Loïs Vanhée
Greyc, Caen
Tuesday March 27th, 2018, 14:00
Room 306, IMAG Building, Saint-Martin d'Hères
Abstract
Multi-agent systems seek to promote successful interactions between autonomous artificial intelligences. A majority of approaches for designing such systems are tightly inspired by social organization in human societies (e.g. protocols, norms, organizations). These mechanisms, although fostering collaboration in specific situations, are still very inflexible when compared against human societies. Towards lifting this limitation, I searched for the key aspects that make human societies more flexible than their artificial counterpart, looking for new concepts with a fundamental approach: what are the basic building blocks that humans have but agents miss for fostering flexible collaboration? An intensive study of social sciences pointed to an obvious yet overlooked aspect of human societies: culture. Sharing a culture is critical for understanding each other and thus for collectively adapting to new situations, many multi-agent failures in situation of collaboration can be explained in terms of mismatching cultures. This talk introduces my research in integrating cultures within multi-agent systems, towards fostering collaboration: extracting the relevant concepts from social science theories; building models of culture; validating these models by simulating the influence of culture on emerging phenomena, such as gender inequalities, organizational performance; exploiting these models for improving collaboration flexibility in multi-agent systems.