Inherent Trade-offs in Non-Classical Reasoning Systems

You are reading version 1 of this direction, frozen 2026-10-04. The latest version is version 2.

Version 1 of 25 papers18 conceptsfrozen 2026-10-04

In classical propositional logic, reasoning is straightforward: if a statement is true, it remains true regardless of what new information you learn. However, real-world knowledge is rarely that clean. We deal with exceptions ("birds fly, but penguins don't"), modalities ("it is possible that it will rain"), and contradictions.

When researchers attempt to formalize these "common-sense" reasoning patterns, they encounter a recurring mathematical wall. It turns out that many properties we find intuitive—such as being able to handle exceptions while still behaving like classical logic—are mathematically incompatible. This has led to a series of "impossibility results" across different branches of logic, forcing a shift from searching for a "perfect" logic to making intentional design trade-offs.

The Conflict Between Rationality and Classical Logic

The most prominent example of this trade-off appears in the study of typicality. We want systems that can reason about "typical" members of a category (e.g., typical birds fly) while remaining grounded in classical logic for non-typical cases.

Richard Booth 0001, Giovanni Casini, Thomas Andreas Meyer, and Ivan Varzinczak demonstrated in On the Entailment Problem for a Logic of Typicality that you cannot have it both ways. There is a fundamental incompatibility between:

  1. Conditional Rationality: A set of logical postulates that ensure "if-then" reasoning behaves predictably.
  2. Strict Classical Entailment: Ensuring that standard facts are never contradicted by the reasoning system.

If you prioritize classical logic, your system will eventually fail certain tests of rational conditional reasoning. If you prioritize rationality, you may lose the ability to preserve classical certainties. This isn't a flaw in the algorithms; it is a mathematical boundary of the logic itself.

The Complexity of Modal Logic

When we move from simple true/false statements to modal logic (which handles concepts like "necessity" and "possibility"), even basic tasks like simplifying formulas become difficult. In propositional logic, we use "prime implicates"—the most essential, non-redundant consequences of a formula—to manage knowledge bases.

Meghyn Bienvenu showed in Prime Implicates and Prime Implicants in Modal Logic that extending this to modal logic $K$ triggers another impossibility. You cannot simultaneously satisfy:

  • Logical Equivalence: The simplified version means exactly the same thing as the original.
  • Finiteness: The set of essential consequences is not infinitely long.
  • Structural Simplicity: The results follow a clean, predictable format.

To use modal logic in automated reasoning, practitioners must choose which of these three to sacrifice.

Why We Can't Perfectly Localize Inconsistency

When a knowledge base contains a contradiction (e.g., "The light is on" and "The light is off"), we need to find the "conflict set" to fix it. Ideally, this set should be small, contain no useless information (tautologies), and be robust to different ways of phrasing the same facts.

However, Glauber De Bona and Anthony Hunter proved in Germane Conflicts: Desirable Properties for Localising Inconsistency that no definition of a "conflict" can satisfy all these requirements at once. Specifically, to ensure that the system is "substitution-robust" (meaning it doesn't break if you swap out equivalent terms), you are often forced to include irrelevant tautologies in your conflict sets. This "iceberg" effect means that the root of a contradiction is often mathematically inseparable from seemingly unrelated, trivial truths.

The Failure of the Or Rule in Hierarchies

In standard reasoning, we expect the Or Rule to hold: if a property is typically true for "Cats" and typically true for "Dogs," it should be typically true for "Cats or Dogs."

But as Krishna Manoorkar, Ni Wayan Switrayni, and Ruoding Wang explore in Defeasible Reasoning on Concepts, this fails in concept lattices. In a hierarchical structure, an object can belong to the "join" (the union) of two categories without belonging to either category individually. This structural quirk breaks the inheritance of typical properties.

This is echoed by Richard Booth 0001 and Ivan Varzinczak in Conditional Inference under Disjunctive Rationality, where they show that no inference method can simultaneously satisfy properties like Disjunctive Rationality and Vacuity.

Conclusion: Logic as Design

The shared insight across these papers is that there is no "universal" logic for complex reasoning. Whether dealing with typicality, modalities, or inconsistencies, the transition from classical logic to more expressive frameworks introduces fundamental conflicts between desirable properties.

The task for the researcher is no longer to find the "correct" logic, but to identify which logical desiderata—be it finiteness, classicality, or rationality—are most critical for a specific application and which can be safely sacrificed.

Papers in this version

History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 7 papers
  2. Version 1 · 2026-10-04 · 5 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_logical_property_trade_offs_acfddbf9,
  title = {logical property trade-offs},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/logical-property-trade-offs/rdr_acfddbf9d0bb44b8b136ad0cdff53801},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_inherent_trade_offs_in_non_classical_rea_acfddbf9_v1,
  title = {Inherent Trade-offs in Non-Classical Reasoning Systems},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/inherent-trade-offs-in-non-classical-reasoning-systems/rdr_acfddbf9d0bb44b8b136ad0cdff53801/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}