In this thesis, we developed a novel long-term social interaction framework for interaction with virtual characters and human-expressive social robots. The thesis describes an expressive robotic tutor for long-term interaction that can remember past exchanges with users and that can respond emotionally based on its initial personality. Previous work on long-term interaction with virtual characters and robots frequently reported that after the novelty effect disappeared, users' interest into the interaction decreased with time. Our primary goal in this thesis was to develop a system that can still keep the attention of the users after the first interaction. We suggest that episodic memory is a core concept for realizing this goal. We designed an experiment to measure effect of memory on social presence and task engagement levels. Different from previous work on long-tem interaction we found out that users' interest in our system did not decrease with time.