A good teacher knows each student as an individual and encourages them accordingly. An Intelligent Virtual Agent (IVA) designed to provide tailored educational and emotional support also needs to reason and respond according to individual student differences. To collect data to develop models of individual differences in students’ preferences and responses to IVAs based on gender, personality and emotional state, we conducted a study with 376 participants using two different virtual advisors (one empathic, one neutral) to “Reduce Study Stress”. The experiment consisted of a control group who received tips via pdf documents and two experimental groups designed with one within-subjects factor (virtual advisors with empathic or neutral dialogue) and one between-subjects factor (different order of receiving empathic and neutral advisors). We also collected students perception of the advisors’ helpfulness and the students’ study stress levels at three time points. Groups using the IVAs reported significantly lower levels of study stress at the end of the study. Some differences were found in preferences for and responses to IVA behaviour based on participants’ gender, personality and levels of depression, anxiety and stress.