The advent of the “Social Web” has provided a rich testbed for studying numerous human behavioral phenomena at scale for social scientists. Data incidentally collected by online social systems and platforms as people use them to work, connect with others, express opinions and emotions, etc., are temporally fine-grained, rich in detail (e.g., networks of interaction, language, etc.), and importantly relatively easy to access for the motivated researcher. However, as with other, more traditional forms of observational data, online digital traces are generally not amenable to studying causal effects. In this dissertation, we use quasi-experimental study designs which can extract causal narratives from observational data under the right conditions, in order to study how people respond to challenges brought on by external shocks. Historically, the wide-use of quasiexperimental methods have been hampered by the limited availability of suitable data, a limitation that is alleviated by the affordances of online digital traces. It is challenging to use experiments to study how people respond to external shocks as well as how different intrinsic and environmental factors influence these adaptations for ethical and practical reasons. This dissertation demonstrates how, in this vacuum, quasi-experimental studies using online data can make meaningful contributions to the causal understanding of this domain. It also highlights the limits of this approach, especially in the context of extending inferences to offline contexts, and discusses potential strategies for how these issues may be mitigated with additional work. More specifically, we study individual and collective adaptive responses to unexpected challenges in two different settings, natural disasters and attention shocks in online crowd environments. In the context of natural disasters, we use novel computational methods to characterize the response of the community in the aftermath and demonstrate that community responses follow several prototypical patterns in different disasters and the intrinsic properties of disasters can explain some differences in community responses. In the same setting, we investigate how different forms of an individual’s online social capital can influence their choice to move away from the affected area and consider their differing implications for individuals and communities. Finally, in a markedly different setting, we investigate how online peer-production teams manage an influx of users and contributors after receiving an attention shock with implications for the sustainability and growth of such virtual teams. In sum, we demonstrate the potential for this data and methodology combination to infer fine-grained and nuanced narratives regarding how people adapt to unforeseen challenges and draw attention to their implications individuals and collectives vulnerable to shocks as well as to policy makers interested in building resilience in communities.