Thomas BASSANETTI – Soutenance de thèse

Soutenance en français

Equipe : Comportements collectifs (CAB), CRCA-CBI

Encadrement : Guy THERAULAZ (CRCA-CBI) & Clément SIRE (LPT)

Jury :

• Laeticia Gauvin, Directrice de Recherches à l’IRD, Aubervilliers, France (Rapporteure)
• David Chavalarias, Directeur de Recherches au CNRS, Institut des Systèmes Complexes de Paris Île-de-France, France (Rapporteur)
• Alain Barrat, Directeur de recherches au CNRS, Centre de Physique Théorique, Marseille, France (Examinateur)
• Umberto Grandi, Professeur, Université Toulouse Capitole, Toulouse, France (Examinateur)
• Clément Sire, CNRS, Directeur de Recherches au CNRS, Université Toulouse Paul Sabatier, France (Co-directeur de thèse)
• Guy Theraulaz, CNRS, Directeur de Recherches au CNRS, Université Toulouse Paul Sabatier, France (Co-directeur de thèse)

Résumé:

Stigmergy is a generic coordination mechanism widely used by animal societies, in which traces left by individuals in the environment guide and stimulate the subsequent actions of the same or different individuals. In the human context, with the digitization of society, new forms of stigmergic processes have emerged through the development of online services that extensively exploit the digital traces left by their users, in particular, using rating-based recommendation systems. Therefore, understanding the impact of these digital traces on both individual and collective decision-making is essential.

This study pursues two main objectives. First, I investigate and develop a model of the interactions of groups of individuals with their digital traces, and determine how they can exploit these traces to cooperate in an information search task. Subsequently, the research explores the impact of intragroup and intergroup competition on the dynamics of cooperation in the framework of this information search task.

To answer these questions, we have developed the online multiplayer Stigmer game, on which we base 16 series of experiments under varying conditions. In this game, groups of individuals leave and exploit digital traces in an information search task that implements a 5-star rating system. This system is similar to recommendation systems used by many online marketplaces and platforms, where users can evaluate products, services, or sellers. In the game, all individuals interact with a grid of hidden values, searching for cells with the highest values, and using only indirect information provided in the form of colored traces resulting from their collective ratings. This controlled environment allows for a thorough and quantitative analysis of individual and collective behaviors, and offers the possibility of manipulating and studying the combined impact of intragroup and intergroup competition on cooperation.

The experimental and modeling results indicate that the type and intensity of competition determine how individuals interpret and use digital traces, and impact the reliability of the information delivered via these traces. This study reveals that individuals can be classified into three behavioral profiles that differ in their degree of cooperation: collaborators, neutrals, and defectors. When there is no competition, digital traces spontaneously induce cooperation among individuals, highlighting the potential for stigmergic processes to foster collaboration in human groups. Likewise, competition between two groups also promotes cooperative behavior among group members who aim to outperform the members of the other group. However, intragroup competition can prompt deceptive behaviors, as individuals may manipulate their ratings to gain a competitive advantage over the other group members. In this situation, the presence of misinformation reinforces the use of private information over social information in the decision-making process. Finally, situations that combine both intragroup and intergroup competition display varying levels of cooperation between individuals, that we explain.

This research establishes the foundations for understanding stigmergic interactions in digital environments, shedding light on the relationships between competition, cooperation, deception, and decision-making. The insights gained may contribute to the development of sustainable and cooperative personalized decision-making algorithms and artificial collective intelligence systems grounded in stigmergy.