In this study, we investigate the role of network structure for voluntary cooperation on networks. Numerous public goods experiments have shown that many people contribute more to the public good than pure self-interest can easily explain. Correspondingly, there exists a large body of experimental evidence and behavioral explanations. As a natural extension, we deploy the standard linear public goods game in a network setting, where only connected neighbors on the network can contribute to and benefit from the same collection of public goods. In the proposed experiment, we utilize a series of locally tree-like regular graphs to examine if the network structure will affect learning and other-regarding preferences (social preferences). Building upon the various network structures in the experiment, we would like to employ structural models that characterize the patterns of local interactions and calibrate the empirical population distribution of such characteristics. Lastly, the experimental evidence will support the construction of ”Empirical agent-based models,” which we plan to use for ”out of sample prediction.” Based on our understanding of local interactions, we are now ready to explore the provision of public goods in networks with other typologies — be it of larger size or different characteristics.