In this study, we manipulated gain/loss framing context during a simulated negotiation between a human user and a virtual agent. Task instructions placed users either in a loss or gain framed context, such that those in the loss frame had to minimize expenses whereas those in the gain frame had to maximize profits. The virtual agent displayed facial emotions so that we could also test how interpersonal emotions interact with framing. Results suggest that individuals are more motivated to minimize their losses than maximizing their gains. The loss frame caused individuals to demand more during the negotiation, hence to minimize expenses. Neurophysiological results suggest that cardiovascular patterns of challenge (i.e., positive motivations) were present in the loss frame condition, most strongly when the virtual human smiled. We discuss these results in regards to Prospect Theory. This work also has implications for designing and rigorously evaluating human-like virtual agents.