Could a robot ever establish a satisfying relationship with a human being? What are the essential engine blocks that would make a robot real and socially present, and what kinds of robot personalities would humans welcome?

The paper starts with a review of previous studies on long-term social interactions between robots and human beings. The authors then present a framework based on the belief-desire-intention architecture and hierarchical task networks, in which virtual characters with different personalities can retrieve past events from episodic memory storage and incorporate them into current interactions with human beings. The framework is then implemented as a virtual expressive tutor designed for long-term human interaction.

The authors conducted tests involving the virtual tutor and groups of human learners, and the results were analyzed statistically. The conclusion is that memory is an important component in building long-term relationships with robots. A robot with memory and a supportive personality has a better long-term social presence than one without, in terms of maintaining user interest at the same level during the test periods.

This paper is a good introduction to human efforts to at least partially achieve, if it is ever possible, singularity--that is, the emergence of a superintelligence that is self-evolving and way above human intelligence.