Enhancing Student Motivation and Engagement Through the Use of a Slovenian-Speaking Social Robot AlphaMini

Abstract

The integration of Artificial Intelligence into education is transforming how abstract and complex concepts are delivered, especially through embodied tools like social robots. This study examines the impact of AlphaMini, a Slovenian-speaking social robot supported by model ChatGPT 4o and trained on structured book-based content, on student engagement during knowledge management lessons. A case study approach was used, including student questionnaires, classroom observations, and post-session discussions, with 70 university students from diverse academic fields. Engagement was assessed across behavioral, emotional, and cognitive dimensions, with comparisons based on prior robot experience. Results show AlphaMini significantly enhanced emotional and behavioral engagement, with moderate cognitive gains. Students familiar with social robots demonstrated higher engagement, interacting more naturally and actively. Informal feedback highlighted positive attitudes toward AlphaMini, especially among students who regularly use generative AI tools like ChatGPT or Copilot. Participants appreciated its human-like gestures, Slovenian language use, and emotionally supportive presence. Many suggested its potential use in primary and inclusive education, where emotional safety and playful interaction are crucial. This study contributes to the growing evidence on AI in education, showing that combining generative AI with social robotics can foster motivation, participation, and emotionally rich learning experiences.

1. Introduction

The integration of Artificial Intelligence (AI) technologies into education is reshaping pedagogical practices, offering innovative and interactive learning experiences. Among these technologies, social robots supported with the generative AI, especially those capable of communicating in the learner’s native language, are proving to be powerful tools for increasing student engagement and consequently their motivation. AI is increasingly involved in many aspects of modern society, influencing education, business, and daily life (Jagodič & Šinkovec, 2021). One such example is AlphaMini, a compact social robot equipped with speech, facial recognition, and emotional expression, capable of interacting in Slovenian language. Social robots like AlphaMini and NAO are designed to facilitate human-like interaction using speech, gestures, facial expressions, and affective cues. While they do not fully replicate human appearance, they are engineered to simulate a sense of social presence and the impression of being in the company of another social being, which helps build emotional bonds and peer-like relationships with students (Breazeal, 2003; Belpaeme et al., 2018). Studies have demonstrated that social robots function effectively as peer tutors, especially in language learning and STEM education (Kanda et al., 2004; Augello et al., 2020). Their repeatable, non-judgmental, and emotionally intelligent interactions can support learning in a way that traditional digital tools or static media cannot. Teachers’ perspectives also shape how and where robots are adopted in classrooms, with recent evidence showing generally positive attitudes toward humanoid robots in education (Flogie et al., 2025).

In Slovenia, the limited number of localized AI tools makes AlphaMini’s Slovenian-language capabilities particularly relevant for inclusion and equity (Nityashree et al., 2023). Robots can also support 21st-century skill and competence development, including collaboration, critical thinking, and digital literacy (Aberšek & Flogie, 2022). Teachers’ creativity and readiness to innovate are equally critical, as recent studies show that innovative training models can strengthen teachers’ ability to integrate knowledge management and digital tools into practice (Skrbinjek et al., 2024). This ability makes it particularly effective in explaining abstract and complex topics, e.g., those related to knowledge management, such as knowledge management processes and knowledge management systems, by articulating them clearly in Slovenian.

Despite the increasing presence of generative AI in educational environments, empirical studies that systematically evaluate its pedagogical impact remain limited (Zhang & Aslan, 2021; Augello et al., 2020). Current research in AI in education predominantly focuses on adaptive learning systems, intelligent tutoring, and chatbot-based writing support (Zawacki-Richter et al., 2019; Holmes et al., 2022). In contrast, significantly fewer studies examine embodied AI, particularly social robots, as instructional tools in formal educational settings (Belpaeme et al., 2018; Mubin et al., 2013). The research exploring how social robots impact multiple dimensions of students’ behavioral, emotional, and cognitive engagement, especially when teaching abstract or conceptually demanding content to diverse student populations, is even more limited (Fridin, 2014; Augello et al., 2020).

Social robots introduce a physical, embodied element of interaction that distinguishes them from screen-based generative AI tools. Their human-like expressiveness, gestures, and the ability to maintain eye contact create a sense of social presence, which can be critical for emotional and behavioral engagement (Belpaeme et al., 2018). When powered by generative AI like ChatGPT, social robots can move beyond scripted dialogues to provide dynamic, context-aware responses, making them particularly useful in exploratory and discussion-driven learning scenarios. The combination of generative AI’s linguistic capabilities with the robot’s physical embodiment enables a more natural, human-like interaction, which has been shown to enhance motivation, trust, and inclusivity in educational contexts (Fridin, 2014; Kanda et al., 2004).

This study was conducted within the framework of the national project Innovative Pedagogy 5.0, funded by the Ministry of Education of the Republic of Slovenia as part of the Recovery and Resilience Plan. The project focuses on developing digital competencies and integrating innovative technologies into Slovenian schools. Given the limited availability of localized AI tools in Slovenian and the strong policy emphasis on inclusion and digital transformation, the Slovenian context provides a distinctive case for examining the role of a Slovenian-speaking social robot in education. Moreover, as a small-language environment, Slovenia can serve as a model for other countries facing similar challenges in adapting generative AI and robotics to local linguistic and cultural needs. Slovenia represents a distinctive sociocultural context for the integration of advanced digital technologies in education. According to the European Commission’s Digital Decade Country Report (European Commission, 2024), Slovenia performs close to the EU average in most digital indicators, suggesting both possibilities for making progress and addressing the remaining gaps, making it a relevant testbed for innovative practices. Furthermore, the OECD (2022) highlighted Slovenia’s efforts to digitalize its economy and education, stressing that systemic investments in digital infrastructure have created a solid foundation for technology-enhanced learning.

Slovenia also plays a unique role internationally: it hosts the International Research Centre on Artificial Intelligence (IRCAI), under the auspices of UNESCO, which further emphasizes its commitment to advancing responsible AI adoption (IRCAI, n.d.). In addition, Slovenia is home to the Vega supercomputer, one of Europe’s most powerful high-performance computing systems, enabling large-scale experimentation and research in AI and digital education (EuroHPC Joint Undertaking, n.d.).

Taken together, these factors show that Slovenia occupies a middle but strategically promising position in digital education integration: it is neither a late adopter nor a global leader, but its infrastructure, policy environment, and cultural orientation toward innovation make it an exemplary case for exploring the educational potential of social robots.

This study addresses this research gap by exploring how a Slovenian-speaking social robot, supported by generative AI, impacts student motivation and engagement when learning complex and abstract topics such as knowledge management.

Based on established theories of engagement and human–robot interaction, we formulated the following research questions (RQ) and corresponding hypotheses (H). Each RQ is accompanied by a short explanation of its aim.

1.

RQ1: How does the presence of the AlphaMini social robot in the classroom engage students (cognitive, behavioral, and emotional engagement)?

This question examines how the use of a social robot as an instructional tool enhances overall student engagement. It is grounded in the assumption that embodied AI tools may promote attention, participation, emotional involvement, and cognitive investment during abstract learning tasks.

H1. 

The use of the AlphaMini social robot will significantly increase students’ emotional and behavioral engagement, with moderate improvement in cognitive engagement.

2.

RQ2: What differences in student engagement can be observed between students who have previously interacted with social robots and those who have not?

This question investigates whether familiarity with social robots influences students’ openness, comfort, and engagement during interaction. Prior exposure might reduce hesitation and increase responsiveness, resulting in higher engagement.

H2. 

Students with prior experience with social robots will report higher emotional engagement.

H3. 

Students with prior experience will report higher behavioral engagement.

H4. 

Students with prior experience will report higher cognitive engagement.

The structure of this article is as follows: In the next section, we present the theoretical background and literature review, focusing on learning engagement and motivation in educational psychology and human–robot interaction. This is followed by research methodology, an overview of the results, and a discussion of the findings. The article concludes with practical implications and suggestions for further research.

2. Theoretical Foundations and Literature Review

The use of social robots in educational settings, such as the AlphaMini robot, is grounded on prior research in educational psychology and human–robot interaction. The broader trend of AI-enhanced education aims to personalize learning, increase student motivation, and improve engagement. However, unlike conventional AI tools like chatbots or adaptive learning platforms, social robots introduce a physical, embodied presence in the classroom. They combine verbal interaction, non-verbal cues (such as gestures and facial expressions), and real-time responsiveness, which together create a rich, multisensory learning experience. This human-like interaction supports students’ emotional connection, fosters social presence, and enhances attention and participation, making social robots especially effective in teaching abstract or unfamiliar content (Belpaeme et al., 2018; Breazeal, 2003; Kanda et al., 2004; Mubin et al., 2013; Fridin, 2014). Furthermore, social robots can act as learning companions or peer tutors, offering consistent and non-judgmental support, which is particularly beneficial for learners who experience anxiety, lack of motivation, or have special educational needs. In this way, social robots bridge the gap between digital technology and human interaction, as they offer a hybrid form of support that taps into the motivational and cognitive benefits of both (Crossman et al., 2018; Kouroupa et al., 2022). Their use is especially relevant in knowledge domains where engagement and conceptual understanding are difficult to achieve through traditional lecture-based instruction alone. Drawing from educational psychology and technology-enhanced learning engagement is recognized as a multidimensional construct, encompassing behavioral, emotional, and cognitive components (Fredricks et al., 2004; Skinner et al., 2009). Behavioral engagement refers to students’ participation in academic tasks, including effort, persistence, attention, and classroom involvement (Fredricks et al., 2004). The robot’s verbal cues, gestures, and real-time feedback mechanisms may help prompt learners to stay attentive and complete tasks. Emotional engagement includes affective responses to learning, such as interest, enthusiasm, a sense of belonging, and emotional connection to the learning environment (Skinner et al., 2009; Salmela-Aro, 2012). Research shows that students are more emotionally engaged when they feel supported and safe. AlphaMini’s ability to communicate in Slovenian, display emotional expressions, and offer social presence creates an emotionally supportive environment, which may foster positive affect and engagement (Belpaeme et al., 2018; Salmela-Aro, 2012). Cognitive engagement refers to the mental effort students invest in understanding complex concepts, characterized by self-regulation, deep learning strategies, and metacognitive processing (Appleton et al., 2006). Engagement is deeply tied to motivation, particularly intrinsic motivation, which drives sustained learning and academic achievement. According to Self-Determination Theory (Ryan & Deci, 2000), motivation flourishes when students experience autonomy, competence, and relatedness. AI-powered social robots offer autonomous learning pathways, timely feedback, and personalized interactions, which have been shown to enhance learners’ motivational states (Fridin, 2014; Benitti, 2012).

In classrooms where content is abstract or perceived as disconnected from daily experience, the novelty and interactivity of social robots can reintroduce relevance and enjoyment, the two key factors in sustaining motivation (Ayeni et al., 2024; Zhang & Aslan, 2021). By making learning tasks more accessible and responsive, AlphaMini increases the likelihood that students remain engaged. Previous studies have shown that social robots exert their most pronounced effects on emotional and behavioral engagement, whereas cognitive improvements typically emerge more gradually through repeated exposure. For instance, Fridin (2014) demonstrated that children interacting with a social robot during storytelling displayed heightened enjoyment and participation, underscoring the importance of emotional safety in learning contexts. Similarly, Belpaeme et al. (2018) and Kanda et al. (2004) found that embodied features, such as gestures, facial expressions, and the direction of the gaze, stimulate behavioral responses such as attention, persistence, and peer discussion.

While prior research has addressed general applications of robotics in education, fewer studies have focused on Slovenian-language social robots delivering book-based, conceptually dense content such as knowledge management. In addition, prior research highlights the importance of technology familiarity in shaping learner responses. Woolf (2009) and Zhang and Aslan (2021) reported that students who have previously engaged with AI-based tools or robots tend to adapt more quickly, exhibit reduced hesitation, and display stronger openness to robot-assisted learning (Chen et al., 2023; Kory-Westlund & Breazeal, 2019).

This study adds to the growing body of research by empirically evaluating how AlphaMini supports behavioral, emotional, and cognitive engagement through structured interactions grounded in verified knowledge sources. It builds on the theoretical foundations of engagement, motivation, and sociable robotics to offer practical insights for the future of AI-enhanced teaching.

3. Research Methodology

This study employs a case study research design, complemented by quantitative methodology and structured classroom observation. To address RQ2, participants were divided into two groups based on their self-reported prior experience with social robots. For each group, average engagement scores were calculated, and comparative statistical analyses (Mann–Whitney U test) were conducted to determine the significance of observed differences.

3.1. Sample and Procedure

A total of 70 graduate students voluntarily participated in the study. All participants were enrolled in graduate-level programs and were selected using purposeful sampling based on their enrollment in courses related to education, management, or technology, where the use of AI and innovation in education was thematically relevant. Participants were drawn from two Slovenian higher education institutions:

• Private institution, International School for Business and Social Studies in Celje—25 participants
• Public institution, Faculty of Natural Sciences and Mathematics in Maribor—45 participants

The data was collected between May and July 2025. In Celje, the study was conducted directly in the classroom, where students interacted with the AlphaMini robot during a live instructional session supported by a Slovenian-language version of ChatGPT. The session was guided by the teacher. Following the session, students completed an online survey about their engagement experience. In Maribor, students had already been introduced to AlphaMini approximately three months earlier during a prior course module. For this study, they were asked to complete the same engagement questionnaire without repeated live interaction, to allow for comparison based on prior exposure to the robot. This structure enabled a comparison between students with recent, direct interaction (Celje) and those with earlier, prior exposure (Maribor), allowing the analysis to investigate how both current experiences and familiarity influence student engagement across behavioral, emotional, and cognitive dimensions.

3.2. Learning Session Description and Background

The learning session followed three main stages. First, the teacher covered the content topic on knowledge management. After they were familiar with the content, the teacher introduced AlphaMini and gave students instructions on how to interact with the robot. In the second stage, students freely engaged with AlphaMini. This activity was intentionally exploratory, with students encouraged to naturally interact with the robot. Finally, in the third stage, a group discussion was held, where students shared their impressions, highlighting both positive and negative aspects of using a social robot in the classroom. These reflections and observations were collected and presented as part of a case study. It should be noted that this was not a traditional lesson with a detailed lesson plan. Instead, the activity was designed as an exploratory session.

To support natural interaction in Slovenian, we developed custom modules for automatic speech recognition (ASR), speech detection, and speech synthesis. These modules enabled AlphaMini to process students’ spoken queries, forward them to ChatGPT for generating responses, and vocalize the answers in Slovenian. This allowed the robot to function as a Slovenian-speaking conversational partner, a capability not natively available in the device. Students were free to formulate their own questions, most of which were related to management and economics, consistent with the course.

3.3. Ethics Considerations

This study was approved by the Ethics Committee of the Institute of Contemporary Technologies, Faculty of Natural Sciences and Mathematics, University of Maribor. Prior to participating in the survey, all participants were provided with detailed information about the purpose, procedures, and nature of the study. Participation was voluntary, and participants had the right to withdraw at any time. Confidentiality of responses was guaranteed. Written informed consent has been obtained from the participants to publish this paper.

3.4. Measurement and Instrument Description

The questionnaire was structured around three key components of learning engagement: behavioral engagement, emotional engagement, and cognitive engagement. This instrument is based on validated constructs of engagement from Fredricks et al. (2004), Skinner et al. (2009), and Appleton et al. (2006), adapted for the Slovenian context and topic area. All three forms of engagement were measured on a 5-point Likert scale with responses ranging from 1 = “strongly disagree” to 5 = “completely agree”.

Behavioral engagement was measured by averaging participants’ responses across four items, with higher values indicating greater engagement. The scale shows excellent internal reliability (Cronbach’s alpha = 0.95). Emotional engagement was calculated by averaging the responses across 6 items, where higher values indicate greater engagement. The scale shows good internal reliability (Cronbach’s alpha = 0.92). Cognitive engagement was calculated by again averaging the responses across 5 items, with higher values reflecting greater engagement. The scale shows excellent internal reliability (Cronbach’s alpha = 0.97).

In addition, sociodemographic characteristics of the respondents were included, including gender (1 = male, 2 = female, 3 = do not want to answer), age (years), field of study (1 = Basic and general educational activities/outcomes, 2 = Educational sciences and teacher education, 3 = Arts and humanities, 4 = Social sciences, journalism and information science, 5 = Business and Administrative Sciences, law, management, 6 = Natural sciences, mathematics, and statistics, 7 = Information and communication technologies (ICT), 8 = Engineering, manufacturing technologies, and construction, 9 = Agriculture, forestry, fisheries, and veterinary science, 10 = Health and social welfare, 11 = Transport, security, hospitality, tourism, and personal services) and respondents’ current employment status with the question “Are you employed?” (1 = yes, 2 = no).

3.5. Methods

The quantitative data collected through the questionnaire were analyzed with descriptive statistics (mean, standard deviation, frequency distributions) for each engagement item and with comparative analysis (e.g., grouped responses regarding prior experiences with the robot).

In addition, we conducted non-intrusive observations focused on students’ observable engagement behaviors, such as:

• verbal interaction with the social robot,
• peer collaboration during social robot-led tasks,
• attentiveness and participation,
• emotional responses (e.g., laughter, curiosity, hesitation).

Observation data were used descriptively to contextualize the findings. Observation notes were guided by a structured observation protocol aligned with the three dimensions of engagement and served as supporting evidence to interpret quantitative results. This comparison was not tested statistically but served as a qualitative cross-check to support the questionnaire data.

3.6. Descriptive Statistics

From

Table 1. Number and Percentage of Students by Field of Study.

	N	%	M (SD)
Gender (total)	68	97.2
Male	10	14.3
Female	58	82.9
Age	70		24.9 (11.4)
Field of study (total)	70	100
Basic and general educational activities/outcomes	17	24.3
Educational sciences and teacher education	28	40
Arts and humanities	2	2.9
Social sciences, journalism, and information science	1	1.4
Business and Administrative Sciences, law, and management	22	31.4
Employment status (total)	70	100
Employed	27	38.6
Not (yet) employed	43	61.4

, the average age of participants was 24.9 years, slightly younger than the national average for Slovenian graduate students, which is 26.5 years. According to the Statistical Office of Slovenia (SORS), the average age of graduate students in 2022 was 26.3 years. Broader OECD data, which include all types of graduate students, indicate an average age of 25.2 years for Slovenia in 2022 (OECD, 2024). According to SORS, 57.9% of all tertiary students were women in the academic year 2023/24. The gender imbalance is even more pronounced at the graduate level: in uniform graduate programs, women represented 73.1% of students, while in graduate studies following an undergraduate degree, 61.8% were women (SORS, 2024). These data confirm that female predominance is especially strong in fields such as education, social sciences, and management, which aligns with the disciplinary composition of our sample. In terms of gender, a large majority of respondents identified as female (82.9%, n = 58), while 14.3% (n = 10) identified as male. A small proportion (2.9%, n = 2) preferred not to disclose their gender. This gender distribution aligns with broader enrollment trends in education and social sciences, where female participation is often higher.

Participants represented a variety of academic disciplines, primarily from the fields of educational sciences, teacher education, management, and business-related programs. The largest group (40%, n = 28) came from the field of educational sciences and teacher education, followed by 31.4% (n = 22) from business, administrative sciences, law, and management. The remaining participants were enrolled in programs categorized as basic and general education (24.3%, n = 17), with a small number from arts and humanities (2.9%, n = 2) and social sciences, journalism, and information science (1.4%, n = 1). This disciplinary composition was in line with our sample’s preferences. Students in educational sciences were included because they represent future educators and are directly relevant to evaluating the pedagogical potential of social robots. Their perspectives provide insight into how such technologies might be implemented in classroom practice. Students from business and management programs were included due to the thematic focus of the robot-led session, which addressed concepts from knowledge management, a subject commonly taught in organizational, business, and managerial curricula. These students were thus expected to have foundational familiarity with the content, allowing the study to isolate the effect of the form of delivery (i.e., robot interaction) on engagement, rather than differences in prior subject knowledge. The inclusion of students from two faculties with complementary disciplinary orientations enabled a more diverse and representative analysis of how social robots influence engagement across both educational and managerial learning contexts. This also reflects the growing interdisciplinary relevance of AI in education, business, and communication settings.

Regarding employment status, 61.4% of participants (n = 43) reported that they were not (yet) employed, while 38.6% (n = 27) were currently employed. This distribution reflects the transitional stage many students are in during or shortly after their graduate studies, which may influence their availability, learning motivations, and responsiveness to educational innovations such as the use of social robots.

3.7. Observational Procedure and Discussion Notes

In addition to the survey, we included a basic exploratory observational component. During the live sessions, we acted as observers and kept handwritten notes. These notes captured students’ visible reactions to the robot, such as laughter, hesitation before speaking, leaning forward, spontaneous questions, or group discussion. Particular attention was given to behaviors and comments that corresponded to the three engagement dimensions (behavioral, emotional, and cognitive). At the end of the activity, the supervisor facilitated an open class discussion (10–15 min) in which students reflected on their experience with AlphaMini and shared their opinions about the potential role of AI and robots in future education. Key statements from this discussion were also written down. Cleaning consisted of removing unclear or repeated statements, clustering related observations, and summarizing representative student behaviors or comments (Miles et al., 2014). The data were then thematically grouped into behavioral, emotional, and cognitive categories, in line with established frameworks of student engagement (Fredricks et al., 2004; Skinner et al., 2009).

3.8. Data Management and Descriptive Analysis

Handwritten observation notes were subsequently digitized and analyzed using an interpretative approach. Following the principles of reflexive thematic analysis (Braun & Clarke, 2006, 2021), two researchers independently reviewed the notes, generated initial codes aligned with the three engagement dimensions (behavioral, emotional, and cognitive), and aimed to reconcile interpretations through discussion, thereby enhancing confirmability (Lincoln & Guba, 1985). Codes were iteratively grouped into themes that summarized repeated behaviors (e.g., leaning forward, spontaneous questions, peer discussion) and representative comments. To strengthen validity, we applied methodological triangulation by comparing these qualitative patterns with quantitative survey results, examining points of convergence and divergence (Creswell & Plano Clark, 2018; Fetters et al., 2013). Our observation protocol and reporting adhered to established guidance for structured classroom/field observations and qualitative transparency (Spradley, 1980; Miles et al., 2014; O’Brien et al., 2014). No qualitative software was used; instead, coding was conducted in a spreadsheet with an auditable trail of category decisions and memo notes.

3.9. Reliability and Validity Considerations

To strengthen qualitative reliability and validity, we adopted several strategies. First, the observational notes were independently reviewed and discussed by the research team to reduce individual bias and enhance confirmability (Lincoln & Guba, 1985). Second, triangulation was applied by comparing observations, open-ended feedback, and survey data to identify convergence or divergence (Creswell & Plano Clark, 2018; Fetters et al., 2013). Third, transparency of procedures was increased by explicitly reporting how data were collected, cleaned, and thematically organized (O’Brien et al., 2014). The interpretative and reflexive approach was applied, following Braun and Clarke (2021), acknowledging subjectivity while emphasizing transparency and team discussion as validity-enhancing practices. While qualitative analysis does not aim at replicability in the strict quantitative sense, we enhanced dependability by providing a clear audit trail of coding decisions, theme development, and researcher memos. This means that another researcher, given access to the same observation notes and coding documentation, could follow the same analytic process and arrive at comparable thematic structures, even if not identical wording of categories. Such transparency increases the trustworthiness of the findings and allows future studies to replicate the procedure in similar classroom contexts.

4. Results

The implementation of the AlphaMini social robot produced encouraging results that emphasize its positive view on student engagement and motivation.

The correlation analysis from

Table 2. Spearman Correlation Matrix Between Dimensions of Student Engagement.

	1	2	3
1. Behavioral engagement	1	0.82 **	0.73 **
2. Emotional engagement		1	0.79 **
3. Cognitive engagement			1

Note: ** p < 0.01.

revealed strong, statistically significant relationships between all three dimensions of student engagement. As shown in Table 2, behavioral engagement was highly correlated with both emotional engagement (Rho = 0.82, p < 0.01) and cognitive engagement (Rho = 0.73, p < 0.01). Similarly, emotional and cognitive engagement demonstrated a strong correlation (Rho = 0.79, p < 0.01). While the correlation analysis (Table 2) highlights the interrelatedness of engagement dimensions, the robot’s impact is more directly evidenced by the increased engagement scores in

Table 3. Mean Differences in Engagement Based on Prior Experience with a Social Robot.

	Experience with Social Robot
	Yes	No
Variables	M (SD)	M (SD)
Behavioral engagement	3.82 (0.7)	2.65 (1.2)
Emotional engagement	3.48 (0.36)	2.49 (1.23)
Cognitive engagement	3.28 (0.73)	2.40 (1.31)

Note: M = Mean, SD = Standard Deviation.

and

Table 4. Mann–Whitney U Test: Engagement Differences Based on Prior Experience with a Social Robot.

	Experience with Social Robot
	Yes (N = 22)	No (N = 48)
Variables	Mean Rank	Mean Rank	U	Z	p
Behavioral engagement	51.27	28.27	181	−4.433	<0.0001
Emotional engagement	50.95	28.42	188	−4.374	<0.001
Cognitive engagement	45.43	30.95	309.5	−2.799	0.005

N = number of respondents, U = Mann-Whitney test statistic, Z = Z-score, p = statistical significance.

and by qualitative observations of student interaction and feedback. These findings also indicate that the three engagement dimensions are closely interconnected and tend to co-occur, suggesting that students who are emotionally invested in the learning experience are also more likely to exhibit behavioral participation and cognitive involvement.

To explore whether prior experience with social robots influenced students’ engagement levels, descriptive statistics were calculated for two groups: those with previous experience and those without. As shown in Table 3, students who had previously interacted with a social robot reported higher engagement across all three dimensions. Specifically, participants with prior experience scored higher in behavioral engagement (M = 3.82, SD = 0.70) compared to those without such experience (M = 2.65, SD = 1.20). A similar pattern was observed for emotional engagement (M = 3.48 vs. 2.49) and cognitive engagement (M = 3.28 vs. 2.40). Although this difference was statistically significant (see Table 4), the absolute scores remained lower than those observed for emotional and behavioral engagement. This suggests that AlphaMini primarily supports affective and participatory aspects of learning, while cognitive gains require deeper or repeated exposure to robot-assisted activities. The results align with prior research indicating that improvements in higher-order cognitive engagement typically emerge gradually, rather than after a single, exploratory session (Fridin, 2014; Belpaeme et al., 2018).

These descriptive results suggest that prior familiarity with social robots may positively influence how students respond to robot-assisted learning activities.

To examine whether prior experience with social robots significantly affected levels of engagement, a non-parametric Mann–Whitney U test was conducted for behavioral, emotional, and cognitive engagement. As presented in Table 4, statistically significant differences were found across all three dimensions. Students with prior experience showed significantly higher behavioral engagement (U = 181.00, Z = −4.433, p < 0.0001), emotional engagement (U = 188.00, Z = −4.374, p < 0.001), and cognitive engagement (U = 309.50, Z = −2.799, p = 0.005) compared to students without prior experience. The implementation of the AlphaMini social robot yielded promising findings that support its positive influence on student engagement. Results are structured according to the tested hypotheses (H1–H4) and grouped around the two core focus areas (RQ1 and RQ2).

H1/RQ1: Spearman correlation analysis (Table 2) revealed strong, statistically significant relationships between all three engagement dimensions:

• Emotional and behavioral engagement: r = 0.89, p < 0.01
• Emotional and cognitive engagement: r = 0.92, p < 0.01
• Behavioral and cognitive engagement: r = 0.84, p < 0.01

These results confirm that emotional engagement tends to co-occur with both behavioral and cognitive involvement, supporting the assumption that the robot stimulated students across multiple engagement domains. Observations further confirmed these effects: students maintained eye contact, leaned forward, responded verbally to the robot, and participated actively in small-group discussions. Emotional responses such as laughter, surprise, and positive verbal feedback were noted during the session. Cognitive engagement was present through question-asking and peer discussion, though less consistently expressed across all students.

Interestingly, several participants suggested that social robots like AlphaMini might be especially beneficial in elementary schools or in educational settings for children with special needs who exhibit anxiety or communication difficulties. They noted the robot’s friendly appearance, calm voice, and expressive gestures as features that could promote emotional safety and engagement among younger or more vulnerable learners. Although some initial discomfort and novelty-induced hesitation were observed, students adapted quickly and expressed appreciation for the innovative, human-centered use of AI.

Overall, findings support H1.

H2–H4/RQ2: Descriptive analysis (Table 3) indicated that students with prior experience working with social robots consistently reported higher levels of engagement:

• Behavioral: 3.82 vs. 2.65
• Emotional: 3.48 vs. 2.49
• Cognitive: 3.28 vs. 2.40

These differences were confirmed using the Mann–Whitney U test (Table 4). All comparisons were statistically significant:

• H2 (emotional): U = 188.00, Z = −4.374, p < 0.001
• H3 (behavioral): U = 181.00, Z = −4.433, p < 0.0001
• H4 (cognitive): U = 309.50, Z = −2.799, p = 0.005

Observational notes documented that survey data further support these observations. Among the behavioral engagement items, the highest-rated statement was “Learning with the social robot increased discussion within the group”, which received the strongest agreement across participants. This finding corroborates the classroom observations, where students spontaneously leaned in, maintained eye contact, and engaged in peer discussion during the activity. Several participants emphasized that AlphaMini’s calming voice, friendly appearance, and expressive gestures were particularly helpful for students who might experience anxiety, communication difficulties, or lower self-confidence. These impressions suggest that students perceived the robot as emotionally supportive and inclusive. During the end-of-session discussions, participants emphasized that AlphaMini’s Slovenian language ability made the interaction feel more personal and inclusive. One student noted: “When the robot spoke Slovenian, it felt like it belonged in our classroom, not just like a foreign gadget”. Several students also suggested potential future applications, particularly in elementary schools and inclusive education, where they believed the robot’s friendly appearance and emotionally supportive presence could be especially valuable. As one participant commented, “Children with anxiety or special needs could really benefit from such a calm and friendly robot”. Open-ended survey responses further confirmed these impressions, highlighting novelty, fun, and curiosity as recurring themes: “It was exciting and fun, I’ve never seen a robot answer me in Slovenian before”. These findings also show that prior familiarity with social robots has a strong and measurable effect on all three engagement dimensions. Participants who had previously interacted with AlphaMini adapted faster, showed fewer signs of hesitation, and engaged more naturally and actively. One such student explained, “Because I had already used a robot in another course, I wasn’t shy, I just started asking questions”. In contrast, students without prior experience admitted initial hesitation before warming up, “At first, I didn’t know what to say, but then I joined in when others started”.

5. Discussion

Our findings align with previous research demonstrating that the most immediate and visible effects of social robots are observed in emotional and behavioral engagement, while cognitive improvements emerge more gradually through repeated exposure and sustained interaction. As Fridin (2014) showed in a storytelling context, social robots can rapidly foster enjoyment and active participation, creating a sense of emotional safety that encourages students to take part without fear of judgment. Similarly, Belpaeme et al. (2018) and Kanda et al. (2004) highlighted that embodied features such as gestures, gaze, and expressive facial cues stimulate behavioral responses, including attention, persistence, and peer discussion. In contrast, the development of deeper cognitive engagement, such as perspective-taking, critical reflection, and conceptual integration, tends to require longer exposure to robot-assisted learning. This distinction underscores the importance of considering not only the immediate motivational impact of social robots but also their potential for fostering sustained cognitive growth when embedded more systematically into instructional design. The observed impact aligns with existing literature that emphasizes the importance of interactivity, embodiment, and emotional resonance in educational technology (Fredricks et al., 2004; Belpaeme et al., 2018). Results clearly support the underlying assumptions of both RQ1 and RQ2 and demonstrate the unique contribution of social robots to fostering a multidimensional learning experience.

Following RQ1, quantitative results indicate that AlphaMini primarily enhanced emotional and behavioral engagement. As shown in Table 2, behavioral and emotional engagement were strongly correlated (r = 0.82, p < 0.01), and observational notes described visible signs of positive interest. The highest-rated item in the behavioral scale was “Learning with the social robot increased discussion within the group,” confirming that the robot stimulated active participation. Together, these findings suggest that AlphaMini fostered both emotional resonance and behavioral involvement, consistent with H1.

The effect of prior experience provided clear differences in engagement between students with and without prior experience using social robots. Participants with previous experience reported higher engagement scores across all three dimensions (behavioral M = 3.82 vs. 2.65; emotional M = 3.48 vs. 2.49; cognitive M = 3.28 vs. 2.40). These differences were statistically significant in the Mann–Whitney U test (Table 4), confirming H2–H4. Observational notes supported this pattern: experienced students engaged without hesitation (“I just started asking questions”), while novices initially expressed uncertainty (“At first, I didn’t know what to say”) before gradually warming up. These data directly support the interpretation that familiarity with robots reduces barriers to engagement and enhances natural interaction.

Emotional engagement was especially strong throughout the learning session, as supported by our observation notes. Students consistently responded with visible signs of interest, surprise, and curiosity, particularly when the robot spoke Slovenian or performed expressive gestures. Several participants expressed how novel and exciting it was to interact with a robot in their native language. Such reactions support theories suggesting that personalization and cultural relevance can increase emotional connection with learning technologies. Despite some initial hesitation among students with no prior experience, emotional barriers were quickly overcome, and most students appeared comfortable, intrigued, and even amused by the robot’s behavior. This sense of emotional safety and enjoyment is critical for fostering positive attitudes toward abstract content and aligns with frameworks emphasizing the importance of emotional presence in education (Salmela-Aro, 2012).

In terms of behavioral engagement, the robot successfully promoted active participation, including verbal responses, body orientation toward the robot, and group discussion. The highest-rated survey statement, “Learning with the social robot increased discussion within the group,” corroborates these observations. Students spontaneously leaned in, made eye contact with the robot, laughed at its expressions, and responded both verbally and nonverbally. These behaviors reflect an environment conducive to participation and social interaction. Such responses also confirm that embodied agents can function as powerful social stimuli that support attention and classroom dynamics, which are central to behavioral engagement (Fredricks et al., 2004). Cognitive engagement, while less pronounced, showed potential. Students reported that the activity prompted them to reflect and approach content from a new perspective, though the depth of conceptual learning appeared limited, likely due to the short duration and exploratory nature of the session. For instance, one of the cognitive engagement items, “Using the social robot motivated me to further explore the topic,” received only moderate agreement (M = 3.12, SD = 0.94), indicating that although curiosity was sparked, deeper conceptual learning would require repeated or extended exposure. And while some signs of cognitive engagement were present, such as reflection and perspective-taking, the observed impact was more modest. This suggests that deeper cognitive development may require repeated or sustained exposure to robot-assisted activities, particularly when dealing with abstract or complex content. If this is the case, the implication is that AlphaMini and similar social robots should not be seen as standalone tools for developing higher-order thinking, but rather as complementary support that initiates curiosity and lowers affective barriers. Sustained integration into the curriculum, with repeated opportunities for interaction, may be necessary to transform initial interest into deeper conceptual understanding. This aligns with prior research showing that, while emotional and behavioral engagement can be enhanced quickly, cognitive gains usually emerge only after long-term exposure and structured scaffolding (Fridin, 2014; Belpaeme et al., 2018). Nevertheless, several students posed follow-up questions to the robot or initiated peer conversations that connected the activity to broader course content. These episodes suggest that even a single interaction with a social robot can ignite moments of cognitive stimulation and perspective-shifting, which could be amplified with more structured integration into the curriculum (Appleton et al., 2006).

Significant differences were also observed between students who had previously interacted with social robots and those who had not, supporting the assumptions of RQ2. Those with prior experience demonstrated faster engagement, more fluid interaction, and greater ease in communicating with the robot. This was confirmed both through qualitative observation and statistical analysis, which revealed higher engagement scores across emotional, behavioral, and cognitive domains for experienced participants. This outcome resonates with research suggesting that technological familiarity can reduce uncertainty, foster comfort, and increase openness to novel learning environments (Woolf, 2009).

Post-activity reflections revealed an interesting contrast: while most students were highly familiar with generative AI tools, particularly ChatGPT, Copilot, and Gemini, their direct interaction with embodied AI remains rare. Many students stated that this was their first time speaking to a robot, especially one using Slovenian. The discrepancy between text-based AI usage and embodied AI interaction underscores the novelty of the experience and highlights the distinct affordances of physical presence, including non-verbal communication and affective responsiveness, which purely textual interfaces cannot replicate. This contrast resonates with broader research on the impact of digitalisation on organizational and societal practices, showing how technological adoption patterns can differ substantially across contexts (Jagodič et al., 2025).

Several students proposed that AlphaMini might be even more impactful in elementary education or among children with special educational needs. These observations are consistent with previous findings indicating that social robots can function as emotionally supportive companions, especially in contexts where relational trust and emotional engagement are foundational (Fridin, 2014; Belpaeme et al., 2018). Our results reinforce this perspective by showing that students themselves highlighted the robot’s calming and inclusive qualities. At the same time, while prior studies have mostly focused on younger children or language learning contexts, our findings extend this evidence to higher education and abstract subject matter, suggesting that the emotional dimension of engagement may generalize across educational levels.

Beyond confirming prior findings, our study offers several unique theoretical contributions to the field of educational robotics and human–robot interaction: AlphaMini’s use of Slovenian notably strengthened students’ emotional engagement, highlighting that social presence is influenced not solely by embodiment but also by linguistic inclusivity, especially crucial in small-language contexts with limited AI tools. Prior work confirms that language variety affects warmth and rapport in human–robot interaction, and others emphasize cultural responsiveness as key to engagement (Louie, 2021). Unlike studies employing scripted dialogue, our integration of ChatGPT enabled dynamic, context-aware responses, paired with embodied non-verbal cues. This aligns with calls for combining large language models with embodied agents for richer interaction (Pinto-Bernal et al., 2025) and advances theories of social presence in robots (Prescott & Robillard, 2021). We found that students with prior robot experience scored higher across all engagement dimensions. This echoes findings that prior exposure shapes trust and attitudes toward robots (Sanders et al., 2017). Most previous studies focus on STEM or language learning with children (Kanda et al., 2004), while our findings extend robot-assisted learning to abstract knowledge management in higher education, demonstrating that robots can engage adult learners in conceptual domains, which is a valuable expansion of application contexts. This multi-method exploratory design, blending survey data, descriptive observations, and post-session reflections, was specifically chosen to capture the interplay between embodiment, language, AI, and familiarity that would likely remain unseen in purely quantitative or scripted designs.

6. Limitations and Future Research

While significant differences were found, these results should be interpreted as indicative rather than conclusive, warranting replication with larger and more balanced samples. The subgroup sizes (N = 22 and N = 48) are below the ideal threshold for robust statistical inference. In addition, the study sample was limited to students from education, management, and related social science fields, with no direct representation from STEAM disciplines. This focus was intentional, as these groups are directly relevant to the pedagogical and organizational applications of social robots. However, the absence of participants from science, technology, engineering, and mathematics limits the robustness and external validity of the study. Furthermore, the sample displayed a pronounced gender imbalance that reflects the demographic structure of the educational sciences and management programs from which participants were recruited.

While the observational component enriched our understanding of student engagement, we acknowledge several limitations. First, the observations were conducted in an exploratory and descriptive manner, without a fully structured ethnographic protocol. Data were limited to handwritten field notes, later digitized and thematically organized. Although this approach is consistent with basic field observation practices (Spradley, 1980) and reflexive thematic analysis (Braun & Clarke, 2006, 2021), it lacks the systematic coding rigor typically associated with qualitative research. Second, no formal inter-rater reliability procedures were employed; instead, reliability was supported through team-based discussions and triangulation with survey data (Lincoln & Guba, 1985; O’Brien et al., 2014). These methodological choices limit replicability and the depth of qualitative insights. Future studies should therefore consider more structured qualitative protocols, the use of coding software, and multiple coders to ensure greater validity and reliability.

It should be noted that AlphaMini has several technical limitations, including a simplified programming interface (comparable to Scratch), a relatively narrow range of predefined behaviors, and dependence on external AI models to extend conversational depth. While these constraints reduce the robot’s flexibility, they also reflect the realities of many educational technologies currently available in schools. Importantly, our study did not aim to test advanced programming or technical features but rather examined how the robot’s basic social and linguistic capabilities in Slovenian influence student engagement. This should be taken into consideration when interpreting the findings.

Limitations of the study also include a short duration of exposure to the robot, the relatively small and context-specific sample, and reliance on self-reported engagement measures. Additionally, the novelty effect cannot be ruled out as a contributing factor to the observed increases in emotional and behavioral engagement.

These constraints limit the generalizability of the findings to broader educational contexts or long-term implementations. Future research should therefore aim to include a more balanced gender representation, as previous studies suggest that gender can play an important role in technology acceptance and interaction. In addition, a more cross-disciplinary approach can be investigated, as prior studies suggest that learners’ academic backgrounds may significantly influence their engagement with technological interventions. Special attention should also be given to inclusive and early childhood settings, where embodied AI may offer unique benefits in promoting emotional safety, social presence, and accessibility.

In addition, the present study should be understood as an exploratory case study with a relatively small and context-specific sample. While the findings provide valuable insights into how linguistic localization, generative AI, and embodiment shape engagement, they are not directly generalizable to other educational contexts or larger populations. The limited scope also means that our conclusions should be treated as indicative rather than definitive. Future research should therefore replicate this work with larger, more diverse samples and across different disciplines, in order to test the robustness and transferability of the patterns identified here.

7. Conclusions

This study explored the role of the Slovenian-speaking social robot AlphaMini in shaping student engagement across emotional, behavioral, and cognitive dimensions. The results demonstrate that the robot’s presence had a particularly strong correlation with emotional and behavioral engagement, enhancing students’ motivation, enjoyment, and classroom participation. These effects were especially notable among students with no prior experience interacting with social robots, highlighting the potential influence of novelty in shaping learner response. Although some signs of cognitive engagement were present, such as reflection and perspective-taking, the observed impact was modest. This suggests that deeper cognitive development may require repeated or sustained exposure to robot-assisted activities, particularly when dealing with abstract or complex content. Overall, the results confirm RQ1 and H1: AlphaMini significantly enhanced emotional and behavioral engagement, while cognitive improvements, although present, were less pronounced and are likely to emerge more strongly with continued exposure. Furthermore, differences in engagement between students with and without previous robot experience indicate that familiarity may play a role in facilitating more seamless interaction and openness to embodied AI in education.

In summary, AlphaMini demonstrates meaningful potential for supporting learner engagement in higher education. However, unlocking its full pedagogical value will require a more comprehensive and iterative integration into the instructional design and further refinement of its technological affordances.