Updating the Belief Promotion Operator

Daniel A. Grimaldi, Maria Vanina Martinez, Ricardo Oscar RodríguezIJCAI 2021 · DOI

TL;DR

This paper introduces "local belief promotion," a novel way for agents to update their beliefs when new information arrives. Unlike traditional methods that prioritize new information or discard old beliefs, local promotion allows for a more nuanced integration, preserving relevant older beliefs while incorporating new data. This offers a more flexible and potentially more realistic model of belief change.

What Problem Does This Paper Address?

Agents constantly receive new information about themselves and their environment. How they integrate this new information into their existing beliefs is crucial for decision-making. Traditional belief change operations like revision and update often assume that new information must be prioritized, potentially leading to the abandonment of previously held, yet still relevant, beliefs. This paper addresses the limitations of these approaches by proposing a "local belief promotion" operator. This operator aims to strike a balance: it allows for the integration of new information without necessarily discarding all prior beliefs, especially those that are "close" or relevant to the new evidence. The authors are inspired by existing "promotion" operators but adapt them to a "local" setting, similar to how "update" is a local version of "revision."

Methods

The core of the paper is the introduction and formalization of the "local belief promotion operator." This operator is defined by a set of postulates, which are adaptations of existing postulates for update and promotion operators. The authors provide a representation theorem that characterizes these operators in terms of two underlying "update" operators and a "trigger function." The trigger function plays a key role in determining which parts of the old beliefs are preserved based on their "closeness" to the new information. The paper also explores various properties of this trigger function and its impact on the local promotion operator, drawing connections to existing concepts like credibility-limited revision.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Nuanced Belief Integration: The most significant insight is the development of a belief change mechanism that doesn't force a complete overhaul of existing beliefs when new information arrives. This "local promotion" allows for a more selective preservation of old beliefs based on their relevance to the new data.
  • The Role of the Trigger Function: The "trigger function" is a novel concept that acts as a filter, determining which old beliefs are considered "close enough" to the new information to be preserved. This function is central to the operator's ability to perform local belief promotion.
  • Duality with Credibility: The paper highlights a dual relationship between the "preservable" nature of old beliefs in promotion and the "credible" nature of new information in other theories. This suggests a deeper connection between different approaches to belief change.
  • Local vs. Global: The authors emphasize the "local" nature of their proposed operator, contrasting it with the "global" behavior of revision. This local perspective means that the belief change is influenced by specific interpretations of the information, rather than a single, overarching order.
  • Example-Driven Motivation: The Funes and Mafalda example effectively illustrates the limitations of traditional belief change operators and motivates the need for a more flexible approach like local promotion.

Key Results

The paper presents a formal characterization of local belief promotion operators through a representation theorem. This theorem shows that any such operator can be expressed as a combination of two underlying update operators and a trigger function. The authors also provide a set of postulates that define local belief promotion and demonstrate how it relates to existing concepts like update and promotion. An illustrative example is used to showcase the differences between local promotion and other belief change mechanisms, suggesting that local promotion offers a more balanced and realistic outcome in certain scenarios.

Limitations and Future Work

The authors acknowledge that some proofs were moved to an appendix due to space constraints. They also suggest several avenues for future research, including:

  • Studying the "erasure operator" associated with local promotion.
  • Extending the local promotion operator in the sense of [Konieczny and Pérez, 2008].
  • Investigating the relationship between postulates of credibility-limited local promotion and properties of trigger functions.
  • Analyzing the connection between local promotion and selective revision.
  • Considering iterated versions of local promotion operators.

Summary updated

BibTeX
@inproceedings{grimaldi2021updating,
  author = {Daniel A. Grimaldi and Maria Vanina Martinez and Ricardo Oscar Rodríguez},
  title = {{Updating the Belief Promotion Operator}},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year = {2021},
  doi = {10.24963/ijcai.2021/262},
}
Figure