Equidad y objetividad en la evaluación asistida por inteligencia artificial en la educación superior dominicana: un estudio cuasiexperimental en universidades de la República Dominicana

Abstract

The expansion of AI-assisted assessment tools in higher education poses significant challenges to ensuring the fairness and objectivity of the results. This research analyzes, in higher education institutions in the Dominican Republic, how the use of AI in feedback, rubric-based grading, and academic integrity monitoring impacts the validity, reliability, and fairness of assessments. A mixed-methods approach is adopted: (a) a review of institutional policies and ethical-regulatory frameworks (responsible use of AI, privacy, and data protection); (b) a quasi-experimental study comparing grades and feedback times in courses with and without AI; and (c) surveys and interviews with faculty and students regarding perceptions of fairness, transparency, and trust. The results show improvements in the timeliness of feedback and greater inter-rater consistency when AI supports the application of rubrics; however, risks of subgroup bias, false positives in detection systems, and information asymmetries regarding AI use that affect the perception of fairness are identified. Clear institutional policies, disclosure and informed consent protocols, human review in sensitive decisions, algorithmic impact assessments, and teacher training in fair assessment design and bias mitigation are recommended to align the adoption of AI with the educational mission and the rights of the university community.