We conduct textual analysis of a sample of more than 200,000 papers written on HIV/AIDS during the past three decades. Using the Latent Dirichlet Allocation method, we disentangle studies that address behavioral and social aspects from other studies and measure the trends of different topics as related to HIV/AIDS. We show that there is a regional variation in scientists' approach to the problem of HIV/AIDS. Our results show that controlling for the economy, proximity to the HIV/AIDS problem correlates with the extent to which scientists look at the behavioral and social aspects of the disease rather than biomedical.