AI-Mediated Generative Art in Primary Education: Emotional Expression, Creativity and the Limits of Visual Reflection

Abstract

This study examines the implementation of an educational workshop on AI-mediated generative art, entitled Programmed Emotions, developed with primary school students. The research aims to explore the potential of generative technologies as tools for emotional expression and creativity, as well as to analyse their implications for students’ reflective processes. A mixed-methods approach was employed, combining quantitative data from self-assessment scales with qualitative data from open-ended responses and image analysis. The study investigated how the use of generative tools influenced engagement, emotional expression, creativity, and reflection. The results revealed high levels of active participation and creative production, together with a clear predominance of positive emotions, particularly joy, and a strong sense of identification between students and their artworks. However, the findings also highlighted significant limitations in reflective depth, characterized by brief responses and limited conceptual elaboration. These results suggest that, although generative art can effectively promote emotional expression and creativity, it does not by itself ensure deeper processes of understanding. Overall, the study underscores the educational potential of generative art in early educational stages while emphasizing the need for explicit pedagogical mediation to foster critical thinking, emotional diversity, and meaningful reflection.

1. Introduction

In recent years, the integration of digital technologies into education has significantly transformed teaching and learning practices, opening new possibilities for the development of creative, emotional and cognitive skills [1]. In particular, the emergence of Generative Artificial Intelligence (hereinafter GAI) has introduced tools capable of automatically generating visual, audio or textual content, presenting new opportunities and challenges for early-stage education. In this context, GAI-mediated generative art emerges as a particularly relevant space for exploring the relationship between creativity, emotion and technology [2]. Few studies have examined how AI-mediated generative art affects the relationship between emotional expression and reflective depth in primary education contexts.

At the same time, emotional education has been recognised as a fundamental element in students’ holistic development, as it fosters the ability to identify, understand and express emotions, as well as to establish healthy interpersonal relationships [3]. In this regard, the integration of creative tools in the classroom can be an effective means of promoting emotional expression, particularly at an early age, when verbal language is still developing. Artistic practices, and particularly those mediated by digital technologies, enable pupils to externalise internal states through visual forms, facilitating processes of self-awareness and emotional communication [4].

However, despite the potential of these tools, the literature also warns of certain risks associated with their use in education. It has been noted that creative output facilitated by automated systems does not necessarily guarantee the development of deep understanding or the cultivation of reflective processes [5,6]. Furthermore, automation may encourage more immediate or superficial responses if not accompanied by appropriate pedagogical guidance [7]. In this regard, a significant tension emerges between these technologies’ capacity to foster engagement and creativity and their potential to limit the development of reflection and critical thinking.

Despite growing interest in the integration of AIG in education, there is still limited empirical evidence regarding its specific impact on the development of emotional expression and reflection in early stages of education, particularly in contexts of artistic creation mediated by generative systems. This gap is particularly significant, given that primary education constitutes a key period for the development of emotional and creative skills.

Rather than examining creativity or emotional expression in isolation, this study specifically explores the tension between high levels of visual creative engagement and the limited depth of reflective discourse observed in AI-mediated educational contexts. In this regard, the research provides exploratory empirical evidence on how generative systems can facilitate emotional engagement without necessarily ensuring deep metacognitive or reflective processes among primary school pupils.

In this context, the research examines the implementation of an educational workshop based on generative art mediated by AI, entitled “Experiences in Programmed Emotions”, carried out with primary school pupils using a mixed-methods approach; the aim is to explore how these tools influence pupils’ active participation, creativity, emotional expression and reflection. It also aims to examine the relationship between emotional engagement and depth of reflection in this type of educational experience, thereby contributing to an understanding of the role of AI in pupils’ holistic development.

The present study aims to analyse the impact of an educational proposal based on generative art mediated by artificial intelligence on primary school pupils, with the aim of examining its influence on active participation, creativity, emotional expression and reflection. It also seeks to explore the relationship between emotional engagement and depth of reflection in this type of educational experience.

In this context, the research seeks to answer the following research questions:

(1)

To what extent does the use of generative art mediated by artificial intelligence promote pupils’ participation, creativity and emotional expression?

(2)

How does the reflective dimension manifest itself in this type of experience?

(3)

What is the relationship between creativity and reflection in the learning process?

2. Theoretical Framework

The intersection between emotion, creativity, and technology constitutes an emerging field in educational research. Within this framework, digital creation can be understood as a space in which students project their emotional states through visual representations, integrating affective and cognitive processes.

2.1. Emotional Education

Emotional education has become established as an essential component of students’ holistic development, complementing the acquisition of academic knowledge with the development of personal and social competencies. In this sense, it involves the ability to recognize, understand, and regulate one’s own emotions, as well as to establish positive interpersonal relationships [3].

Since the 1990s, the concept of emotional intelligence has contributed to highlighting the importance of these competencies in education. Emotional intelligence includes dimensions such as self-awareness, self-regulation, motivation, empathy, and social skills, all of which are fundamental for learning and adaptation to the environment [8]. In this regard, recent studies have shown that emotional intelligence is significantly related to academic performance and student well-being, regardless of sociodemographic variables [9,10].

Likewise, schools represent a key environment for emotional development, as they provide contexts of interaction in which students can experience, express, and regulate their emotions. A positive educational environment, based on respect and emotional safety, promotes active participation and a sense of belonging, thereby contributing to students’ psychological well-being [11,12].

2.2. Creativity in Education

Creativity constitutes a complex cognitive process that enables individuals to generate original ideas and establish meaningful connections between experiences, knowledge, and contexts. In education, creativity is not limited to the production of novel outcomes but also involves the ability to reinterpret reality and construct meaning from experience [13].

From a psychological perspective, creativity has been defined as a process integrating multiple cognitive functions, including flexibility, fluency, and originality [14]. In this sense, all individuals possess creative potential, which can be developed through appropriate educational contexts that encourage curiosity, exploration, and personal expression [15,16].

In schools, creativity plays a central role in the development of twenty-first-century competencies such as critical thinking, problem-solving, and adaptation to changing contexts [17]. However, several authors warn that creativity should not be understood solely as spontaneous production, but rather as a process that requires pedagogical mediation to become meaningful learning [5].

2.3. Art and Emotional Expression

The arts constitute one of the most effective means of integrating emotional and cognitive dimensions into learning processes. Through artistic creation, students can explore, represent, and reinterpret their emotional experiences, fostering processes of self-knowledge and personal expression [4].

In this sense, arts education provides access to forms of knowledge that are not strictly discursive, thereby expanding possibilities for representing and understanding reality. As Gardner [18] argues, the arts promote the development of multiple forms of intelligence and contribute to the construction of complex meanings.

Likewise, graphic creativity, understood as the ability to generate original and meaningful visual representations, represents a relevant dimension in students’ creative development. This involves not only the production of images but also the attribution of symbolic meaning to them, integrating perceptual, cognitive, and emotional aspects [19].

2.4. Generative Art and Artificial Intelligence

Generative art can be defined as an artistic process in which the final outcome is produced wholly or partially through autonomous systems or algorithms. In this type of practice, the artist establishes the rules and parameters that govern the process while partially relinquishing control over the final result to the generative system [20]. Unlike other forms of digital art, the defining feature of generative art is not merely the use of technological tools, but the rule-based system through which the artwork is created [20]. This characteristic introduces a conception of creativity as a distributed and mediated process between the individual and the technological system, both of which actively contribute to the production of the work. Such an approach is particularly relevant in educational contexts, where students can interact with systems that expand their expressive possibilities and facilitate experimentation. Previous studies have shown that these technologies can increase motivation and support creative production [2,21], enabling students to represent ideas and emotions in accessible and immediate ways without requiring advanced technical or representational skills.

In these practices, the artist designs the system but relinquishes part of the control to the generative process, implying a redefinition of the traditional concept of authorship [22,23]. As these authors note [20], creativity can be understood as a distributed process between the creator and the system, in which the degree of human intervention may vary from minimal to highly significant.

This conception is exemplified by pioneering developments such as the AARON system, created by Harold Cohen, which could generate complex visual works from predefined rules [24]. Such systems demonstrate that creativity does not reside exclusively in the result, but also in the design of the process that generates it.

In educational contexts, the incorporation of AI-based tools expands students’ creative possibilities, enabling them to explore new forms of visual expression and experiment with systems capable of generating complex outcomes from simple instructions. These tools may act as catalysts for creativity and motivation by providing immediate and visually attractive results [21].

However, several studies warn that the use of generative technologies also poses pedagogical challenges. It has been argued that the automation of certain creative processes may reduce students’ cognitive engagement if it is not accompanied by strategies that foster reflection and critical thinking [7]. In this regard, creativity mediated by digital systems does not in itself guarantee the construction of deep knowledge but rather requires pedagogical mediation to guide the learning process.

3. Materials and Methods

3.1. Design

The present study is framed within a mixed-methods approach, integrating quantitative and qualitative techniques in a complementary manner to obtain a broad and in-depth understanding of the impact of an educational proposal based on AI-mediated generative art. This type of design makes it possible not only to measure variables in a structured way but also to explore students’ meanings, perceptions, and subjective processes [25,26].

The quantitative component focused on the systematic collection of students’ perceptions using Likert-type scales [27], whilst the qualitative component focused on the analysis of open-ended questions and students’ work, which facilitated a contextualised interpretation of the educational experience [28].

The research was developed through the implementation of the workshop Programmed Emotions: The Psychology Behind Generative Art, designed to explore the relationship between emotion, creativity, and technology among students.

The intervention was structured into four main phases (see Figure 1):

• Emotional activation phase: introduction to the concept of generative art and connection with students’ emotional experiences through the identification of personal emotions.
• Guided exploration phase: presentation of examples of generative art and familiarization with digital tools.
• Individual creation phase: each student selected an emotion and generated a visual artwork using digital platforms, employing elements such as color, shape, or movement to represent it.
• Reflection and dialogue phase: presentation of the productions and verbalization of creative decisions, represented emotions, and experiences lived throughout the process.

The images presented below are intended as illustrative examples of the visual productions generated by students. Although they are informed by the thematic and visual patterns identified during the analysis, they should not be interpreted as representing a systematically selected visual corpus.

The students interacted with the Bing Image Creator platform. This AI-powered image-generation tool, which uses text prompts, was chosen because it is particularly well-suited for working with children, combining ease of use with free access. Its simple interface allows users to generate images from descriptions in natural language without the need for technical knowledge. During the workshop, participants were given initial examples of descriptive instructions and basic support for modifying the visual results through iterative textual instructions. Each student generated several visual versions before selecting the final image for presentation and discussion. The teacher provided guidance throughout the process to support emotional interpretation, the construction of prompts and reflective dialogue.

3.2. Sample

The sample consisted of 54 primary school students (40 girls and 14 boys), aged between 9 and 12 years old (M = 11.2), corresponding to the 5th and 6th grades of Primary Education, from two schools. Participants were selected through purposive sampling linked to a specific educational context, within the framework of a workshop developed at the Faculty of Education of the University of Zaragoza.

Four groups from the two schools were included. Groups 1 and 2 (Group 1 N = 16; Group 2 N = 19) came from a private all-girls primary school whose families presented a high socioeconomic status. Groups 3 and 4 (Group 3 N = 13; Group 4 N = 6) came from a public school characterized by high cultural diversity.

This type of sample is appropriate in exploratory studies of pedagogical innovation, where the main objective is the understanding of educational processes in situated contexts rather than the generalization of results [29].

3.3. Instruments

For data collection, a questionnaire structured into two complementary sections was designed.

Quantitative section: included Likert-scale items (1–4) aimed at evaluating five key dimensions:

• Personal engagement: a person’s level of commitment, interest and active participation in an activity, encompassing cognitive, emotional and behavioral dimensions. In education, it relates to students’ motivation and meaningful connection with learning [30].
• Creativity: the ability to produce original and contextually appropriate ideas, responses, or productions through imagination and divergent thinking [31].
• Active participation: direct, conscious, and continuous involvement in an activity, demonstrating initiative, interest, and collaboration in the development of proposed tasks. Active participation promotes meaningful learning and engagement with the educational process [32].
• Emotional expression: the ability to identify, express, and communicate emotions and feelings through different forms of expression, such as verbal, bodily, artistic, or musical language [8].
• Reflection: a process of critical and conscious analysis of experiences, actions, or prior knowledge that enables learning, understanding, and personal or professional improvement [33].

These dimensions have been identified in the literature as relevant indicators in creative and experiential learning contexts [13,18]. Some examples of the items included were: “I actively participated in the activity”, “The activity helped me express my emotions” and “I reflected on my feelings during the process”, which were assessed using a four-point Likert scale. The questionnaire was designed specifically for the exploratory context of the workshop, drawing on recent research into creativity and artistic practices mediated by generative artificial intelligence, generative image AI and collaborative creativity [34]. Furthermore, the wording of the items was adapted linguistically to ensure they could be understood by primary school pupils.

Given the exploratory nature of the study, the instrument was not intended to constitute a standardised psychometric scale, an aspect that is explicitly recognised as a limitation of the research.

Qualitative section: included an open-ended question in which students were asked to:

• Justify the selected emotion;
• Describe the generative artwork they created.

This type of instrument makes it possible to access meaning-making processes and the narrative dimension of learning [35].

Additionally, analyses of the visual productions generated by the students were conducted and evaluated through a rubric [36,37]. Four researchers took part in the coding process; they reviewed the material independently and discussed any discrepancies until a consensus was reached. The section on visual analysis considered aspects such as emotional coherence, symbolic representation, visual composition, and the correspondence between the textual explanation and the image produced. The images included in the manuscript serve primarily an illustrative and exemplary function, rather than constituting a corpus subject to systematic and exhaustive iconographic analysis.

Throughout the study, the fundamental ethical principles of educational research were respected. Given that the study involved primary school students, the questionnaires were designed in accordance with the confidentiality and data protection principles established by the University of Zaragoza. Attention was paid to ensuring participants’ anonymity, the secure handling of personal information, and the ethical management of data throughout the research process. The instrument was reviewed and authorized by the Data Protection Unit of the University of Zaragoza under reference code RAT 2025-86, guaranteeing compliance with current regulations regarding educational research involving minors. Confidentiality and anonymity were ensured, and the collected data were used exclusively for academic purposes. Informed consent was obtained from both families and participating schools, in accordance with current regulations concerning data protection and children’s rights. Student participation was voluntary and took place within a regular educational context, without involving risks for participants.

The full questionnaire used in the study is included in Appendix A.1 to ensure methodological transparency and the future reproducibility of the research. Furthermore, the criteria used to analyse the visual works are set out in Appendix A.2.

3.4. Data Analysis

Data analysis was carried out through the integration of quantitative and qualitative techniques [38,39]:

• Quantitative analysis: descriptive statistical techniques were applied, including the calculation of means and the analysis of response distributions, to identify general trends in the evaluated dimensions. Inferential analyses were also conducted to examine differences between groups and relationships among variables.
• Qualitative analysis: open-ended responses were analyzed through a thematic categorization process that included the identification, coding, and grouping of meaning units into emerging categories [28]. This procedure made it possible to detect patterns related to expressed emotions, levels of reflection, and students’ discourse types.

4. Results

4.1. Quantitative Analysis

The analysis of the collected data combines quantitative and qualitative evidence to provide a comprehensive view of the impact of the workshop on students. This approach makes it possible not only to identify general trends in responses but also to understand the meanings and underlying processes involved in the educational experience (see

Table 1. Mean scores by group.

	Group	Average	Standard Deviation
personal engagement	1	3.38	0.71
	2	3.42	0.60
	3	2.77	0.92
	4	3.17	0.40
	Total	3.22	0.74
creativity	1	3.19	1.10
	2	3.68	0.58
	3	3.23	0.59
	4	3.33	0.51
	Total	3.39	0.78
active participation	1	3.69	0.47
	2	3.74	0.56
	3	3.08	0.86
	4	3.00	0.89
	Total	3.48	0.72
emotional expression	1	3.56	1.03
	2	3.68	0.47
	3	2.77	0.83
	4	2.67	0.81
	Total	3.31	0.88
reflection	1	2.69	1.53
	2	3.22	0.54
	3	2.85	0.89
	4	3.00	0.63
	Total	2.94	1.02

).

Before inferential analyses, the assumptions underlying parametric testing were examined. Normality and homogeneity of variances were assessed for the main study variables. Since the assumption of homogeneity of variances was violated for the reflection variable, Welch’s ANOVA was applied for this dimension. Given the exploratory nature of the study and the relatively small subgroup sizes, statistical findings were interpreted cautiously and complemented with effect size indicators.

First, the quantitative results show high levels across most of the analyzed dimensions. Active participation reached the highest mean score (M = 3.48), followed by creativity (M = 3.39 in both cases). These values indicate that the activity promoted a high degree of student engagement, fostering both active involvement in the process and the production of creative responses. Similarly, emotional expression obtained a mean score of 3.30, suggesting that the generated environment facilitated the externalization of emotions and personal connection with the activity.

However, the reflection dimension obtained a lower mean score (M = 2.94) and showed the greatest dispersion in responses. This result points to greater heterogeneity in students’ ability to develop reflective processes, in contrast to the homogeneity observed in active participation and creativity. The difference between these dimensions suggests that, while the activity effectively stimulated engagement and expression, deeper reflective processes did not occur uniformly. Descriptive results are presented in Table 1.

The homogeneity of variances test showed that the variables of personal engagement, creativity, active participation, and emotional expression met the assumption of homogeneity. However, the reflection variable did not meet this assumption; consequently, a Welch ANOVA was applied (see

Table 2. Group differences across study variables using ANOVA and Welch’s ANOVA.

Variable	Test	F	df	p	η2
Personal engagement	ANOVA	2.49	3.5	0.071	0.130
Creativity	ANOVA	1.54	3.5	0.215	0.085
Active participation	ANOVA	4.11	3.5	0.011	0.198
Emotional expression	ANOVA	6.87	3.5	0.001	0.292
Reflection	Welch ANOVA	0.27	3.49	0.848	0.016

).

The ANOVA results (see Table 2) indicated statistically significant group differences in active participation (F(3,50) = 4.11, p = 0.011, η² = 0.198) and emotional expression (F(3,50) = 6.87, p = 0.001, η² = 0.292), reflecting large effect sizes, particularly in emotional expression. No statistically significant differences were observed for personal engagement, creativity, or reflection.

Because the assumption of homogeneity of variances was violated for reflection according to Levene’s test, Welch’s ANOVA was applied for this variable. Post hoc comparisons using the Bonferroni adjustment revealed significant differences in emotional expression between Groups 1 and 3, Groups 1 and 4, Groups 2 and 3, and Groups 2 and 4. In addition, a statistically significant difference in active participation was observed between Groups 2 and 3 (p = 0.048). The relatively large effect sizes observed in active participation and emotional expression suggest meaningful group-related patterns; however, these findings should be interpreted cautiously given the exploratory nature of the study and the unequal subgroup sizes.

At the same time, given the different nature of the schools, differences between them could be expected. In this regard, the independent samples t-test showed statistically significant differences between schools in personal engagement (t(52) = 2.498, p = 0.016), active participation (t(52) = 3.562, p = 0.001), and emotional expression (t(52) = 3.998, p < 0.001), with the private all-girls school obtaining higher scores in all these variables. No significant differences were found in creativity or reflection (p > 0.05).

Since one of the schools (Groups 1 and 2) was exclusively female, a Student’s t-test was conducted to determine the existence of gender differences. Students from the predominantly female groups showed higher scores (M = 3.54; SD = 0.809) compared to boys (M = 2.62; SD = 0.768), with a significant difference between both groups (t = 3.618; p = 0.001). No statistically significant differences were observed in the remaining variables—personal engagement, creativity, active participation, and reflection (p > 0.05). These differences should be interpreted cautiously, since gender composition was closely associated with institutional context and school type. Therefore, the observed patterns may reflect a combination of sociocultural, educational, and contextual factors rather than gender alone. In other variables, such as creativity and reflection, gender differences were less pronounced, suggesting greater homogeneity in these dimensions (see

Table 3. Analysis of correlations between variables.

	1	2	3	4
1. personal engagement	1
2. creativity	0.01	1
3. active participation	0.43 **	0.16	1
4. emotional expression	0.32 *	0.28 *	0.29 *	1
5. reflection	0.14	0.26	0.29 *	0.39 **

* p < 0.05; ** p < 0.01.

).

Relationships between the variables were analyzed through correlation analyses. The results showed that personal engagement was positively and significantly associated with active participation (r = 0.430, p < 0.01) and emotional expression (r = 0.321, p < 0.05), indicating that higher levels of engagement are related to greater active participation and emotional expression.

Creativity did not show a significant relationship with personal engagement (r = 0.011, p > 0.05), although it did present a positive correlation with emotional expression (r = 0.281, p < 0.05). In other words, students may display high levels of creativity without this necessarily implying a greater capacity for analysis or reflection on their own process. This finding reinforces the idea that creativity and metacognition, although related, require differentiated pedagogical stimuli.

Active participation, in turn, was significantly related to both emotional expression (r = 0.290, p < 0.05) and reflection (r = 0.296, p < 0.05), suggesting that greater behavioral engagement is linked to reflective and emotional processes.

Finally, emotional expression showed a moderate positive correlation with reflection (r = 0.398, p < 0.01), indicating that both variables tend to increase together.

4.2. Qualitative Results Analysis

The qualitative analysis followed an exploratory thematic approach that included the identification of units of meaning, initial coding, grouping into emerging categories, and an interpretative review of recurring patterns. Given the exploratory nature of the study and the age of the participants, the analysis focused on identifying dominant trends in emotional expression and reflective discourse rather than establishing exhaustive or fully saturated categories. Consequently, interpretations relating to reflective depth were treated with caution, recognising that brief responses may also reflect developmental or expressive limitations associated with the participants’ age.

4.2.1. Analysis of Open-Ended Questions

The qualitative results complement and enrich these findings. The analysis of the open-ended responses reveals a clear predominance of positive emotions, especially joy and happiness (see Figure 2).

These emotions were repeatedly expressed through justifications such as “Because I like being happy,” “Because it makes me feel good,” or “Because it is my favorite.” In addition, these emotions were associated with meaningful everyday experiences for the students, such as sports, celebrations, or social relationships, showing a direct connection between the activity and their personal lives.

Another relevant pattern was the strong presence of personal identification discourses, in which students linked the selected emotion to their own identity or emotional state. Expressions such as “Because that’s how I am,” “Because it represents me,” or “Because it is how I feel right now” reflect a level of subjective appropriation that goes beyond a simple emotional choice, revealing processes of self-expression and self-knowledge.

4.2.2. Analysis of Artistic Productions

However, the qualitative analysis also revealed the existence of different levels of elaboration in the responses. Alongside more developed productions, brief or poorly elaborated responses appeared, such as “Because I like it” or “The first one I thought of,” highlighting significant variability in the students’ reflective depth. This heterogeneity coincides with the dispersion observed in the reflection variable in the quantitative analysis.

In addition, although to a lesser extent, unpleasant or complex emotions such as fear, frustration, or anger also emerged (see Figure 3) (“Fear at school,” “Because I get frustrated many times,” “The anger inside”). These responses are particularly relevant because they demonstrate students’ ability to explore diverse emotional registers and not only positive emotions, thus broadening the expressive scope of the activity.

Finally, elements of personal and aspirational projection were identified, particularly linked to sports and the future (“I want to make it to the NBA,” “When I become a teenager, I won’t have as much joy anymore”). This type of discourse suggests that the activity not only facilitates present emotional expression but also acts as a space for the construction of expectations, desires, and future identities.

The overall analysis of the visual productions made it possible to identify consistent patterns in the representation of different emotions by the students, both pleasant and unpleasant. In general terms, emotions were expressed through intensified, exaggerated, intense, visible, and easily recognizable features, indicating a conceptualization centered on external expressiveness.

Thus, in representations of pleasant emotions, as shown in Figure 2, happiness appeared associated with broad smiles, expansive body gestures (raised arms, movement, identified with triumph and victory), and bright contexts linked to play, nature, or social interaction. These productions reflect a conception of emotion as an active and shared experience. In terms of framing, most of these representations were depicted through close-up shots, with some exceptions corresponding to medium or long shots.

In contrast, representations of unpleasant emotions—as observed in Figure 3—were characterized by facial expressions of tension (wide-open eyes, furrowed brows), defensive or aggressive postures, and environments associated with danger or threat, such as dark spaces, aggressive animals, or situations of interpersonal conflict. In this case, emotion was constructed not only through the represented figure but also through the visual context. From a framing perspective, most images relied on close-ups, extreme close-ups focused solely on the face, long shots, and full-body shots.

Despite these differences in emotional content, both sets of images tended to represent emotions in their most extreme or prototypical forms, with little representation of intermediate or nuanced emotional states. Likewise, different levels of elaboration were identified in the productions, ranging from realistic representations centered on everyday situations to more symbolic or metaphorical compositions, in which emotions were projected onto external figures such as animals, fictional characters, or imaginary environments. From the perspective of framing and photographic aesthetics—largely characterized by the simulation of technical images—the compositions in both cases focused mainly on emphasizing facial features, with close-ups and long shots being the most common framing devices, alongside the exceptions illustrated in Figure 2 and Figure 3.

Another aspect observed during the analysis was the recurrence of similar visual conventions in emotional representations, such as smiling faces associated with joy or exaggerated facial expressions linked to anger. This convergence may reflect both culturally shared emotional codes and biases present in AI image-generation systems. However, many students simultaneously linked these visual representations to personal experiences and subjective emotional narratives, suggesting the coexistence of algorithmically mediated visual conventions and genuine processes of emotional appropriation.

Overall, the results show that the workshop generated high levels of participation, creativity, and emotional expression, while also revealing greater variability in reflective processes. The integration of quantitative and qualitative data confirms that the experience was meaningful for students, although with differing levels of depth in the elaboration of their responses and image productions.

5. Discussion

The results we obtained position this study within current debates on creativity, emotion and reflection in educational contexts mediated by artistic and generative technologies. One of the main findings concerns the tension observed between emotional engagement and reflective depth, a phenomenon previously identified in research on creative learning [3,4,5,6,20,40]. Existing literature suggests that emotional participation alone does not necessarily lead to deep reflective or metacognitive processes. As Flavell [6] and Kolb [40] argue, meaningful learning requires conscious reflection and metacognitive regulation. Similarly, research in arts education has shown that emotional expression may emerge without deep interpretative elaboration when explicit pedagogical mediation is absent [4].

On the one hand, the data reveal high levels of active participation, creativity, and emotional expression, consistent with studies highlighting the potential of arts-based methodologies to foster student engagement [41]. However, these results should be interpreted with caution. In contexts of AI-mediated visual creation, positive assessments of creativity may partly reflect students’ attraction to visually appealing generated images, rather than exclusively conceptually elaborate works. Furthermore, generative systems may produce standardised visual outcomes derived from common patterns present in their training data. Consequently, the creativity scores obtained in this study should be understood primarily as indicators of perceived creative engagement and expressive participation, rather than as objective measures of artistic originality.

Eisner [4] emphasizes that arts-based experiences promote more open, expressive, and personal forms of knowledge, which is clearly reflected in students’ responses, where references to personal experiences and meaningful emotions predominate. Likewise, studies such as those by Gardner [42] point out that creativity and emotional expression constitute key dimensions of learning, especially in contexts that allow multiple forms of representation.

However, the lower score obtained in the reflection variable and its weak correlation with creativity suggest that emotional engagement does not automatically translate into deep metacognitive processes. This result is consistent with Flavell’s [6] distinction between cognitive experience and metacognitive awareness, the latter requiring explicit processes of regulation and conscious reflection. In the same vein, Perkins [5] argues that deep thinking does not emerge spontaneously but rather requires support structures that guide reflection.

The tension observed between creativity and reflection may be particularly relevant in AI-mediated artistic contexts. As Boden and Edmonds [20] argue, creative processes in generative systems may be partially distributed between the user and the technological system itself. Consequently, visually sophisticated outputs do not necessarily imply deep cognitive elaboration or reflective understanding on the part of students. In this study, high levels of creative engagement coexisted with comparatively limited reflective depth, suggesting that generative systems may facilitate emotional participation and visual production while not automatically promoting metacognitive processes.

In this sense, the results suggest that the activity fostered creative production and emotional engagement, but not necessarily deep conceptual understanding. Previous educational research has shown that participation in creative activities alone does not automatically generate meaningful learning, particularly when reflective guidance is limited [7,8,13]. Therefore, the educational value of generative artistic practices appears to depend largely on how these activities are pedagogically structured and mediated.

On the other hand, the predominance of positive and pleasant emotions, especially joy and positive emotional states, is consistent with previous findings in emotional education. Research such as that of Bisquerra [5] indicates that students tend to express positive emotions more easily, whereas negative emotions require safer and more guided contexts for exploration. In this study, although emotions such as fear or frustration emerged—as well as terror expressed through myths such as witches’ covens—their lower frequency suggests a certain avoidance of more complex emotional registers, which may limit the depth of the reflective process.

The variability observed in students’ qualitative responses—from elaborated explanations to minimal expressions such as “Because I like it”—suggests important differences in reflective and discursive competence. From a sociocultural perspective, reflective thinking depends largely on opportunities for dialogue, interaction and guided verbalization within the educational environment [43,44]. Consequently, while generative technologies may facilitate emotional expression and idea generation, reflective depth appears to require explicit pedagogical mediation and structured opportunities for metacognitive dialogue.

These findings also highlight the importance of promoting critical AI literacy within educational contexts that utilise generative systems. Beyond their creative potential, AI tools can reproduce aesthetic biases, stereotypical emotional representations and simplified visual conventions derived from their training data. Consequently, pedagogical mediation is essential not only to foster reflection and emotional dialogue, but also to promote a critical understanding of how algorithmic systems shape processes of visual production and authorship. Furthermore, the ethical implications arising from children’s interaction with generative systems constitute a relevant dimension for future educational research.

Finally, the gender differences observed in emotional expression may be interpreted in line with studies on emotional socialization, which indicate differentiated patterns in emotional expression according to sociocultural variables [45]. These differences should not be interpreted exclusively in terms of gender or socioeconomic background. Other contextual and pedagogical factors may also have influenced students’ participation and emotional expression, including classroom dynamics, teacher facilitation strategies, previous familiarity with digital tools, and differences in instructional mediation across educational settings.

The results related to the image analysis suggest that students tend to represent emotions through intense and prototypical forms, indicating a predominance of the identification and expression of basic emotions over their complex elaboration. This tendency is consistent with Bisquerra’s [3] model of emotional education, which distinguishes between different levels of emotional competence, placing understanding and regulation at more advanced developmental stages.

Likewise, the presence of symbolic resources—such as the projection of emotions onto animals or environments—points to the role of creativity as a mediator in emotional externalization, in line with Eisner [4], who highlights the value of art in representing internal experiences. However, as Perkins [5] warns, creative production does not necessarily imply understanding, which is consistent with the limited reflective depth observed in this study. It should be noted that the figures included in the manuscript primarily serve an illustrative purpose. The interpretation of visual tendencies reported in this study derives from the broader thematic analysis rather than from a formal iconographic analysis of the specific images reproduced in the manuscript.

These findings should be interpreted cautiously, given that the questionnaire was developed specifically for the present exploratory study and was not subjected to a full psychometric validation process.

In this sense, the results reinforce the idea that emotional expression may develop relatively easily in creative contexts, whereas the diversification and understanding of emotions require explicit pedagogical mediation, especially in early educational stages.

The findings confirm the potential of creative practices mediated by generative technologies to foster student participation and emotional expression, in line with previous literature [4,13]. However, they also demonstrate that reflection requires explicit pedagogical mediation, as it does not emerge automatically from creative experience [6,39]. Consistent with Gardner’s [18] perspective, the educational challenge lies not only in promoting activity, but also in ensuring that such activity leads to deep understanding. Furthermore, from the perspective of constructive alignment, meaningful learning has been shown to depend on the articulation between objectives, activities, and assessment [46], reinforcing the need to structure these proposals appropriately.

Overall, the findings suggest that generative artistic technologies may constitute valuable tools for promoting participation and emotional expression in primary education contexts. However, creative engagement alone does not guarantee reflective depth or meaningful understanding. These processes appear to depend fundamentally on pedagogical mediation, reflective scaffolding and structured dialogue practices integrated into the learning design. Consequently, future educational interventions involving generative AI should prioritize not only creative production, but also critical reflection, emotional interpretation and metacognitive development.

6. Conclusions

The present study made it possible to analyze the impact of an educational proposal based on AI-mediated generative art with primary school students, providing relevant evidence regarding both its potential and its limitations in the development of emotional, creative, and reflective competencies.

Regarding the first research question—to what extent does the use of AI-mediated generative art foster student participation, creativity, and emotional expression?—the results show that this type of proposal generates high levels of engagement, perceived creativity, and emotional expression. The high level of student participation demonstrates the workshop’s ability to create a motivating and meaningful learning environment in which students actively engage in the process.

Likewise, students reported high levels of creative engagement and produced a variety of visual representations, suggesting that generative art tools may expand their expressive possibilities and facilitate experimentation. In this sense, technology acts as a mediator that reduces technical barriers and allows students to focus more fully on expression.

Emotional expression was also clearly enhanced, not only in terms of frequency but also in connection with students’ personal experiences. The analyzed productions and justifications show that students not only represented emotions, but also linked them to their own experiences, highlighting the potential of these methodologies as spaces for expression and identity construction. Nevertheless, a predominance of positive emotions, especially joy, was observed, suggesting the need to broaden the work toward greater emotional diversity by more explicitly incorporating complex or ambivalent emotions.

Regarding the second research question—how does the reflective dimension manifest itself in this type of experience?—the results indicate that reflection develops unevenly and is generally less consolidated than the other dimensions. Although some students elaborated complex responses and demonstrated analytical capacity, a significant proportion showed difficulties in deepening the explanation of their decisions, relying instead on brief or underdeveloped responses.

This heterogeneity demonstrates that reflection does not emerge automatically from creative activity but rather requires explicit pedagogical mediation. In this sense, students showed a willingness to explore their emotions and creative processes, but required tools that encouraged verbalization, analysis, and awareness. These findings reinforce the idea that reflection constitutes a competence that must be taught and supported, especially in early educational stages.

Regarding the third research question—what relationship is established between creativity and reflection in the learning process?—the results revealed a positive but weak relationship between both dimensions, indicating that they do not necessarily evolve in parallel. Students may engage actively in creative processes and express emotions effectively without this necessarily implying greater reflective or metacognitive capacity.

This finding is particularly relevant in the context of generative art, where creativity can be understood as a process partially distributed between the subject and the system. The ease with which visually sophisticated results can be generated may facilitate students’ creative engagement and visual expression, but it may also reduce the need for deep cognitive elaboration if not accompanied by adequate mediation. In this regard, high levels of perceived creativity and creative engagement do not always translate into deeper understanding, highlighting the need to differentiate between creative production and meaningful learning.

Furthermore, the results reveal the influence of variables such as age and gender on students’ experiences. Overall, some differences were observed across groups; however, given the exploratory nature of the study and the overlap between institutional, sociocultural, and demographic variables, these patterns should be interpreted cautiously. The study was not designed to isolate the independent effects of age, gender, or school context. Likewise, gender differences in emotional expression point to the influence of emotional socialization processes, reinforcing the need to design inclusive educational proposals that consider different styles of participation and expression.

Based on these findings, several directions for improvement in future implementations are proposed. First, it is necessary to incorporate specific reflective scaffolding strategies, such as open-ended questions, response models, or structured dialogue spaces that encourage conceptual elaboration. Second, it is recommended to promote greater emotional diversity by guiding students in the exploration of complex emotions that broaden their expressive repertoire. Third, it is advisable to maintain spaces for creative flexibility by reducing rigidity in instructions and encouraging autonomous exploration. Finally, the incorporation of collaborative and oral dynamics may contribute to enriching processes of reflection and shared meaning-making.

Overall, the results suggest that AI-mediated generative art may constitute a valuable educational tool for promoting student engagement, perceived creativity, and emotional expression. However, the findings also indicate that reflective and metacognitive processes do not emerge automatically from creative activity and require explicit pedagogical support.

In conclusion, this study contributes to understanding the potential role of generative artificial intelligence in emotional education. Its educational value appears to lie not only in supporting creative engagement and visual expression, but also in the ways these technologies are embedded within pedagogical designs that promote reflection, critical thinking, and meaningful learning. Given the exploratory nature of the study and the characteristics of the instrument employed, these findings should be interpreted cautiously and viewed as a basis for future research rather than as definitive evidence of educational effectiveness.

6.1. Limitations

This study presents several limitations that should be acknowledged. First, the sample size was relatively small and based on intentional sampling, which limits the generalizability of the findings. In addition, the participating schools differed considerably in terms of sociocultural and educational context, including variations in socioeconomic background, school type, and gender composition, which may have influenced the results. The intervention was also developed through a short-term workshop, preventing the analysis of long-term effects on creativity, emotional expression, or reflective processes. Furthermore, some responses may have been affected by social desirability bias, particularly in relation to the expression of positive emotions.

Furthermore, gender composition was closely associated with institutional context, since one of the participating schools consisted exclusively of girls. Consequently, gender-related findings cannot be interpreted independently from school type, socioeconomic background, or educational environment, and should therefore be considered exploratory.

Another limitation of the study concerns the relatively small sample size and the unequal distribution of participants across groups, including one subgroup composed of only six students. These conditions may reduce the robustness and stability of subgroup comparisons and limit the generalizability of the findings. Furthermore, the lack of homogeneity between groups in terms of gender composition, educational context, and sociocultural background may have influenced some of the observed differences. Consequently, the quantitative results should be interpreted cautiously and understood as exploratory rather than confirmatory.

An additional limitation concerns the absence of formal psychometric validation procedures for the questionnaire. Because the instrument was specifically designed for the exploratory purposes of this study, internal consistency indices and construct validation analyses were not conducted. Consequently, the quantitative findings should be interpreted as exploratory indicators rather than as measurements derived from validated scales.

In addition, the study did not systematically control other potentially relevant pedagogical variables, such as teacher facilitation styles, classroom interaction dynamics, or students’ previous familiarity with AI-based digital tools, which may also have influenced the observed differences between groups.

Finally, the absence of a longitudinal design limits the possibility of examining the stability and evolution of these processes over time.

6.2. Future Research Directions

Future research should explore longitudinal approaches to examine the sustained impact of AI-mediated generative art on emotional, creative, and reflective development. Comparative studies involving different educational stages, sociocultural contexts, and instructional models would also contribute to a deeper understanding of these practices. In addition, future work could investigate multimodal forms of reflection that combine verbal, visual, and collaborative dimensions. Further attention should also be given to the development of AI literacy and critical understanding of generative systems in educational settings. Finally, collaborative generative art experiences may provide valuable opportunities to explore shared creativity, dialogue, and collective meaning-making processes.

6.3. Practical Implications

The results suggest that generative artificial intelligence may be a valuable educational tool for fostering creativity, emotional expression and critical reflection in educational settings. However, its integration into the classroom should not be approached from a technocentric or purely instrumental perspective, but rather as a mediating tool that requires pedagogical support, didactic structuring and ethical guidance from teachers. In this regard, the study’s findings allow for the identification of various practical implications for the educational sector.

Applications for developing emotional intelligence in primary education

The recognition and verbalisation of emotions, both one’s own and those of others, form the basis of emotional intelligence (understood as a skill that can be acquired). Due to the very nature of human neurological development, metacognition regarding emotional states requires explicit teaching. In line with the classic self-report models derived from the principles of cognitive-behavioural therapy, it is necessary to structure three sets of questions: (a) the preceding situation, (b) recognition and expression of the emotion, and (c) management of the emotion.

(a) Background: this helps to identify and contextualise the events that trigger emotional changes; therefore, the questions for the educational intervention would be:

What were you doing before you felt this emotion? Were you alone or with other people? What emotion were you experiencing before?

(b) Recognition and expression of emotion: verbalising and understanding one’s own and others’ emotional states is complex during childhood and adolescence. Visual representation can help students express their emotional state and communicate it to others, explicitly working on the principles of cognitive empathy. Some questions might be: Why did you choose this emotion? Which visual elements best represent your feelings? Is there a connection between the actual emotion and the image created? How would the meaning of the image change if we altered certain symbols or colours?

(c) Emotion management: learning to detach oneself from an unpleasant emotion is a complex process that usually requires extensive experience. However, as emotional intelligence can be developed and cultivated, it is possible to begin addressing this in primary education. This is essential, as it ensures that behaviour is not driven by impulsivity, but by reasoning and metacognition. However, the emotion felt mustn’t be underestimated during this process. Some questions might be: “I understand that you felt angry when your classmate took your pencil case and broke your paints. You take very good care of all your materials and neither approve of nor engage in such behaviour. Would you like to express and tell your classmates how you felt? And now, how would you like to feel? What could you do to feel a different emotion? Would you like to calm down again? What can you do if a situation like this happens again?

This type of activity can be particularly useful during periods of developmental change, such as the transition from childhood to adolescence, or in response to new or stressful situations that have caused an emotional reaction or distress. In this way, using images, they can express their emotional state, share it with others and understand that of their peers.