Thesis: This thesis combines experiments and theoretical models with data analysis to answer ques-tions on the role of social network and aggregated information. Chapter 1 studies a multi-state binary choice experiment in which in each state, one alternative has well understood consequences whereas the other alternative has unknown consequences. Subjects repeatedly receive feedback from past choices about the consequences of unfamiliar alternatives but this feedback is aggregated over states. Varying the payoffs attached to the various al-ternatives in various states allows us to test whether unfamiliar ones are discounted and whether subjects’ use of feedback is better explained by similarity-based reinforcement learning models (in the spirit of the valuation equilibrium, Jehiel and Samet 2007) or by some variant of Bayesian learning model. Our experimental data suggest that there is no discount attached to the unfamiliar alter-natives and that similarity-based reinforcement learning models have a better explanatory power than their Bayesian counterparts. Chapter 2 studies a natural follow up to the question to understand how these findings would change in the face of the feedback being individual specific in the spirit of learning by doing. The follow up experiment allows subject to experiment and generate individual level feedback unlike the endogenous group feedback in the original one. Our experimental data suggest that there is not much difference in learning and the choice in the proportion of Bayesian learners. Chapter 3 studies the demand for monitoring and its effectiveness across different group compositions. In developing countries where formal institutions are often weak or non-existent, the community is responsible to enforce local agreements. Peer monitoring represents a natural mechanism for the enforcement of social norms and local agreements in such a setting. In this paper we collect original network data in 19 villages in rural Nepal and conduct an experiment to study who is elected as a monitor in a public good game.Individuals play in groups of three, both with their close friends and with people socially distant in the network. They receive the opportunity to anonymously choose their preferred"institution". We combine a theoretical model and a unique lab-in-the-field experiment to show that closely knit groups are significantly more likely to not choose any monitor, while sparse groups tend to prefer a monitor who is highly central in their network. Low central monitors are seldom chosen. Endogenous election of the high central monitor ensures higher cooperation compared to an exogenous assignment, but only in sparse groups.