Skip to main content

Fateh BOULMAIZ

Wednesday December 4, 2024

Multi-target learning for predictive justice

Abstract:

The thesis addresses the design of engaging interactive decision-support systems, a subclass of decision-support systems aimed at inducing and supporting behavior change toward more sustainable practices. This is achieved by involving users in a constructive rather than prescriptive approach through a decision-making process that avoids coercion or deception. The design lies at the intersection of three fields: knowledge management, to develop the reasoning engine that generates recommendations; persuasive technologies, to design features encouraging behavior modification; and human-computer interaction, to seamlessly integrate recommendations and engaging features.

However, even though these systems are applied in diverse fields such as healthcare, education, and more, designing a generic engaging recommendation system is not feasible, as the application domain can significantly influence the effectiveness of engaging features. Therefore, this research specifically focuses on designing engaging recommendation systems to promote behavior change in energy use within inhabited spaces. These systems are grounded in human behavior, studied for years in disciplines such as computer science, philosophy, psychology, sociology, and even rhetoric. These fields provide theories and models to understand and explain the processes involved in behavior change, highlighting its complexity and variability due to numerous influencing factors.

The current state-of-the-art in designing such systems only partially addresses their design, often considering the three key aspects of reasoning, engagement, and interaction in isolation. Even with this limited perspective, existing approaches to implementing these aspects remain improvable. The challenge, therefore, is to devise integrated and synergistic approaches that account for the complex interconnections between these aspects to create effective engaging interactive decision-support systems.

To address these challenges, we propose a holistic approach comprising:
a) a reasoning engine based on the case-based reasoning paradigm, enhanced to improve retrieval, adaptation processes, and data quality;
b) various engaging features such as user inclusion in the decision-making process, explanations, feedback, and confidence indicators; and
c) an interaction design specifically tailored to support behavior change processes, integrating contributions from the reasoning engine and engaging features while considering the specific application domain of energy management.

Date et lieu

Wednesday December 4 at 14:00
Maison Jean Kuntzmann, campus universitaire
et Zoom

Composition du jury

PATRICK  REIGNIER

Directeur de thèse, PROFESSEUR DES UNIVERSITES, GRENOBLE INP - UGA 

MEHDI ADDA

Rapporteur, FULL PROFESSOR UNIVERSITE DU QUEBEC A RIMOUSKI 

GUILLAUME SANDOU

Rapporteur, PROFESSEUR DES UNIVERSITES, ECOLE CENTRALESUPELEC

AMINI  MASSIH-REZA

Examinateur, PROFESSEUR DES UNIVERSITES, UNIVERSITE GRENOBLE-ALPES

AMELIE CORDIER

Examinateur, Docteur en informatique, LYON-iS-Ai

FREDERIC WURTZ

Examinateur, DIRECTEUR DE RECHERCHE, CNRS DELEGATION ALPES

UGO COMIGNANI

MAITRE DE CONFERENCE, UNIVERSITE GRENOBLE-ALPES

TEPHANE  PLOIX

PROFESSEUR DES UNIVERSITES, GRENOBLE INP - UGA

Submitted on December 2, 2024

Updated on December 2, 2024