Formalization of belief change and stories in AI

Formalization of belief change and stories in AI

by Florence Dupin de Saint-Cyr

Formal & Physical Sciences Mathematics PBCMathematical foundationsPBCDMathematical logic
🎙 Florence Dupin de Saint-Cyr 👥 2K 📅 July 9, 2026 ⏱ 83 min 👁 54 📄 expert opinion 🧭 2026-08-15
Available in: English (current) Français

Keywords

belief changebelief revisionbelief updatenonmonotonic reasoningnarrative tension

Summary

Florence Dupin de Saint-Cyr, a researcher from Toulouse, presents a talk on the formalization of belief change and its application to story understanding in AI. She begins by motivating the formalization of reasoning, distinguishing between data-driven AI (system 1) and knowledge representation and reasoning (system 2). She introduces the concepts of belief revision and update, contrasting them with examples. She explains the representation theorems that characterize these operators using postulates and plausibility rankings. She then discusses nonmonotonic reasoning and its connection to belief revision. The second part of the talk focuses on stories, narrative tension, and the concept of ‘candid engagement’ where listeners become immersed in a story. She argues that listening to a story can be modeled as a belief revision process, and she illustrates this with a joke. The talk concludes by suggesting that formalizing belief change can help explain the persuasive power of stories and potentially enhance critical thinking.

152 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talk provides a clear and rigorous introduction to belief change theories, including belief revision and update, with illustrative examples. The argumentation is solid, building on well-established formal frameworks (AGM postulates, Katsuno-Mendelzon update) and connecting them to nonmonotonic reasoning. The link to story understanding is innovative and well-motivated, though the formalization of narrative tension is only sketched. The speaker effectively argues for the importance of formalizing reasoning for AI and for understanding human cognition.

Scientific Rigor, Source Quality, Title Accuracy

The presentation is scientifically rigorous, referencing foundational work in belief revision (Alchourrón, Gärdenfors, Makinson), update (Katsuno, Mendelzon), and nonmonotonic reasoning (Kraus, Lehmann, Magidor). The title accurately reflects the content, as the talk formalizes belief change and applies it to stories. The speaker does not cite specific sources during the talk, but the theoretical foundations are well-known. The talk is an expert opinion rather than a peer-reviewed study, but it is based on established research.

163 words

Title / Content Match

The title accurately reflects the content: the talk formalizes belief change and connects it to story understanding, with a focus on narrative tension and jokes.

Quality & Reliability

8/10

Presentation by a recognized researcher in knowledge representation and reasoning, based on established formal theories (belief revision, update, nonmonotonic reasoning) and illustrated with examples. The talk is technical and rigorous, but it is an expert opinion rather than a peer-reviewed publication.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

Concurring Sources

Contribution & Novelties

The talk bridges belief change theory and narrative understanding, proposing that story comprehension can be modeled as belief revision. It highlights the role of narrative tension and candid engagement in persuasion. The speaker suggests that formalizing these mechanisms could help in designing AI systems that explain persuasion and enhance critical thinking.

Pour aller plus loin :

  • Belief revision — Overview of AGM theory and related approaches.
  • Non-monotonic logic — Introduction to nonmonotonic reasoning and its properties.
  • Narrative transportation — Psychological theory related to immersion in stories.
  • Computational narratology — Research area on computational analysis of narratives.

96 words

Radar Profile

The radar profile shows high scores in quality of information, technical level, and global reliability, with slightly lower scores in quantity of information and novelty. This indicates a technically rigorous presentation with a focused scope, but with limited breadth and originality in the context of existing research.

Reliability 8/10

💬 No comments were provided for analysis.