A new formal tool: Functorial variables representing assertions and presuppositions

Ingolf MaxCOLING 1988

TL;DR

This paper introduces "functorial variables" (FVs) as a novel syntactic category in classical propositional logic. FVs offer a unified way to represent both the asserted content and the presupposed background of a statement within a single logical formula, addressing a long-standing challenge in formalizing natural language semantics.

What Problem Does This Paper Address?

Formalizing natural language utterances often requires distinguishing between what is explicitly stated (assertion) and what is implicitly assumed to be true (presupposition). While classical logic excels at handling assertions, integrating presuppositions has proven difficult. Early attempts, like Russell's, treated assertion and presupposition at the same logical level, which was later deemed unsatisfactory. Later approaches suggested separate logical representations, but struggled to combine them into a single, coherent expression. This paper tackles the challenge of creating a unified logical framework that can syntactically represent both assertion and presupposition simultaneously.

Methods

The paper extends classical propositional logic by introducing a new syntactic category: functorial variables (FVs). These FVs are designed to take classical propositional functors (like AND, OR, NOT) as their values. Specifically, the paper focuses on 2-placed FVs of the form $G^{f,g}_{p,q}$, where $f$ and $g$ are "components" (representing functors) and $p$ and $q$ are propositional variables.

The core idea is that these FVs can represent sets of two functors. Semantically, they "unite" the properties of these functors. The paper defines a new language $G$ with formation rules that allow for the integration of these FVs alongside classical propositional variables and functors. A key aspect is the definition of "connection conditions" for FVs, which ensures that any formula containing FVs can be systematically translated into two equivalent classical (FV-free) formulas. This translation process is crucial for evaluating the validity of formulas with FVs.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Unified Representation: Functorial variables provide a syntactical mechanism to represent assertion and presupposition within a single logical formula, overcoming the limitations of previous approaches that kept them separate or conflated them.
  • Syntactic vs. Semantic Integration: Unlike some prior work, this approach integrates assertion and presupposition syntactically, offering a distinct formal tool.
  • "Uniting" Properties: FVs are presented as entities that semantically "unite" the properties of the functors they represent, allowing for a more nuanced logical treatment.
  • Systematic Translation: The paper establishes a clear method for translating formulas with FVs into equivalent classical logic formulas, ensuring that the new system is semantically grounded and comparable to classical logic.
  • New Negation Types: The framework allows for the interpretation of specific FVs as different types of negation, including "presupposition-rejecting" and "presupposition-preserving" negation, which is a significant semantic insight.

Key Results

The paper demonstrates that by introducing functorial variables, a new formal language $G$ is created that is semantically equivalent to classical propositional logic. This equivalence is established through a process of "closing" formulas with FVs, where each formula $A$ in $G$ corresponds to two classical formulas, $A^{E1}$ and $A^{E2}$. A formula $A$ in $G$ is valid if and only if both $A^{E1}$ and $A^{E2}$ are valid in classical logic.

The paper also shows that formulas involving FVs can be stronger than classical conjunction (e.g., $(p \land q)$) but weaker than classical disjunction (e.g., $(p \lor q)$), depending on the specific FVs used. For instance, the formula $(p \rightarrow q)$ is not valid in the new system, highlighting a difference from classical logic when FVs are involved.

Limitations and Future Work

The paper acknowledges that while the system is semantically equivalent to classical logic, formulas with FVs that are neither tautologies nor contradictions remain "unsaturated" until they are translated into their closed classical forms. This is presented as a specific difference in interpretation.

Future work is outlined in Section 6, suggesting extensions to the language $G$:

  • Dropping index-equality for propositional variables to allow for more direct formalization of sentences with shared presuppositions.
  • Representing multiple presuppositions connected by propositional functors.
  • Admitting functional dependence between assertion and presupposition components.
  • Allowing FVs with more than two components, enabling more complex representations of presuppositions.
  • Combinations of these extensions.

Summary updated

BibTeX
@inproceedings{max1988new,
  author = {Ingolf Max},
  title = {{A new formal tool: Functorial variables representing assertions and presuppositions}},
  booktitle = {International Conference on Computational Linguistics},
  year = {1988},
}
Figure