Скачать или смотреть Game Theory and the Dynamics of Complex Agentic Systems

Game Theory and the Dynamics of Complex Agentic Systems
By Maxim Yakimenko
Assisted by Gemini 2.5 pro

Slides on GDrive: https://drive.google.com/drive/folder...

Slides on GitHub - https://github.com/lselector/seminar
(click on pptx file, then on "raw" or download button on the right)

Multi-agent systems (MAS) fail with standard
reinforcement learning because agent outcomes
depend on each other's actions in constantly
changing environments.

Solution: Use the game theory to model
inter-dependencies.

Evolutionary Game Theory (EGT) - shows how
successful strategies spread through
populations via replicator dynamics - what works
gets replicated (survives and grows).

Key Insight: "Survival of the flattest" means
robust strategies often out-compete optimal but
fragile ones in realistic, high-mutation
environments.

Fundamental Limit: Turing's Halting Problem
proves complete prediction of complex multi-agent
behavior is mathematically impossible. We can
simulate probabilities but never achieve
certainty about emergent dynamics.

Game theory provides powerful analytical tools
while acknowledging inherent unpredictability.