This project focuses on quantum automorphism groups of graphs and their connection to isomorphism games. We will investigate physically observable quantum behaviours, prove self-testing theorems and foster a two-way exchange between quantum groups and quantum information. The latter connection relies on quantum isomorphisms of graphs, which are defined as perfect quantum strategies of isomorphism games. Quantum isomorphisms from a graph to itself have been shown to be equivalent to quantum automorphisms, in the setting of quantum automorphism groups. One of the main questions that the project aims to understand are nonlocal symmetries of graphs. Those are quantum symmetries of graphs coming from non-classical quantum strategies of the corresponding isomorphism game. Interestingly, there are graphs that have quantum symmetry, but all quantum strategies are equivalent to classical ones. Thus, there is a difference between the considered model of reality and our observations of reality. Understanding this phenomenon will enable us to provide new examples of pairs of quantum isomorphic, non-isomorphic graphs. Another objective is to obtain self-testing theorems for isomorphism games. In the language of nonlocal games, self-testing means that any near perfect strategy is close to some fixed reference strategy. Self-testing theorems of linear binary constraint systems will be studied to transfer them to the isomorphism game setting. The project aims also to use recent results in quantum information theory for addressing open questions in quantum groups. On the one hand, giving examples of quantum automorphism groups of graphs that are not residually finite dimensional. On the other hand, figuring out the complexity of computing quantum symmetry and nonlocal symmetry.