Introducing sentinel assessment: An AI-powered approach for evaluating writing competence in French as a Foreign Language

Abstract

The evaluation of writing competence in French as a Foreign Language is a pedagogical challenge, traditionally hampered by high subjectivity and time constraints. The emergence of Artificial Intelligence presents a new paradigm for educational assessment, particularly within educational contexts where traditional methods still struggle with complex competences evaluation. This research addresses the conflict between the limitations of traditional writing assessment and the potential of AI-driven solutions. The central problem is determining how an AI-based system can effectively support the formative eAssessment of FFL writing competence for elementary school students, ensuring pedagogical value beyond simple automation.
Using a Design-Based Research (DBR) and sequential mixed-methods approach, this study developed and implemented “eCorrige,” an AI-powered mobile ecosystem which analyzes students’ written works based on four criteria: relevance, coherence, cohesion, and linguistic proficiency. Its performance was evaluated by comparing its three repeated automated gradings of 62 student compositions against assessments from three expert human teachers. The findings demonstrate significant inter-rater variability and subjectivity among the human teachers. In contrast, the eCorrige AI system exhibited high stability and internal consistency across repeated assessments. The AI’s evaluations aligned within the range of human grading practices, proving to be a reliable and consistent assessment tool.
This research introduces “sentinel assessment,” a novel conceptual framework where AI acts as a predictive tool to identify potential learning difficulties before they become entrenched. This model moves beyond traditional formative assessment by enabling anticipatory remediation and preventive pedagogical intervention.