Offre en lien avec l’Action/le Réseau : – — –/– — –
Laboratoire/Entreprise : (UR 4108 / FR CNRS 3638)
Durée : 6 mois
Contact : asma.dhaouadi@insa-rouen.fr
Date limite de publication : 2026-02-27
Contexte :
Facility Management (FM) encompasses all activities aimed at improving the efficiency of the working
environment (building management, energy consumption, employee comfort, safety, etc.). These practices
directly influence a company’s economic, environmental, and societal performance [1, 2]. However, the
relationships between these indicators are complex, often implicit, and rarely described in clear quantitative
terms. In addition, company executives need to identify these relationships and their influence on
the overall work environment.
The emergence of Large Language Models (LLMs) and qualitative reasoning approaches (qualitative
influence diagrams, fuzzy logic, Bayesian networks) paves the way for more detailed modeling of dependencies
between indicators, facilitating analysis and strategic decision-making.
Sujet :
Goals
• Automatic extraction of indicators from Quality of Life and Working Conditions (QLWC) documents
(scientific publications, reports, CSR documents, audits).
• Identification of qualitative relationships of influence between these indicators using LLM (e.g.,
“better air quality improves employee productivity”).
• Construction of an influence graph representing these relationships in the form of Qualitative Influence
Diagrams (QID) [3], using two different approaches :
• Fuzzy Logic [4]
• Bayesian Networks [5]
• Analysis of the graph to detect key indicators (those that strongly influence others) and their
influences linking FM practices to overall performance (economic, environmental, societal).
• Prototype decision-making tool for visualizing these graphs and simulating the impact of a change
in indicators.
Bibliography
1. AFNOR, “NF EN 15221-1 : Facility Management — Part 1 : Terms and Definitions,” French Standard,
Association Française de Normalisation, Dec. 2006. Withdrawn on Jul. 13, 2018.
2. Mouvement des Entreprises de France (MEDEF), “GUIDE RSE – Introduction à la Qualité de Vie
et des Conditions de Travail (QVCT),” Paris, France : MEDEF, 2023.
3. Renooij, S., & van der Gaag, L. C. (1998, May). Decision Making in Qualitative Influence Diagrams.
In FLAIRS (pp. 410-414).
4. Klir, G. J., & Yuan, B. (1995). Fuzzy Sets and Fuzzy Logic : Theory and Applications. Prentice
Hall.
5. Koller, D., & Friedman, N. (2009). Probabilistic Graphical Models : Principles and Techniques.
MIT Press.
Profil du candidat :
Expected Skills
• Good knowledge of Python 3 and interest in LLMs.
• Basics of Semantic Web (ontologies, RDF, OWL, SPARQL).
• Interest in fuzzy set theory and probabilistic reasoning.
• Strong motivation for collaboration and teamwork.
Formation et compétences requises :
Adresse d’emploi :
INSA Rouen Normandie
Equipe MIND – LITIS (UR 4108 / FR CNRS 3638)
Avenue de l’Université, BP 8, 76801 Saint-Étienne-du-Rouvray cedex, France
Document attaché : 202511201327_2026_stage.pdf