Networks are of most importance for a broad range of social and economic interactions, such as the spread of information and ideas and the organization of logistic value chains. This thesis aims to investigate the dynamics of networks formation. Specifically, we want to determine which networks are likely to emerge and persist in the long run and in particular when introducing Farsightedness (the ability for (some) agents to anticipate the (re)actions of the others). This thesis contributes to the literature on network formation through three papers. The first one presents a stability concept that intends to select the networks that are likely to emerge and persist when some agents are myopic and some others are farsighted. The paper shows that the concept always exists (i.e., it always makes a prediction). Yet, the concept may, in some cases, get quite inclusive. That is the reason why the second paper presents new stability concepts when all agents are farsighted but with the aim of being much narrower. The concepts there are based on a minimal requirement called minimally unstable networks. In the first and second papers, we assume that the farsighted agents have full knowledge about all feasible networks. The third paper relaxes this assumption and presents stability concepts when all agents are farsighted but have limited information.