In late 1960, Joseph Weizenbaum authored a program called Eliza that simulated a therapist in carrying out a conversation with the user. The program did not understand anything, but relied on keyword matches, and a few simple heuristics to keep the flow of conversation. However, Weizenbaum was taken aback by the intensity of emotional attachment users felt towards this program, prompting him to highlight this negative aspect of technology in his thought provoking book “Computer Power and Human Reason”. In recent years, however, there has been a revival of Eliza-like systems and interfaces in cognitive robotics. We will look at some such systems and argue that they often have a positive and therapeutic effect on the user, and that in some situations at least this kind of robot-human interaction transcends human-human interaction. We will also examine some ethical implications of this approach.