Introducing Sentinel Assessment: An AI-Powered Functionality for Evaluating Writing Competence in French as a Foreign Language
Saved in:
| Title: | Introducing Sentinel Assessment: An AI-Powered Functionality for Evaluating Writing Competence in French as a Foreign Language |
|---|---|
| Language: | English |
| Authors: | Abdelghani Es-sarghini (ORCID |
| Source: | Language Education & Assessment. 2026 9. |
| Availability: | Castledown Publishers. Ground Level, 470 St Kilda Road, Melbourne, 3004, Australia. Tel: 646-520-0676; e-mail: contact@castledown.com; Web site: https://www.castledown.com/journals/lea/ |
| Peer Reviewed: | Y |
| Page Count: | 22 |
| Publication Date: | 2026 |
| Document Type: | Journal Articles Reports - Research |
| Education Level: | Elementary Education Grade 6 Intermediate Grades Middle Schools |
| Descriptors: | Artificial Intelligence, French, Second Language Learning, Writing Skills, Writing Evaluation, Elementary School Students, Grade 6, Foreign Countries, Technology Uses in Education, Formative Evaluation, Interrater Reliability |
| Geographic Terms: | Morocco |
| ISSN: | 2209-3591 |
| Abstract: | The evaluation of writing competence in French as a Foreign Language remains a complex pedagogical challenge, traditionally constrained by high inter-rater variability and significant time burdens. The emergence of Artificial Intelligence (AI) offers an opportunity to shift the paradigm, yet empirical evidence regarding its reliability remains limited. This research addresses the tension between the limitations of traditional manual assessment and the potential of automated solutions to enhance formative practices. Adopting a Design-Based Research approach, this research developed "eCorrige," a mobile ecosystem powered by AI designed to analyze student-written compositions based on the criteria of relevance, coherence, cohesion, and linguistic proficiency. The performance of eCorrige was rigorously evaluated by comparing the assessment data from two groups of evaluators: three expert human teachers and three repeated iterations of the automated system, applied to a corpus of sixty-two compositions produced by students in the sixth grade, the final year of elementary education in Morocco, totaling 3,740 tokens and 1,118 unique types. Paired-sample t-tests revealed a marked contrast in reliability. While human evaluators exhibited significant heterogeneity and high standard deviations, eCorrige demonstrated superior stability and consistency across repeated assessment trials. Statistical tests confirmed a significant concordance between human and automated scoring on objective criteria such as errors and structure, revealing that the automated system applies stricter linguistic standards in the assessment of qualitative language mastery. These findings suggest that the algorithm models a "conservative" expert rater, offering a standardized complement to human evaluation. Finally, this research proposes the "sentinel functionality," a metaphorical concept where AI serves not just as a grader, but as a predictive tool, which forecasts potential learning trajectories and enables anticipatory remediation. |
| Abstractor: | As Provided |
| Entry Date: | 2026 |
| Accession Number: | EJ1502003 |
| Database: | ERIC |
Be the first to leave a comment!
