Beyond teaching: how an AI-based conversational model shapes university students’ structural understanding of the environmental crisis

Purpose This study investigates the influence of artificial intelligence on environmental awareness and perceptions of climate responsibility among university students. A quasi-experimental design was employed, involving two groups: an experimental group that interacted with Alma de Tezontle, an AI-based conversational model and a control group that received only an informative presentation. Participants (n = 92) completed an adapted version of the Andalusian Ecobarometer. The data were analyzed using machine learning classification models, specifically decision trees, to identify patterns in levels of environmental concern. Design/methodology/approach The study aimed to explore how interactions with AI-based narratives influence university students’ understanding of environmental issues, with particular emphasis on their perceptions of responsibility for climate change. It sought to determine whether AI tools promote deeper cognitive engagement with environmental problems and support a shift toward more structural interpretations of ecological crises. Findings Students in the experimental group (n = 60) who interacted with Alma de Tezontle displayed greater variability in their levels of environmental concern (levels 6–10) than the control group (n = 32), which ranged from levels 7 to 10. Decision tree analysis revealed that factors such as perceiving the crisis as inevitable, attributing responsibility to companies and practicing sustainable behaviors were key in classifying concern levels. The decision tree model successfully classified students’ concern levels with 87.5% accuracy, revealing distinct predictive pathways for each group. This pattern was particularly pronounced in the experimental group, where AI interaction was associated with more structurally oriented interpretations of climate responsibility. Research limitations/implications The study’s main limitations include the use of a single institutional sample and the brief duration of interaction with the AI model, limiting generalizability and long-term analysis. Future research should expand exposure time and include diverse educational and cultural contexts, incorporating qualitative methods to better understand changes in perception and engagement. Practical implications The results support the integration of AI-based conversational models into environmental education programs to enhance critical thinking and foster deeper understanding of climate issues. The use of decision trees to analyze perception data can inform the design of more tailored and impactful educational interventions. Social implications By promoting a critical and structured understanding of the environmental crisis, AI tools can contribute to forming more informed and engaged citizens. This approach supports collective responsibility and may strengthen sustainability policies by highlighting the roles of individuals, companies and governments in addressing climate change. Originality/value This study introduces an innovative approach by combining AI-based dialogue with environmental education to foster critical reflection on climate issues. Unlike traditional educational tools, Alma de Tezontle enables personalized, symbolic interaction. The use of decision trees provides a novel methodological contribution by segmenting ecological thinking patterns, offering deeper insights into how AI can mediate environmental understanding.