Rabbi: Exploring the Inner World through Stories

AAAI 1996

TL;DR

This paper introduces "Rabbi," a conversational storytelling system designed to explore users' inner worlds through indirect advice from Hasidic stories. It demonstrates that a system can foster meaningful interpretation and self-awareness by creating a believable persona and socio-cultural context, rather than relying on complex AI matching.

What Problem Does This Paper Address?

The paper addresses the gap between traditional storytelling, used for indirect advice and wisdom, and modern digital systems that often focus on entertainment or direct information delivery. While some research explores storytelling systems through drama or adventure metaphors, this work investigates the potential of storytelling for counseling and self-awareness. The authors aim to create a system that can facilitate personal reflection and understanding by simulating a wise elder figure, a role traditionally filled by storytellers in oral traditions.

Methods

The core of the "Rabbi" system is its conversational approach, simulating an encounter with a rabbi. Instead of employing complex matching algorithms, "Rabbi" focuses on creating a rich interaction and a believable persona. The system parses user input, identifies keywords, and augments them with synonyms, hyponyms, and hypernyms from WordNet to find relevant Hasidic stories. Each story is indexed by descriptors (nouns, verbs, Commandments) that establish a socio-cultural domain. The system's "persona" is constructed through conversational turns, aiming to build trust and maintain the user's suspension of disbelief, mirroring how human storytellers build rapport.

Datasets

The paper mentions a "data-base of stories" containing Hasidic stories. No specific details about the size or origin of this dataset are provided, beyond its indexing with nouns, verbs, and Commandments.

Key Insights and Ideas

  • Persona over Intelligence: The system's effectiveness hinges on its ability to project a believable personality and establish a socio-cultural context, rather than on sophisticated AI intelligence or complex matching techniques.
  • Indirect Advice as a Tool for Self-Awareness: Simulating traditional storytelling roles, like that of a wise elder, can be a powerful mechanism for users to explore their own experiences and gain self-insight.
  • Socio-Cultural Context is Key: The success of interpretation relies heavily on the shared socio-cultural context between the system and the user, even with a primitive matching technique.
  • Emotional Believability: Constructing emotionally believable characters is crucial for maintaining user engagement and the suspension of disbelief, a principle applied to "Rabbi's" persona.

Key Results

The paper reports that users expressed a need to tell personal stories when interacting with "Rabbi." The subjects found the Hasidic stories to be coherent with their own narratives, suggesting that the system successfully facilitated meaningful interpretation. The authors conclude that the system, by projecting a personality, appears believable and can foster self-awareness without needing to be "intelligent" in a traditional AI sense.

Limitations and Future Work

The paper acknowledges that the matching technique used is "primitive." Future work is focused on exploring what kind of discourse structures the system should parse to better abstract the core message or "point" of a personal story. The authors also imply a need to further refine how the system constructs its persona and manages conversational turns to enhance the user experience.

Summary updated

BibTeX
@inproceedings{1996rabbi,
  title = {{Rabbi: Exploring the Inner World through Stories}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {1996},
}
Figure