In this talk, I will introduce my online experiments aimed at examining how people coordinate with each other in the presence of machine intelligence. I will begin by discussing the findings published by Nature in 2017, which involved the injection of bots into experimental networks where people (N = 4000 in total) were playing a coordination game. The study showed that the performance of human groups improved when bots acted with seemingly irrational behaviors. Using the study, I will discuss the benefits of using bots for social experiments. I will then introduce a recent study investigating the social implications of intelligent assistance, such as emergency automatic steering assistance, in a novel cyber-physical experiment. Based on these findings, I will discuss my hypothesis of how AI could potentially impact human prosociality by decoupling social foci.