This paper describes a pilot study on the role of expectations in human-computer interaction on a decision-making task. Participants (N=70) were randomly assigned to one of 5 different computer partners or to a human partner. After completing the rankings for the Desert Survival Task, participants engaged in a dialog with their computer or human partners. Results revealed that interaction with human partners was more expected and more positively evaluated than interaction with computer agents. In addition, the addition of human-like qualities to computer interfaces did not increase expectedness or evaluations as predicted. Correlation analysis for the five computer conditions demonstrated that expectations and evaluations do effect influence and perceptions of the partner. Discussion focuses on ways to coordinate expectations, interface design, and task objectives.