized conditions
and within the same time slot to all participants in
both groups. During the intervention phase
(weeks one to twelve), the experimental group
developed the artificial intelligence-mediated
physical education program, while the control
group followed conventional physical education
with identical frequency and volume. In the post-
intervention phase (week thirteen), the
measurements were repeated with the same
instruments, the same evaluators, and the same
application conditions as in the baseline. To
reduce biases, the evaluators were unaware of the
condition of each classroom, the order of test
application was kept constant, and data were
coded anonymously before analysis.
Sample Size and Statistical Power
The sample size (N=240; n=120 per group)
was considered adequate to detect effects of
moderate magnitude. A sensitivity analysis
indicates that, with two groups, a significance
level of 0.05, and a power of 0.80, the design
allows detecting standardized differences from d
≈ 0.37, a value lower than the effects expected
according to the literature. The statistical power
achieved (post hoc) for each contrast is reported
in the results section.
Statistical Analysis
Data management and cleaning were
performed using Microsoft Excel, and statistical
processing with the R language and environment
(version 4.4.1). Data organization and
transformation were carried out with the
tidyverse ecosystem (dplyr and tidyr packages),
and descriptive statistics—means, standard
deviations, and intervals—were obtained with the
psych package. To ensure reproducibility, a
random seed was set (set.seed) before any
resampling-based procedure.
The inferential analysis plan comprised: (a)
verification of baseline equivalence between
groups using independent samples t-tests and the
chi-square test for sex; (b) verification of
normality assumptions with the Shapiro-Wilk test
(shapiro.test function) and homogeneity of
variances with Levene's test (car package); (c) an
analysis of covariance (ANCOVA) on the post-
measurement, with the baseline as covariate,
estimated with the car package (type III sums of
squares) and whose adjusted marginal means and
95% confidence intervals were obtained with the
emmeans package; (d) a mixed two-factor (group
× time) analysis of variance with the afex package
(aov_ez function) and partial eta squared; (e) t-
tests for related and independent samples—the
latter with Welch's correction for unequal
variances—and, when assumptions were not met,
their non-parametric equivalent (Wilcoxon tests);
(f) calculation of Cohen's d effect size and partial
eta squared with the effect size package, with 95%
RIAF Journal ISSN: 2953-6693 Vol – 4 No. 2, July 2026
158
confidence intervals obtained by bootstrap
resampling (2000 repetitions) using the boot
package; (g) statistical power analysis—both a
priori sensitivity calculation and achieved
power—with the pwr package; (h) moderation
analysis, evaluating group × educational level and
group × sex interactions on change scores; (i)
test-retest stability, estimated with pre-post
correlation in the control group; and (j) Pearson
correlations between changes in physical fitness
and emotional well-being, with their confidence
intervals and power, calculated with the
correlation package. Figures were generated with
the ggplot2 package. A significance level of 0.05
was established. As an essential methodological
limitation, it is reiterated that the data are
simulated, so the results illustrate the analytical
procedure and do not constitute empirical
evidence.
Introduction
Los análisis se realizaron sobre la totalidad de
los 240 casos simulados, sin valores perdidos. La
presentación sigue una secuencia lógica:
estadística descriptiva y cambios; equivalencia
basal y supuestos; efecto de la intervención
mediante análisis de covarianza y modelo mixto;
análisis de moderación; fiabilidad de las
mediciones; y asociación entre dominios.
The analyses were performed on all 240
simulated cases, with no missing values. The
presentation follows a logical sequence:
descriptive statistics and changes; baseline
equivalence and assumptions; intervention effect
via analysis of covariance and mixed model;
moderation analysis; reliability of measurements;
and association between domains.
Descriptive Statistics and Pre-Post
Changes
varianza mixto de dos factores (grupo × tiempo)
con el paquete afex (función aov_ez) y eta
cuadrado parcial; (e) pruebas t para muestras
relacionadas e independientes —estas últimas con
corrección de Welch ante varianzas desiguales—
y, cuando los supuestos no se cumplieron, su
equivalente no paramétrico (pruebas de
Wilcoxon); (f) el cálculo del tamaño del efecto d
de Cohen y del eta cuadrado parcial con el
paquete effectsize, con intervalos de confianza
del 95 % obtenidos por remuestreo bootstrap
(2000 repeticiones) mediante el paquete boot; (g)
el análisis de la potencia estadística —tanto el
cálculo a priori de sensibilidad como la potencia
alcanzada— con el paquete pwr; (h) el análisis de
moderación, evaluando las interacciones grupo ×
nivel educativo y grupo × sexo sobre las
puntuaciones de cambio; (i) la estabilidad test-
retest, estimada con la correlación pre-post en el
grupo control; y (j) las correlaciones de Pearson
entre los cambios de la condición física y del
bienestar emocional, con sus intervalos de
confianza y potencia, calculadas con el paquete
correlation. Las figuras se elaboraron con el
paquete ggplot2. Se fijó un nivel de significación
de 0,05. Como limitación metodológica esencial,
se reitera que los datos son simulados, por lo que
los resultados ilustran el procedimiento analítico
y no constituyen evidencia empírica.
Introducción
Los análisis se realizaron sobre la totalidad de
los 240 casos simulados, sin valores perdidos. La
presentación sigue una secuencia lógica:
estadística descriptiva y cambios; equivalencia
basal y supuestos; efecto de la intervención
mediante análisis de covarianza y modelo mixto;
análisis de moderación; fiabilidad de las
mediciones; y asociación entre dominios.
Estadística descriptiva y cambios pre-post
RIAF Journal ISSN: 2953-6693 Vol – 4 No. 2, July 2026
159
Examination of the descriptive statistics
reveals a consistent pattern of greater
improvement in the experimental group. In
relative terms, this group increased aerobic
capacity by approximately 4.3% (compared to
1.5% for the control), muscular endurance by
about 14% (compared to 4.5%), speed by 3.2%
(compared to 1.6%), and emotional well-being by
10.4% (compared to 2.8%), while estimated
maximal oxygen consumption increased by 5.9%
(compared to 2.1%). Intragroup contrasts using
related-samples t-tests confirmed that these
changes were statistically significant in the
experimental group for all four variables (all p <
0.001), with large effect sizes (Cohen's d between
1.00 and 1.24); in the control group, changes also
reached significance, although with small to
moderate magnitudes (Cohen's d between 0.36
and 0.61). This preliminary contrast in the
magnitude of changes anticipates the differential
effect formally confirmed in subsequent
inferential analyses.
After the intervention, the experimental group
increased the distance in the Cooper test by 87.4
meters compared to 31.2 for the control,
improved by 2.4 repetitions in push-ups
compared to 0.7, reduced the fifty-meter speed
time by 0.28 seconds compared to 0.14, and
increased well-being by 6.4 points compared to
1.7. Figure 1 illustrates the evolution of the
means, and Figure 2 shows the complete
distributions using box plots, evidencing the
favorable shift of the experimental group without
influential outliers.
Table 1.
Descriptive statistics (mean and standard deviation) by group and time point. (N = 240).
Variable
Group
Pre M (SD)
Post M (SD)
Change
(Δ)
Cooper Test (m)
Experimental
2013.9 (283.1)
2101.3 (288.5)
+87.4
Cooper (m)
Control
2018.1 (286.5)
2049.3 (291.8)
+31.2
Estimated VO2max
(ml/kg/min)
Experimental
33.7 (6.3)
35.7 (6.5)
+2.0
Estimated VO2max
(ml/kg/min)
Control
33.8 (6.4)
34.5 (6.5)
+0.7
Elbow Push-ups
30 s (rep)
Experimental
17.1 (4.8)
19.4 (5.3)
+2.4
Elbow Push-ups
30 s (rep)
Control
15.7 (5.9)
16.4 (6.3)
+0.7
Speed 50 m (s)
Experimental
8.88 (1.11)
8.60 (1.15)
−0.28
Speed 50 m (s)
Control
8.77 (1.03)
8.63 (1.05)
−0.14
WHO-5 Well-being (0–100)
Experimental
61.2 (13.3)
67.5 (14.2)
+6.4
WHO-5 Well-being (0–100)
Control
61.4 (13.4)
63.1 (13.8)
+1.7
Figure 1. Means (with standard error) before and after the intervention, by group.
RIAF Journal ISSN: 2953-6693 Vol – 4 No. 2, July 2026
160
Figure 2. Distributions (box plots) pre and post by group.
Baseline Equivalence and Assumption Verification
Before the intervention, the groups did not differ significantly in most variables, with trivial effect sizes
(Table 2). Only elbow push-ups showed a small baseline difference (p=0.041; d=0.27), and sex
distribution was homogeneous between groups (χ²=2.40; p=0.121). This equivalence, together with the
use of analysis of covariance, controls for the possible selection bias inherent in quasi-experimental
designs.
Table 2.
Baseline Equivalence and Assumption Verification
Variable
M Exp.
M Control
t
p
d
Cooper Test (m)
2013.9
2018.1
−0.11
0.910
0.01
Elbow Push-ups 30 s (rep)
17.1
15.7
2.05
0.041
0.27
Speed 50 m (s)
8.88
8.77
0.83
0.409
0.11
WHO-5 Well-being (0–100)
61.2
61.4
−0.13
0.896
0.02
The Shapiro-Wilk test indicated normality of change scores for the Cooper test, speed, and
well-being (p>0.05), while push-ups deviated from normality; Levene's test revealed homogeneity
of variances for Cooper and well-being, and unequal variances for push-ups and speed (Table 3).
Consequently, between-group comparisons were performed with Welch's correction and the main
contrasts were complemented with analysis of covariance, robust to these deviations; results were
confirmed through non-parametric tests, with equivalent conclusions.
Tabla 3.
Verification of assumptions about change scores
Variable
Shapiro p
(Exp.)
Shapiro p (Ctrl.)
Levene p
Decision
Cooper Test (m)
0,206
0,802
0,158
Met
Elbow Push-ups 30 s
0,007
< 0,001
0,002
Welch/non-param.
Speed 50 m (s)
0,296
0,456
0,035
Welch
WHO-5 Well-being
0,409
0,505
0,117
Met
RIAF Journal ISSN: 2953-6693 Vol – 4 No. 2, July 2026
161
Intervention Effect: ANCOVA and Mixed Model
The analysis of covariance, adjusting the post-measurement for the baseline, confirmed a significant
group effect in all four variables, with systematically more favorable adjusted means in the experimental
group (Table 4). Convergently, the mixed analysis of variance evidenced significant group × time
interactions, and change scores differed between groups with effect sizes ranging from moderate to large,
confidence intervals excluding zero, and achieved statistical power close to one (Table 5; Figure 3).
Table 4.
Analysis of covariance (post-measurement adjusted by baseline) and adjusted marginal means with
95% CI.
Variable
F(1,2
37)
η²
partial
Adj. M Exp. [95% CI]
Adj. M Ctrl. [95% CI]
Cooper Test (m)
36,25
0,133
2103,4 [2090,4–2116,4]
2047,3 [2034,3–2060,3]
Push-ups 30 s (rep)
55,65
0,190
18,70 [18,40–18,99]
17,10 [16,80–17,39]
Speed 50 m (s)
18,38
0,072
8,55 [8,50–8,59]
8,69 [8,64–8,73]
WHO-5 Well-being
49,48
0,173
67,63 [66,70–68,55]
62,98 [62,06–63,90]
All ANCOVA effects were significant (p<0.001). Partial eta squared values, between 0.072 and 0.190,
correspond to medium to large effects.
Table 5.
Group × time interaction, between-group difference in change, effect size with 95% CI, and power
Variable
Interaction
F(1,238) [η²p]
Cohen's d [95% CI]
Power
p
Cooper Test (m)
36.38 [0,133]
0.78 [0,52–1,04]
1.00
< 0.001
Push-ups 30 s (rep)
59.77 [0,201]
1.00 [0,76–1,29]
1.00
< 0.001
Speed 50 m (s)
18.63 [0,073]
−0.56 [−0,82–−0,30]
0.99
< 0.001
WHO-5 Well-being
49.62 [0,173]
0.91 [0,64–1,21]
1.00
< 0.001
Figure 3. Between-group effect size (Cohen's d) with 95% confidence intervals.
RIAF Journal ISSN: 2953-6693 Vol – 4 No. 2, July 2026
162
Moderation Analysis
Group × educational level and group × sex
interactions were not significant in any of the
variables (all p > 0.28 and partial eta squared ≤
0.005), indicating that the magnitude of the
intervention effect was homogeneous between
primary and secondary levels and between male
and female students. This consistency supports
the robustness and potential generalizability of
the result pattern within the simulated model.
Measurement Reliability
Test-retest stability, estimated by the
correlation between pre- and post-measurements
in the control group, was high for physical tests (r
= 0.98 for the Cooper test, push-ups, and speed)
and high for emotional well-being (r = 0.94),
suggesting that instruments offered consistent
measurements and that observed changes in the
in the experimental group are not attributable to
random instrument fluctuations.
Association between Physical Fitness and
Emotional Well-being
In the experimental group, changes in physical
fitness were associated with improvements in
emotional well-being (Table 6; Figure 4). The
highest correlation corresponded to the
relationship between change in aerobic capacity
and change in well-being, followed by strength;
improvement in speed (negative change in time)
was also associated with greater well-being. The
correlation matrix among all changes (Figure 5)
shows a coherent pattern of low to moderate
magnitude associations.
Table 6.
Pearson correlations between changes in physical fitness and emotional well-being (experimental
group, n = 120).
Relationship between changes
r [95 % CI]
p
Power
Δ Aerobic capacity (Cooper) – Δ Well-being
0.38 [0,21–0,52]
< 0.001
0.99
Δ Strength (push-ups)– Δ Well-being
0.28 [0,11–0,44]
0.002
0.88
Δ Speed (50 m) – Δ Well-being
−0.25 [−0,41–−0,08]
0.006
0.80
Figure 4. Scatter plots with regression line between changes in physical fitness and well-being
(experimental group).
RIAF Journal ISSN: 2953-6693 Vol – 4 No. 2, July 2026
163
Figure 5. Correlation matrix of changes (experimental group).
Discussion
The present study evaluated, using a
methodological model with simulated data, the
effect of a physical education program mediated
by artificial intelligence on the physical fitness and
emotional well-being of basic education students.
The simulated results showed significantly greater
improvements in the experimental group across
all four variables, with significant group-by-time
interaction effects, favorable adjusted means after
controlling for baseline measurement, effect sizes
ranging from moderate to large with precise
confidence intervals, and positive associations
between physical and emotional gains. These
findings should be interpreted as an illustration of
the expected behavior under the study
hypotheses, and not as empirical evidence.
The direction of the simulated results is
consistent with the available literature. Regarding
physical fitness, the observed favorable effects
align with evidence on school-based programs
supported by wearable devices, which have
reported small to moderate increases in daily
physical activity (Casado-Robles et al., 2022),
although with heterogeneous results depending
on the design and type of device (Chen et al.,
2025). Similarly, active video games have shown
improvements in physical fitness, enjoyment, and
psychological well-being in schoolchildren
(Marsigliante et al., 2024; Rosi et al., 2025),
supporting the plausibility of the gamification
component included in the intervention.
Regarding emotional well-being, the simulated
improvement in the experimental group is
consistent with the body of evidence linking
physical activity with benefits in child and
adolescent mental health, particularly on stress
and social competence (Fu et al., 2025), and with
school-based trials that have documented
reductions in depressive symptoms and increases
in life satisfaction (Ahmed et al., 2023). The
incorporation of a conversational assistant
oriented towards emotional self-regulation aligns
with the robust meta-analytic evidence on social
and emotional learning programs (Cipriano et al.,
2023) and with the still emerging and cautious
perspectives on the use of artificial intelligence in
the mental health of minors (Dritsona et al.,
2025).
RIAF Journal ISSN: 2953-6693 Vol – 4 No. 2, July 2026
164
The positive association between physical and
emotional gains in the experimental group
suggests, within the limits of the model, that both
domains could benefit jointly when artificial
intelligence articulates training personalization
with socio-emotional support (Lee & Lee, 2021;
Li et al., 2024; Zhang et al., 2022). However, an
alternative possible explanation is that the
improvement in well-being derives from greater
enjoyment and technological novelty rather than
from the physical gain itself, which should be
clarified in empirical studies with mediation
analysis.
From the perspective of analytical robustness,
the results were consistent across complementary
procedures: the analysis of covariance, which
controls for baseline measurement, coincided
with the mixed model and the comparison of
change scores; effects were maintained after
applying Welch's correction for unequal variances
and non-parametric tests; effect size confidence
intervals were precise and achieved statistical
power was close to unity; and the moderation
analysis did not reveal differences according to
educational level or sex, pointing to a stable
pattern. The high test-retest stability reinforces
measurement reliability. Together, these elements
illustrate the type of methodological control
required of quality research, although their value
remains conditioned by the simulated nature of
the data.
Among the strengths of the approach are the
integration of two domains usually studied
separately, the use of recognized field tests and a
validated well-being index, and the application of
a comprehensive and reproducible analytical plan.
As limitations, the most relevant is that the data
are simulated, so conclusions cannot be
generalized to real populations; the quasi-
experimental design does not allow strong causal
inferences; field tests provide estimates and not
direct measures, especially the Cooper test in
preadolescents, whose validity has been
questioned (Martínez-Lemos et al., 2024); and the
twelve-week period does not inform about the
sustainability of the effects. Likewise, the use of
artificial intelligence with minors poses ethical
challenges of privacy, equity, and human
supervision that condition its implementation
(Farrelly & Baker, 2023; Ng et al., 2024).
The implications of this model are primarily
methodological and pedagogical.
Methodologically, the study provides a replicable
template for articulating the objective assessment
of physical fitness with the measurement of
emotional well-being. Pedagogically, it offers an
intervention scheme that combines adaptive
monitoring, motor gesture analysis, gamification,
and socio-emotional support. Future lines of
research should execute the protocol with
empirical data, employ cluster-randomized
designs with long-term follow-up, incorporate
objective measures of physical activity, analyze
mediation mechanisms between physical fitness
and well-being, and evaluate equity of access to
technological tools. In summary, the scope of this
work is limited to demonstrating the coherence
of an analytical model; its significance will depend
on subsequent empirical verification.
Conclusions
In line with the stated objective, the results
obtained from simulated data suggest that a
physical education program mediated by artificial
intelligence could be associated with superior
improvements in physical fitness and emotional
well-being in basic education students, compared
to conventional physical education, and that both
improvements would tend to be positively
related. These claims are illustrative and cautious
in nature, as they are not derived from real
measurements.
RIAF Journal ISSN: 2953-6693 Vol – 4 No. 2, July 2026
165
Therefore, it is concluded that artificial
intelligence, used as a pedagogical tool,
constitutes a promising pathway to integrate
physical development and emotional balance
within basic education, provided that its adoption
is accompanied by methodological rigor, ethical
safeguards, and teacher supervision. Research
with empirical data, robust designs, and long-
term follow-up is required to confirm these
trends and specify the mechanisms involved.
References
Ahmed, K. R., Horwood, S., & Khan, A. (2023).
Effects of a school-based physical activity
intervention on adolescents’ mental health: A
cluster randomized controlled trial. Journal of
Physical Activity and Health, 20(12), 1102–
1108.
https://journals.humankinetics.com/view/jo
urnals/jpah/20/12/article-p1102.xml
Aubert, S., Barnes, J. D., Demchenko, I.,
Hawthorne, M., … Tremblay, M. S. (2022).
Global Matrix 4.0 physical activity report card
grades for children and adolescents: Results
and analyses from 57 countries. Journal of
Physical Activity and Health, 19(11), 700–728.
https://doi.org/10.1123/jpah.2022-0456
Bull, F. C., Al-Ansari, S. S., Biddle, S., Borodulin,
K., Buman, M. P., Cardon, G., … Willumsen,
J. F. (2020). World Health Organization 2020
guidelines on physical activity and sedentary
behaviour. British Journal of Sports Medicine,
54(24), 1451–1462.
https://doi.org/10.1136/bjsports-2020-
102955
Casado-Robles, C., Viciana, J., Guijarro-Romero,
S., & Mayorga-Vega, D. (2022). Effects of
consumer-wearable activity tracker-based
programs on objectively measured daily
physical activity and sedentary behavior
among school-aged children: A systematic
review and meta-analysis. Sports Medicine -
Open, 8, 18.
https://doi.org/10.1186/s40798-021-00407-
6
Chen, X., et al. (2025). Effectiveness of wearable
activity trackers on physical activity among
adolescents in school-based settings: A
systematic review and meta-analysis. BMC
Public Health, 25(1).
https://doi.org/10.1186/s12889-025-22170-
z
Cipriano, C., Strambler, M. J., Naples, L. H., Ha,
C., Kirk, M., Wood, M., … Durlak, J. A.
(2023). The state of evidence for social and
emotional learning: A contemporary meta-
analysis of universal school-based SEL
interventions. Child Development, 94(5),
1181–1204.
https://doi.org/10.1111/cdev.13968
Dritsona, A., Koutroumpa, M. E., Sergentanis, T.
N., & Tsitsika, A. K. (2025). Artificial
intelligence and mental health of children and
adolescents: Current status and perspectives.
Advances in Experimental Medicine and
Biology, 1489, 477–484.
https://doi.org/10.1007/978-3-032-03394-
9_45
Farrelly, T., & Baker, N. (2023). Generative
artificial intelligence: Implications and
considerations for higher education practice.
Education Sciences, 13(11), 1109.
https://doi.org/10.3390/educsci13111109
Fu, Q., Li, L., Li, Q., & Wang, J. (2025). The
effects of physical activity on the mental health
of typically developing children and
adolescents: A systematic review and meta-
analysis. BMC Public Health, 25(1), 1514.
https://doi.org/10.1186/s12889-025-22690-
8
Lee, H. S., & Lee, J. (2021). Applying artificial
intelligence in physical education and future
perspectives. Sustainability, 13(1), 351.
https://doi.org/10.3390/su13010351
Li, S., Wang, C., & Wang, Y. (2024). Fuzzy
evaluation model for physical education
teaching methods in colleges and universities
using artificial intelligence. Scientific Reports,
14. https://doi.org/10.1038/s41598-024-
53177-y
RIAF Journal ISSN: 2953-6693 Vol – 4 No. 2, July 2026
166
Marques, A., Henriques-Neto, D., Peralta, M.,
Martins, J., Demetriou, Y., Schönbach, D. M.
I., & Gaspar de Matos, M. (2020). Prevalence
of physical activity among adolescents from
105 low, middle, and high-income countries.
International Journal of Environmental
Research and Public Health, 17(9), 3145.
https://doi.org/10.3390/ijerph17093145
Marsigliante, S., My, G., Mazzotta, G., &
Muscella, A. (2024). The effects of exergames
on physical fitness, body composition and
enjoyment in children: A six-month
intervention study. Children, 11(10), 1172.
https://doi.org/10.3390/children11101172
Martínez-Lemos, I., et al. (2024). Reliability and
criterion-related validity of the Cooper test in
pre-adolescents and adolescents: A systematic
review and meta-analysis. Journal of Sports
Sciences, 42(3), 222–236.
https://doi.org/10.1080/02640414.2024.232
6352
Mayorga-Vega, D., Aguilar-Soto, P., & Viciana, J.
(2015). Criterion-related validity of the 20-m
shuttle run test for estimating
cardiorespiratory fitness: A meta-analysis.
Journal of Sports Science & Medicine, 14(3),
536–547.
https://www.jssm.org/hfpdf.php?volume=1
4&issue=3&page=536
Ng, D. T. K., Su, J., Leung, J. K. L., & Chu, S. K.
W. (2024). Artificial intelligence (AI) literacy
education in secondary schools: A review.
Interactive Learning Environments, 32(10),
6204–6224.
https://doi.org/10.1080/10494820.2023.225
5228
Rosi, E., Bianchi, V., Baù, I., Nuzzo, R.,
Valsecchi, S., Molteni, M., & Colombo, P.
(2025). Effects of exergames on motor skills,
psychological well-being, and cognitive
abilities in schoolchildren and adolescents:
Scoping review. JMIR Pediatrics and
Parenting, 8, e71416.
https://doi.org/10.2196/71416
Topp, C. W., Østergaard, S. D., Søndergaard, S.,
& Bech, P. (2015). The WHO-5 Well-Being
Index: A systematic review of the literature.
Psychotherapy and Psychosomatics, 84(3),
167–176.
https://doi.org/10.1159/000376585
van Sluijs, E. M. F., Ekelund, U., Crochemore-
Silva, I., Guthold, R., Ha, A., Lubans, D., et al.
(2021). Physical activity behaviours in
adolescence: Current evidence and
opportunities for intervention. The Lancet,
398(10298), 429–442.
https://doi.org/10.1016/S0140-
6736(21)01259-9
Xu, K., et al. (2023). Predicting academic
performance associated with physical fitness
of primary school students using machine
learning methods. Complementary Therapies
in Clinical Practice, 51, 101736.
https://doi.org/10.1016/j.ctcp.2023.101736
Zhang, B., Jin, H., & Duan, X. (2022). Physical
education movement and comprehensive
health quality intervention under the
background of artificial intelligence. Frontiers
in Public Health, 10, 947731.
https://doi.org/10.3389/fpubh.2022.947731
Declaration of conflicts of interest:
Through this document, the authors of this article
assume the regulatory and normative provisions
of the RIAF Journal's publication and therefore
declare no conflicts of interest in this regard.
Authors' Contribution:
Jorge Andrés Bayona Martínez: Elaborated
the research methodology, developed the
diagnostic instruments and techniques, and
carried out the application of the program in the
experimental group.
Edwin Esneider Andrade Perdomo:
Conducted the bibliographic search and
construction of citations, references, and
theoretical systematization of the study variables.
RIAF Journal ISSN: 2953-6693 Vol – 4 No. 2, July 2026
167
Steven Arturo Torres Burgos: Participated in
the review, analysis, and decision-making in the
writing of the literature, theoretical
systematization of the study variables, application
of empirical instruments, and in the results and
conclusions of the research.