Search for a command to run...
Objective: The study examined how recent literature has addressed learning analytics to understand and predict student dropout in virtual environments. Academic and behavioral indicators, institutional conditions, and the ethical implications of using artificial intelligence were considered. Methodology: A qualitative study with a bibliographic-documentary approach was conducted. Publications were intentionally selected, and the information was organized using an extraction matrix. Thematic coding and content analysis were applied to identify similarities, differences, and gaps in the field. Results: The research focused on 2021–2025, and the analysis identified four main patterns: (1) reasons and indicators of school dropout; (2) use of data for educational decision-making; (3) contextual conditions (participation, inequalities, and infrastructure); and (4) ethics, biases, and educational objectives of AI. It was noted that the ethical dimension was very prominent, while evidence of predictive models with clear technical validation was less prevalent by comparison. Additionally, risks of misinterpretation were noted when the platform’s metrics do not include access and infrastructure variables. Conclusions: It was concluded that dropout prediction requires integrating participation and performance indicators with pedagogical interpretation, contextual controls, and operational ethical frameworks. This is necessary to sustain transparent and equitable institutional decisions. It was suggested to advance long-term validations, evaluate biases by subgroup, and study the effectiveness of interventions triggered by early warnings.