With the emergence of web applications that enabled user-generated content and social interactions, the Web became a place where people can engage in a number of new activities. With the success of smart enabled devices people now actively share their location data through various applications. However, as this thesis reveals, location plays a primal role in linking and inferring new information about people, often without their knowledge or consent. Due to this inferential power of location data new privacy concerns arise, as the actual affordances of their data are far greater than people are even aware of. Regardless of the numerous controversies around privacy, people keep on sharing their data on the Web. However, privacy systems themselves (and the ways in which individuals express their preferences) have changed very little. This thesis argues that understanding the mechanisms that people employ in their privacy decisions can provide fundamental insight for the design of privacy systems. The main focus of this thesis is to understand the underlying reasons why people share their location and whether their disclosure behaviour is paradoxical when compared with their stated attitudes towards location sharing. The first part of this thesis involves a study comprising of an online survey that addresses these two issues. The findings provide supporting evidence that people's location sharing decisions are indeed paradoxical in comparison with their stated attitudes and that privacy decision-making can be seen as a process of structuration, in the sense that people's decisions are tempered by contextual factors (external structures). The second part comprises of a series of focus groups that act as a follow-up study and aim to explore in more detail the underlying reasons behind people's sharing decisions. The findings show that people's decisions are influenced by a number of different contextual factors, grouped together into three main categories; social capital, trust in the application and functionality. Based on the outcomes of the two studies, a conceptual model was developed, called the Isorropic Model, that points out the prominent role of context in privacy decision-making and stresses the need for more dynamic privacy systems.