Title: Inter-Model Explanations for Student Modeling

Start Date: Preferably between February and April 2025

This research internship focuses on the fields of eXplainable Artificial Intelligence (XAI) and Artificial Intelligence in Education (AIED). The goal is to design and evaluate XAI methods to explain the predictions made by student models, as well as any disparities between them.

A student model is an AI model that aims to infer a learner's knowledge state based on exercises they have already completed. Such a model can then predict whether the learner will succeed on a new exercise. For example, predicting if a 6th-grade student has the required knowledge to succeed in a fraction addition exercise. There are several well-established student models in the literature, e.g., based on Bayesian models, regression, or neural networks. The internship aims to focus on the design and application of model-agnostic XAI methods to explain the predictions of these student models. Subsequently, the internship will explore how to provide explanations about similarities and disparities between multiple student models simultaneously to achieve a better understanding of these models, as opposed to explaining the predictions of each student model independently.

The expected tasks for the internship include:

Conducting a brief literature review on XAI methods applicable to student models, as well as on inter-model explanations.
Getting familiar with the most popular student models in AIED, as well as open datasets and libraries in Python (e.g., PyKT) to train them.
Applying XAI methods identified in the literature review to each trained student model.
Studying, and if necessary developing, one or more methods for providing inter-model explanations.
The intern is expected to build an open-source repository with his/her code for applying the XAI methods per model and inter-model.

Required Skills:

Strong knowledge of machine learning
Strong programming skills in Python, including with at least one AI library (e.g., Scikit-learn, Keras, PyTorch)
Completion of a course on XAI and/or Bayesian models would be a plus (but not required).
The internship will take place at the LIP6 laboratory at Sorbonne University in Paris, one of the main computer science research laboratories in France. The intern will be supervised by Marie-Jeanne Lesot and Sébastien Lallé, who are respectively part of the LFI (https://www.lip6.fr/recherche/team.php?acronyme=LFI) and MOCAH (https://www.lip6.fr/recherche/team.php?acronyme=MOCAH) team. The intern is expected to work mostly in person on the campus, 4 Place Jussieu 75005 Paris. The start date is flexible, but ideally February to April 2025, for a 5 to 6-months duration, paid about 580€/month.

To apply, please send a CV, your Master’s transcript of records, and a motivation statement to:
· Marie-Jeanne.Lesot@lip6.fr
· sebastien.lalle@lip6.fr