Every day, decisions are made within a social context, whether it's selecting a coffee shop or choosing to compete in a chess event. These decisions are shaped by external factors that can influence their success. To make better decisions, it is important to understand how individuals incorporate social signals into their decision making process, and how these interactions impact broader outcomes. Further, decisions occur within a complex system, each with its own unique risks and constraints. This dissertation uses networks as a foundational framework to examine the characteristics of exploration spaces that influence human decision making. This approach allows us to uncover common patterns and gain actionable insights to improve decision outcomes across three domains: innovation, mobility, and competition.
‍

First, I examine innovation in clinical trials, focusing on the social processes that influence drug discovery. By analyzing the Protein-Protein Interaction (PPI) network, I uncover that exploration is often biased toward proteins that have been previously tested or those linked to experimental targets. This biased exploration restricts clinical exposure to a small fraction of the network. To overcome this limitation, I propose network-driven search strategies that could enhance drug discovery and help identify novel drug candidates for rare diseases.
‍

Next, I investigate the mobility network within the metaverse to identify the key patterns that define human movement in virtual spaces. By analyzing the exploration trajectory of individuals across 250 million movements in a large scale virtual world, I find that, despite the absence of commuting costs, individual's inclination to explore new locations decreases over time. I introduce a network-based mobility model that predicts both individual and collective patterns, revealing that individual mobility is more driven by popularity of locations than by distance.
‍

Finally, I examine the impact of gender-segregated venues on individual performance and outcomes in intellectual pursuits. Using historical chess data, I construct the player network that reveals a divide in contests between men and women, primarily driven by individuals deciding to participate in women-only events. Leveraging an incremental propensity score framework, I find that women-only events can foster skill development, particularly for beginners and early career players. I find that the gender performance gap is linked with inaccuracies in game play, which ultimately contribute to negative outcomes, providing a mechanistic explanation for these results.
‍

Each chapter of this dissertation presents novel datasets on human behavior, collectively demonstrating that network-based models can predict individual decisions more accurately. These models also offer policy-driven insights that can reduce bias and enhance successful outcomes.