MODEL OF AUTOMATED TEST EVALUATION USING GENERATIVE AI TOOLS
SCIENCE International Journal · 27.05.2026 · 5 (2) · 227-232
DOI: 10.35120/sciencej0502227a ↗Kurz gefasst
Fachartikel in der Zeitschrift SCIENCE International Journal, online erschienen am 27.05.2026. Verfasst von Goran Aritonović. Die Arbeit nennt 10 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).
Abstract (Originalsprache)
This paper presents a model for automated test evaluation based on the application of generative AI tools. The proposed approach is structured as a sequential evaluation pipeline that uses predefined prompts to process student responses through clearly defined phases. The evaluation process includes defining the answer key, applying scoring rules, analyzing responses at the question level, generating individual reports, and aggregating results. A key contribution of the paper is the definition of a controlled prompt protocol that decomposes a complex evaluation task into smaller, verifiable steps, improving transparency and reliability. The model also supports human-in-the-loop interaction, allowing instructors to intervene at each stage of the evaluation process and maintain pedagogical control. Unlike traditional approaches where AI is used as a general-purpose assistant, the proposed model defines a structured workflow that ensures consistent and repeatable evaluation. This design reduces the risk of uncontrolled AI behavior and enables precise identification of potential errors within individual phases of the pipeline. The model was applied in a real educational setting, where test evaluation was performed using the proposed pipeline. The results indicate a significant improvement in efficiency, with an approximate 95% reduction in evaluation time compared to manual grading. Additionally, the system demonstrates a high level of consistency in applying grading criteria, reducing subjectivity in assessment. It was observed that system reliability depends on the quality of input data, particularly in cases involving manually entered student identification data. The findings suggest that limitations are not solely related to the AI model itself, but also to data acquisition and preprocessing. The proposed model shows strong potential for application in various educational contexts where efficient and transparent test evaluation is required.
Zitieren
APA
Aritonović, G. (2026). MODEL OF AUTOMATED TEST EVALUATION USING GENERATIVE AI TOOLS. SCIENCE International Journal, 5(2), 227-232. https://doi.org/10.35120/sciencej0502227a
Vancouver
Aritonović G. MODEL OF AUTOMATED TEST EVALUATION USING GENERATIVE AI TOOLS. SCIENCE International Journal. 2026;5(2):227-232. doi:10.35120/sciencej0502227a
BibTeX
@article{Aritonovi2026,
author = {Aritonović, Goran},
title = {MODEL OF AUTOMATED TEST EVALUATION USING GENERATIVE AI TOOLS},
journal = {SCIENCE International Journal},
year = {2026},
volume = {5},
number = {2},
pages = {227--232},
doi = {10.35120/sciencej0502227a}
}
EUVIMED