Technological innovation and socioemotional education: evaluation of the impact of a GPT on the formation of empathy and recognition Open Access

Empathy and recognition are key competencies for building fairer and more inclusive societies. They foster social cohesion, prevent conflict, and promote coexistence based on respect and human dignity [1]. In the face of persistent inequalities and humanitarian crises, cultivating these skills is essential across all educational and professional fields. Universities play a pivotal role in developing these competencies, not only through technical training but by shaping socially conscious citizens committed to justice and inclusion, especially in regions like Latin America, where inequality and violence remain entrenched [2]. Technological innovation, particularly artificial intelligence, offers new avenues to support socio-emotional education. Narrative-based AI tools can facilitate dialogic learning environments that promote mutual understanding and critical reflection on complex realities [3].

This article explores the following questions: To what extent can interaction with an AI model enhance empathy and recognition? What features must such a tool possess to be effective in educational and social contexts? To address these, the study analyzes the impact of the Cattleya Llanera model on developing these competencies through pedagogical and peacebuilding lens. The originality of this work lies in merging AI and pedagogy to promote socio-emotional skills beyond knowledge transmission. By fostering reflective and meaningful interactions, Cattleya Llanera contributes to current debates on AI’s role in social transformation and inclusive education.

Current global crises, armed conflicts, forced displacement, and rising hate speech—have exacerbated inequality and social fragmentation, eroded cohesion and limiting community development [4]. In Latin America, these dynamics are particularly acute, with deep-rooted economic disparities and structural violence affecting vulnerable groups such as women, Indigenous peoples, migrants, and LGBTQ+ communities [5]. In response, empathy and recognition emerge as vital socio-emotional competencies. Empathy, understood as the capacity to understand and share others’ emotions, fosters solidarity and narrows gaps in understanding that sustain inequality [6]. Recognition, meanwhile, entails the validation of others’ histories, rights, and dignity within social frameworks. In contexts marked by exclusion, both are essential to enabling dialogue, reducing violence, and promoting mutual respect [7].

Beyond interpersonal benefits, these competencies influence inclusive policymaking and conflict resolution. Societies that cultivate empathy and recognition through education and community processes tend to develop more resilient responses to social and environmental crises [8]. In Latin America, where historical violence has shaped collective memory, these skills are foundational for reconciliation and reparative justice. Education and social innovation thus play a central role in advancing a culture of peace through critical awareness and recognition of others’ realities [9].

Universities today play a critical role in driving social transformation, not only through academic training but by addressing real-world challenges tied to equity, sustainability, and peacebuilding [10, 11]. Technological innovation, particularly through AI, has expanded this mission, offering tools to confront exclusion, injustice, and violence in new ways [12].

Recent scholarship has emphasized the ethical and normative dimensions of AI, particularly in dialogical and pedagogical applications. Authors such as Floridi, Mittelstadt, and Wachter highlight the importance of transparency, fairness, and contextual sensitivity in systems designed for human development [13–15]. These principles are especially relevant for models like Cattleya Llanera, which moves beyond affective computing or therapeutic AI to address historical violence, collective memory, and social justice through context-driven dialogue.

By drawing on real testimonies rather than generalized prompts, Cattleya offers a more situated and ethically grounded experience. This positions the model within a growing field of responsible AI aligned with transitional justice and peace education. Comparative initiatives, such as digital storytelling platforms in South Africa or trauma-informed dialogue tools in Rwanda, underscore both the potential and limits of technology in reconciliation work. Cattleya Llanera adds to this conversation by illustrating how AI can promote critical reflection, empathy, and recognition in divided societies.

Ultimately, projects like Cattleya demonstrate how universities can mobilize interdisciplinary research, community alliances, and emerging technologies to fulfill their social mission and contribute meaningfully to processes of healing and civic engagement [16, 17].

Cattleya Llanera is an artificial intelligence model developed to facilitate dialogue on peace, reconciliation, and historical memory in Colombia. It responds to the need for tools that strengthen empathy and recognition in a society emerging from decades of armed conflict [18]. Following the 2016 Peace Agreement, the country has faced the challenge of creating inclusive narratives and spaces for mutual understanding [19]. In this context, Cattleya was conceived as a mediating technology to support these transformative dialogues.

Rooted in an eco-historical-biographical methodology, the model incorporates subjectivity, complexity, and territoriality to recover the silenced voices of women affected by violence. Through life testimonies and intersubjective exchanges, Cattleya Llanera foster’s identity reconstruction and dialogue grounded in empathy and mutual recognition, avoiding revictimization and emphasizing resistance and resilience [20]. Developed as part of a research and innovation project on AI for peace education, the model aligns with digital mediation strategies that preserve memory through context-sensitive, reflective interactions. Instead of offering merely informational responses, it generates narrative exchanges that promote critical awareness and the recognition of diverse perspectives on peace [18].

Technically, Cattleya Llanera is based on advanced natural language processing tailored for ethical, context-aware dialogue. It integrates personalized generative techniques and a curated corpus of testimonies, historical documents, and conflict analyses. A feedback mechanism further refines its responses to ensure coherence, empathy, and alignment with social dialogue needs [18]. Its main innovation lies in generating meaningful conversations that deepen users’ understanding of reconciliation. Designed for use in educational, community, and advocacy contexts, the model contributes to constructing a more inclusive collective memory. As a digital tool rooted in peace education, Cattleya Llanera exemplifies how AI can foster empathy and recognition in societies seeking to overcome violent legacies. Beyond education, its potential extends to transitional justice, reintegration of ex-combatants, and intergenerational dialogue. Its development marks a significant step toward integrating technology in peacebuilding, reinforcing academia’s role in reconciliation and opening new pathways for digital mediation in post-conflict settings.

The study involved 66 participants, predominantly women (69.7%), with a wide age range, over half (54.5%) were older than 35, and the rest were distributed across younger age groups. This intergenerational and gender-diverse composition allowed for a more comprehensive analysis of Cattleya Llanera’s impact on empathy and recognition, particularly among populations historically active in Colombia’s reconciliation efforts. The sample reflects the diversity of users who may engage with AI tools in complex social contexts, enabling insights into how different age and gender groups appropriate technology for socio-emotional learning. While the sample provides valuable exploratory insights, its relatively small size (n = 66) should be considered a limitation, particularly in terms of statistical generalizability. As such, the results are intended to inform initial reflections rather than establish definitive trends.

This research is part of the project “Resignifying the Territory: Women’s Reincorporation and Reintegration into Peaceful and Sustainable Communities” (Project No. 109816, Call 948/2024), funded by Colombia’s Ministry of Science through the Orchids Women in Science Program. The initiative leverages tools like Cattleya Llanera to support the reintegration of women affected by conflict, fostering competencies essential for peacebuilding, innovation, and complex thinking.

The instrument used to assess the impact of Cattleya Llanera on the development of empathy and recognition competencies was based on an assessment rubric designed at Tecnologico de Monterrey [21]. This rubric establishes different competency development levels, allowing for the identification of the degree to which a person demonstrates skills in understanding, valuing, and recognizing others. The proposed levels range from basic, where the person demonstrates a limited understanding of others’ perspectives and struggles to recognize realities different from their own, to advanced, where the person can actively interpret and respond to various situations with sensitivity, respect, and openness to dialogue. Between these extremes are intermediate levels that reflect progress in active listening skills, the generation of empathetic responses, and the willingness to recognize and validate others’ experiences (Table 1).

To facilitate its application and analysis, the rubric was digitized and adapted into a Likert scale survey. Thus, it provided quantifiable data on participants’ perceptions of their development of empathy and recognition after interacting with Cattleya Llanera. The final instrument comprised a series of items assessing different aspects of competency, including the ability to understand complex social realities, sensitivity to human rights, the capacity to analyze conflict situations from a balanced perspective, and the willingness to reflect on one’s role in promoting reconciliation. Each item was designed to capture different nuances of participants’ interactions with the tool, allowing for a comprehensive measurement of its impact on the development of the assessed competency (see Table 2).

The instrument’s reliability analysis was verified by its Cronbach’s alpha of 0.839, indicating high internal consistency in measuring empathy and recognition competency. This value exceeds the recommended threshold of 0.8 in social science studies, suggesting that the questionnaire items are aligned and consistently measure the variable of interest. Furthermore, the Spearman correlation analysis between the items yielded a mean of 0.467, indicating a moderate association between them that is not redundant. These results support the instrument’s reliability and reinforce its validity in assessing the impact of interaction with Cattleya Llanera on the development of the assessed competency. The methodology of the “Resignifying the Territory” project is articulated with Cattleya Llanera’s pedagogical strategy, as both employ qualitative and quantitative techniques that allow not only the assessment of socio-emotional competencies, but also the implementation of educational processes based on life stories, using semi-structured interviews and audiovisual techniques that enrich interaction with the GPT model.

The study adhered to the ethical principles of the Declaration of Helsinki and the guidelines of Tecnológico de Monterrey and Minciencias [22]. Participants gave informed consent after being briefed on the study’s scope and their right to withdraw at any time. Data was anonymized and presented in aggregate to protect confidentiality. Interactions with Cattleya Llanera occurred in a respectful, secure setting, ensuring ethical dialogue and compliance with privacy and data protection standards. No ethical approval was received for this article, as the Institutional Research Ethics Committee (Comité Institucional de Ética en la Investigación) for Tecnológico de Monterrey, School of Humanities and Education, Guadalajara, Jalisco, which evaluated our protocol, granted it the status of exemption due to its classification as low risk.

The study employed multiple analyses to assess Cattleya Llanera’s impact on participants’ empathy and recognition competencies. Descriptive statistics provided initial insights into user perceptions, followed by comparisons by gender, age, and prior AI experience to identify potential differences. Item correlations were examined to explore relationships among competencies, and participants were classified by development level using the study’s rubric. These analyses offered a comprehensive view of the tool’s effectiveness in fostering reflection and social awareness across diverse profiles.

Given the nature of this study as an exploratory evaluation of a novel AI-based educational tool, no inferential statistical analyses (such as t-tests or ANOVA) were conducted. The primary aim was not to establish statistically significant differences between subgroups, but rather to understand general patterns of perception and self-reported competency development. The sample size and its unbalanced distribution across demographic categories further limit the reliability of inferential testing. Additionally, the purpose of this research is to assess perceived socio-emotional impact within a specific pedagogical and peacebuilding framework, where qualitative interpretation and reflective outcomes are prioritized over statistical generalization. For these reasons, we do not plan to conduct additional inferential analyses in subsequent versions of this study, although future related research with different objectives may incorporate such methods if supported by the design and sample characteristics.

In the first analysis, descriptive statistics were calculated for the instrument items, including the mean, standard deviation, minimum and maximum values, and the 25th, 50th (median), and 75th percentiles. The results show that item scores fall within the high range of the Likert scale (4 and 5), indicating a positive assessment of the interaction with Cattleya Llanera. The mean responses ranged from 3.91 to 4.22, indicating a general agreement with the questionnaire’s statements. The standard deviations were relatively low, ranging from 0.47 to 0.57, suggesting that there was little dispersion in the responses and that participants tended to evaluate the experience consistently (Table 3).

The results indicate that participants generally perceived a positive impact from interacting with Cattleya Llanera, with most responses clustered at the upper end of the scale. This suggests the tool effectively fostered meaningful reflection. However, the low variability in responses may reflect a tendency toward favorable evaluations, which future studies should explore for potential perception biases.

Gender and age comparisons revealed no significant differences between men and women, as both groups reported similarly positive experiences. Slightly higher scores among older participants suggest they may have engaged more deeply with the tool, possibly due to greater contextual sensitivity or openness to dialogue (Figures 1 and 2).

The findings confirm that Cattleya Llanera was positively evaluated regardless of gender, suggesting a cross-sectional impact. However, slightly higher ratings among older participants may reflect greater sensitivity to social issues, stronger willingness to engage in dialogue, or different patterns of interaction with AI. While the tool proves effective across all groups, future adaptations could enhance engagement among younger users.

The third analysis, based on a Spearman correlation matrix, revealed moderate to high relationships among several items, confirming the interconnection between dimensions of empathy and recognition. A strong correlation (r = 0.79) was observed between participants’ understanding of the socio-political context and their sensitivity to human rights. Additionally, two correlations (r = 0.66) linked this contextual awareness to both the ability to analyze conflict empathetically and recognition of women’s roles in peacebuilding. Conversely, the lowest correlations were associated with the item on generating constructive solutions, suggesting this aspect may represent a more specific, less integrated dimension of the competency, which could be further strengthened in future iterations of the tool (Table 4).

The findings indicate that participants’ perceptions of understanding the social and political context are strongly related to their sensitivity to human rights, suggesting that the Cattleya Llanera tool facilitates the connection between understanding the environment and social awareness. Furthermore, the high correlation between recognition of women’s role in peace processes and understanding of the context suggests that participants who improved in this area also achieved greater general awareness of reconciliation.

The lower correlation in generating constructive solutions may indicate that this aspect requires further development in the tool. Although participants recognize the importance of empathy and historical memory, they do not necessarily translate these reflections into concrete action strategies. This finding may be helpful for future improvements in the implementation of Cattleya Llanera, with a focus on promoting a more direct link between reflection and action in reconciliation processes.

In the fourth analysis, participants’ responses were compared based on their previous experience with artificial intelligence (AI). The data were grouped into two categories: those who had previously interacted with a GPT or AI tool and those who had not. The results were presented in boxplots to visualize the differences in the distribution of average responses on the Likert scale (Figure 3).

The findings indicate that Cattleya Llanera generated a similar impact on participants with and without prior experience with AI, suggesting that the tool is accessible and effective for diverse user profiles. This is relevant because it indicates that its use is not limited to people with prior knowledge of technology but can be used by anyone interested in engaging in dialogue that produces empathy and recognition. The fact that there were no significant differences also reinforces the idea that interaction with Cattleya Llanera is intuitive and that its design allows for meaningful reflections without the need for prior AI training. However, it would be interesting for future research to investigate whether prior experience affects how participants interpret or utilize the information provided by the model.

Finally, the level of the participants’ development of empathy and recognition competencies was analyzed. The average responses on the Likert scale was calculated, and they were classified according to the levels established in the evaluation rubric. Four levels were defined:

1. Basic: GPA below 2.5.

2. Lower Intermediate: GPA between 2.5 and 3.4.

3. Upper Intermediate: GPA between 3.5 and 4.4.

4. Advanced: GPA above 4.5.

The analysis reveals that the majority of participants were at the Upper Intermediate and Advanced levels, indicating significant skill development after interacting with Cattleya Llanera. The number of participants at the lower levels was small, suggesting that the tool was effective in promoting reflections on empathy and recognition (Figure 4).

Thus, the results indicate that most participants rated their development in empathy and recognition as high, supporting the effectiveness of Cattleya Llanera in strengthening these skills. The concentration of responses at the Upper Intermediate and Advanced levels suggests that users perceived the tool’s positive impact on their understanding of the social context and their ability to analyze conflicts empathically.

The fact that very few participants were at the lower levels indicates that the tool was successful in generating a reflective process among participants in most cases. However, it would be relevant for future studies to analyze whether this distribution holds in other populations and contexts and to explore strategies that strengthen the impact on individuals at intermediate levels.

The findings reaffirm the central role of empathy and recognition in building inclusive and just societies [4, 6]. Participants reported a positive experience with Cattleya Llanera, highlighting its impact on interpersonal awareness and its potential to support collective decision-making and inclusive public policy [8]. The strong correlation between contextual understanding and sensitivity to human rights underscores AI’s value as a peace education tool [1, 7].

Aligned with the principles of university social responsibility, the study demonstrates how technological innovations can extend the social impact of academic work [10, 11]. Cattleya Llanera exemplifies how AI can move beyond technical applications to nurture socio-emotional competencies [9]. Its effectiveness across participant profiles, regardless of prior AI exposure, reinforces its accessibility and scalability in educational and community contexts. Nonetheless, the weaker correlation with the generation of constructive solutions reflects a known challenge in socio-emotional learning: translating reflection into concrete action [23]. Most participants were classified in upper-intermediate and advanced levels of competent development, confirming the model’s effectiveness in promoting meaningful reflection. These results echo existing literature on AI’s capacity to foster empathy through dialogic interaction [3]. They also provide a foundation for future research on the long-term effects of such tools across diverse contexts [18].

The relatively weaker correlation observed between empathy-related competencies and the item on generating constructive solutions highlights a recurring challenge in socio-emotional education: the gap between reflective awareness and behavioral action. While Cattleya Llanera appears to effectively promote critical reflection and contextual sensitivity, this internal process does not necessarily translate into tangible strategies or actions for social change. This finding aligns with broader debates in the literature, which caution that empathy alone is not sufficient to drive engagement unless supported by participatory, action-oriented pedagogies. Future iterations of the model could explore ways to strengthen this linkage, for instance by incorporating scenario-based prompts, collaborative decision-making tasks, or real-world problem-solving simulations. Likewise, the integration of Cattleya into broader peace education or transitional justice frameworks could enhance its potential to move users from understanding toward agency, thus reinforcing the connection between emotional insight and civic engagement.

This study offers empirical support for the potential of narrative-driven AI models to foster empathy, critical reflection, and social awareness. Cattleya Llanera repositions artificial intelligence not only as a knowledge interface but also as a vehicle for cultivating values essential to coexistence and peacebuilding. Its accessible and dialogical design allows for use in diverse contexts with minimal training, reinforcing the role of higher education institutions in promoting social responsibility and civic engagement.

To operationalize its impact, Cattleya Llanera could be embedded in ethics, citizenship, or peace education curricula, as well as in NGO-led reconciliation workshops and local government initiatives focused on historical memory and social dialogue. Implementation could be supported by training materials, facilitation protocols, and offline deployment options to ensure broader reach. Particularly in post-conflict communities, the model supports processes of reintegration and empowerment by facilitating context-sensitive, reflective engagement.

Future research should explore the long-term effects of such tools through longitudinal studies. Mixed-method approaches that combine perception surveys with in-depth interviews could offer deeper insights into user experiences and transformation processes. Additionally, applying user-centered design methodologies may improve the model’s adaptability to varied demographic groups. These directions would enhance both the scholarly robustness and practical scalability of Cattleya Llanera across educational and civic domains.

While the findings point to meaningful reflections facilitated by the use of Cattleya Llanera, several methodological limitations should be acknowledged. First, the absence of a pretest/posttest design or control group restricts the ability to draw causal conclusions about participants’ developmental change. The results reflect perceived growth rather than objectively measured improvement. Second, the sample size, although adequate for exploratory analysis, limits the generalizability of the results and suggests caution when extrapolating the findings to broader populations. Third, the exclusive use of self-reported data introduces potential bias, including social desirability effects or overestimations of personal transformation. These limitations are inherent to the design and context of this pilot implementation, yet acknowledging them transparently contributes to the empirical rigor of the study and provides a foundation for future research to expand on this work through longitudinal, comparative, or mixed-methods approaches.

Future iterations of this research would benefit from triangulation with qualitative data, such as participant narratives or focus groups, to enrich the understanding of how users interpret and internalize their interactions with the model.

This study highlights the potential of the Cattleya Llanera model to foster empathy and recognition through reflective, narrative-based interactions. Participants reported a high degree of perceived socio-emotional development, especially in their awareness of social issues and the value of reconciliation. These findings point to the relevance of dialogical AI tools in educational and peacebuilding contexts. Moving forward, future research may explore the long-term effects of such models and examine their applicability across diverse populations and environments. Cattleya Llanera opens new pathways for integrating artificial intelligence into social transformation efforts rooted in empathy, memory, and inclusive dialogue.

No ethical approval was received for this article, as the Institutional Research Ethics Committee (Comité Institucional de Ética en la Investigación) for Tecnológico de Monterrey, School of Humanities and Education, Guadalajara, Jalisco, which evaluated our protocol, granted it the status of exemption due to its classification as low risk.

The authors acknowledge the financial and technical support of the Writing Lab, Institute for the Future of Education, Tecnologico de Monterrey, Mexico, in the production of this work. The authors acknowledge the Ministerio de Ciencia, Tecnologia e Innovacion de Colombia for funding the project “Resignifying the Territory: Female Reintegration and Reincorporation into Peaceful and Sustainable Communities within the Framework of the Peace Process in Colombia” (Project No. 109816, Call 948/2024) through the Orquideas Program: Women in Science 2024.