In the field of Artificial Intelligence many problem-solving strategies depend primarily on distributed collective systems. In a collective system, a large group of agents autonomously takes decisions with no central decision-maker. However, the final collective decision is strongly dependent on the underlying group communication network that may support or inhibit a rapid spread of information.
Many real-world networks (e.g. social networks) exhibit non-trivial clustering and community structures associated with a high number of triangular connections, called triadic patterns. Yet the influence of such connection patterns on decision-making dynamics is still not fully understood. Therefore, to study collective decision-making, it is useful to construct networks from distinct, closed, directed triangular communication patterns. Such networks can be realized by generating Triadic Random Graphs (TRGs). Finally, to model the spread of information, existing models of opinion dynamics can be applied to different instances of TRGs (e.g. hierarchical, random, bi-directional).
- Model opinion dynamics on TRGs
- Analyse the influence of different triangular communication patterns (e.g. feedforward or feedback loops) on information spreading
- Real-world social network topologies may be tested for the occurrence of such triangular patterns and the corresponding opinion dynamics compared to that implemented in TRGs
The work in this master thesis entails:
- Literature study on opinion dynamics and the generation of TRGs
- Writing simple programs for construction of TRGs and simulation of opinion spread
- Evaluating the emergent collective decisions with respect to the different properties of TRGs (e.g. directionality, clustering coefficient, hierarchality, network distance, etc.)
- Deducing the influence of particular triangular communication loops (e.g. feed-forward, feed-back, bidirectional) on the speed and accuracy of collective decision-making
- Collective decision-making on triadic graphsI Rausch, Y Khaluf, P Simoens, Complex networks XI: proceedings of the 11th Conference on Complex Networks
- Alon, U.: Network motifs: theory and experimental approaches. Nat. Rev. Genet. 8(6), 450–461 (2007)
- Chen, L., Huepe, C., Gross, T.: Adaptive network models of collective decision making in swarming systems. Phys. Rev. E 94(2), 022415 (2016)
- Milo, R., Shen-Orr, S., Itzkovitz, S., Kashtan, N., Chklovskii, D., Alon, U.: Network motifs: Simple building blocks of complex networks. Science 298(5594), 824–827 (2002)
- Courses: Programming in Python (INF-22306)
- Required skills/knowledge: Computational modelling, data analytics, interest about collective systems, bio-inspired systems.
Key words: Agent-based modelling, (Agent) Simulation Modelling, Collective systems, decision-making, computational societies
Yara Khaluf (firstname.lastname@example.org)