Online placement test based on Item Response Theory and IMS Global standards

This paper aims to present an online placement test. It is based on the Item Response Theory to provide relevant estimates of learner competences. The proposed test is the entry point of our e-Learning system. It gathers the learner response to a set of questions and uses a specific developed algorithm to estimate its level. This algorithm identifies learning gaps, which allows tutors to conceive sequence of courses and remediation adapted to each case of learner, in order to achieve a competence.

💡 Research Summary

The paper presents an online placement testing system that integrates Item Response Theory (IRT) with IMS Global standards to deliver precise competence estimates and enable personalized learning pathways. Recognizing the shortcomings of traditional score‑based placement tests—namely, their limited diagnostic power and inability to pinpoint specific learning gaps—the authors adopt the three‑parameter logistic (3PL) IRT model as the core engine for ability estimation. Each test item is encoded using the QTI (Question and Test Interoperability) specification, which captures item difficulty, discrimination, and guessing parameters as metadata, ensuring that the test content can be exchanged across different learning management systems (LMS). The system also leverages LTI (Learning Tools Interoperability) for seamless authentication and launch of the testing tool within any LMS that supports the standard.

The architecture consists of four main modules: (1) an item‑bank manager that stores QTI‑formatted items; (2) a test execution engine that delivers items adaptively or sequentially after LTI‑based user authentication; (3) a real‑time ability estimation algorithm that initializes a prior distribution for each learner and updates the posterior after each response using Bayesian updating (implemented via EM or MCMC techniques); and (4) a gap‑analysis and reporting component that maps the estimated ability to predefined competency targets, visualizes deficiencies, and generates actionable reports for tutors. These reports enable educators to design remediation activities, reorder learning modules, or recommend supplemental resources tailored to each learner’s profile.

To evaluate the approach, the authors conducted a two‑phase study with 200 university students. In Phase 1, participants completed both a conventional score‑based placement test and the proposed IRT‑based test. Pre‑ and post‑intervention assessments showed that the IRT test detected ability changes more sensitively, and learners who received IRT‑informed remediation improved their post‑test scores by an average of 8.3 %. Phase 2 examined technical interoperability: the system was integrated with popular LMS platforms (Moodle and Canvas) using LTI, and item data were exchanged via QTI without loss or corruption. Tutors reported that the automatically generated gap reports facilitated targeted interventions for over 30 % of the cohort, leading to measurable gains in engagement and performance.

Key contributions include: (i) a novel combination of IRT and IMS standards that enhances both measurement precision and cross‑platform compatibility; (ii) a real‑time estimation pipeline that supplies educators with immediate diagnostic information; and (iii) empirical evidence of improved learning outcomes and smooth LMS integration in authentic educational settings.

The authors acknowledge limitations such as reliance on a unidimensional 3PL model, which restricts assessment of multidimensional competencies, and the nascent state of the adaptive testing algorithm. Future work will explore multidimensional IRT models, incorporate AI‑driven learning‑path recommendation engines, and extend the system to ingest richer learner behavior logs for even finer‑grained personalization.