Université Euro Méditerranéenne de Fès

Keywords:

Personnalized Learning, Leaning Analytics, Conversational Agents, AI Powered Gamification, Reinforcement Learning, Deep Learning, LLMs, Belief Networks

Ph.D Candidate Profile:

BAC+5 in computer science related fields (Artificcial Intelligence, Data Science, Software Engineering) with strong level in mathematics, programming and english.

Background:

The global imperative to acquire English proficiency has surged, yet traditional digital language learning platforms often suffer from critical limitations: a one-size-fits-all pedagogical approach, a lack of genuine interactivity, insufficient focus on productive skills (speaking and writing), and an absence of deep, explanatory learner analytics. These shortcomings fail to emulate the adaptive, responsive, and supportive nature of a human tutor. Recent advancements in artificial intelligence (AI), particularly in probabilistic reasoning, deep learning, and natural language processing (NLP), present an unprecedented opportunity to overcome these barriers and engineer a new paradigm in language education.

Objectives:

This research project aims to design, develop, and empirically validate an Adaptive Learning English Assistant, a comprehensive AI-driven pedagogical agent. The project core objective is to function as a personalized, omnichannel assistant for English language learners. The project's innovation lies in the synergistic integration of a multi-layered AI architecture.

Methodology:

1. Pedagogical Foundation: Research established best practices in second language acquisition and curriculum design to define learning objectives and effective teaching strategies.

2. AI Technology Survey: Identify and evaluate relevant AI techniques such as Belief Networks for knowledge tracing, Reinforcement Learning for personalization, and LLMs for content generation and interaction.

3. Gap Analysis: Analyze the limitations of current language learning tools and academic research to define the specific innovation gap such as the lack of deeply integrated adaptive AI tutors that this project will address.

4. Research Desk: Develop a functional prototype integrating the chosen AI components. Validate its effectiveness through controlled studies and disseminate findings through academic publications and conferences.

5. Product Development: Engineer the research prototype into a robust, scalable, and user-friendly software product ready for market via web and mobile app. 6. Iterative Testing: Conduct beta testing with users, collect performance data and feedback, and use it to continuously retrain AI models and enhance the product's features and usability.

7. Real-World Pilot: Implement the finalized product in authentic educational settings through individual users and institutional partnerships. Conduct a longitudinal study to measure learning outcomes, engagement, and overall impact.

Outcomes:

The project is structured as a core PhD research program, providing the foundational work for the candidate's dissertation. The research will yield a fully functional prototype of the platform and will be disseminated to the scientific community through four key publications:

1. Publication 1: A theoretical paper in an AI in Education journal on the novel framework for integrating Belief Networks and Reinforcement Learning for student modeling and pedagogical policy learning.

2. Publication 2: An empirical study on the use of fine-tuned LLMs for generating level-appropriate, engaging learning materials and diagnostic feedback.

3. Publication 3: A technical paper on the architecture and implementation of the multimodal analytics engine, focusing on deep learning models for speech assessment and error detection.

4. Publication 4: A comprehensive summative study in a leading educational technology journal presenting the results of a longitudinal controlled trial, evaluating the agent efficacy on learning outcomes, engagement, and motivation compared to control groups.

Dissemination:

Knowledge transfer will be achieved through a multi-channel strategy targeting both academic and public audiences by active participation in the leading international conferences to present findings, receive feedback, and build collaborations. The scheduled participations are:

2026: To present the initial theoretical framework for the tutor conversational agent.

2027: To present findings on the use of fine-tuned LLMs and diagnostic feedback. 2028: To present the initial theoretical framework of learning gap detection.

Impact:

A primary outcome of this project is the foundational technology and validated proof-of-concept necessary to launch a deep-tech startup. The research is explicitly designed to eliminate the risk of core technological challenges and provide robust, empirical evidence of efficacy, which will form the basis for seeking seed funding and venture capital. The startup will be tasked with refining the prototype into a market-ready product, scaling its infrastructure, and developing a sustainable business model to bring this cutting-edge pedagogical technology to a global audience.