A Computational Theory of Dispositions

Lotfi A. ZadehCOLING 1984

TL;DR

This paper introduces a computational framework for understanding "dispositions" – propositions that are generally, but not always, true (e.g., "birds can fly"). It proposes representing these as propositions with implicit fuzzy quantifiers, which are then "restored" and processed using test-score semantics and fuzzy syllogisms, offering a novel way to model commonsense reasoning and uncertainty.

What Problem Does This Paper Address?

Human knowledge and reasoning heavily rely on "dispositions," which are statements that hold true most of the time but have exceptions. Traditional logic struggles to formally capture this inherent fuzziness and the exceptions. For instance, "birds can fly" is a disposition, but it doesn't account for penguins or injured birds. This paper aims to bridge this gap by developing a computational theory that can represent and reason with these inherently imprecise, yet crucial, aspects of human cognition and knowledge. The challenge lies in moving beyond strict logical truth conditions to accommodate the probabilistic and approximate nature of everyday statements.

Methods

The core of the paper lies in two main techniques:

  • Restoration of Fuzzy Quantifiers: Dispositions are treated as propositions with suppressed fuzzy quantifiers (like "most," "almost all," "frequently"). The first step is to "restore" these quantifiers, making the implicit explicit. For example, "birds can fly" might be restored to "most birds can fly." This restoration is interpretation-dependent.
  • Test-Score Semantics: The meaning of a restored proposition is represented as a procedure that tests "elastic constraints" within an explanatory database. This procedure assigns a test score reflecting the degree of compatibility of the proposition with the data. This approach moves away from traditional truth-conditional semantics.
  • Fuzzy Syllogisms: The paper extends reasoning capabilities by introducing fuzzy syllogisms. These are inference rules that allow for the derivation of new dispositions from existing ones, operating on fuzzy quantifiers and predicates. This is analogous to classical syllogisms but handles the inherent vagueness.

Datasets

Not discussed in the paper. The paper focuses on a theoretical framework and uses hypothetical examples and explanatory databases (EDF) to illustrate its concepts rather than specific, pre-existing datasets.

Key Insights and Ideas

  • Dispositions as Fuzzy Quantifiers: The fundamental insight is viewing dispositions not as absolute truths but as propositions implicitly governed by fuzzy quantifiers. This reframes how we can computationally model everyday knowledge.
  • Test-Score Semantics for Meaning: Shifting from strict truth values to "test scores" provides a more flexible and realistic way to represent the meaning of fuzzy statements and their compatibility with data.
  • Fuzzy Syllogisms for Reasoning: The development of fuzzy syllogisms offers a powerful mechanism for commonsense reasoning, allowing for the propagation of uncertainty and vagueness through logical structures.
  • Scope of Fuzzy Quantifiers: The paper formalizes the concept of "scope" for fuzzy quantifiers, analogous to its role in classical logic, but defined semantically through n-sets and b-sets.
  • Typicality as a Disposition: The paper demonstrates how concepts like "typicality" can be defined as dispositions, highlighting the broad applicability of the framework to cognitive concepts.

Key Results

The paper doesn't present quantitative results in terms of accuracy or performance metrics on specific tasks. Instead, its key results are theoretical:

  • A formal computational framework for representing and reasoning with dispositions.
  • Demonstration of how fuzzy quantifiers can be manipulated as fuzzy numbers.
  • Development of fuzzy syllogisms as an inference mechanism for dispositions.
  • Illustrations of how the framework can be applied to model concepts like "typicality" and dispositional commands.

Limitations and Future Work

  • Interpretation Dependence: The paper acknowledges that the "restoration" of fuzzy quantifiers is interpretation-dependent, meaning different interpretations of a disposition can lead to different formal representations.
  • Computational Complexity: While not explicitly detailed, the complexity of implementing test-score semantics and fuzzy syllogisms for large-scale databases would be a practical consideration.
  • Future Work: The authors suggest that this framework has implications for default reasoning, non-monotonic logic, and the management of uncertainty in expert systems. Further exploration of various types of fuzzy syllogisms and their applications is implied. The paper also touches upon the definition of "typicality" and suggests it can be further explored within this framework.

Summary updated

BibTeX
@inproceedings{zadeh1984computational,
  author = {Lotfi A. Zadeh},
  title = {{A Computational Theory of Dispositions}},
  booktitle = {International Conference on Computational Linguistics},
  year = {1984},
}
Figure