Ceteris paribus logic in counterfactual reasoning

Patrick Girard, Marcus Anthony TriplettTARK 2015 · DOI · OpenReview

TL;DR

This paper introduces a formal logic for "ceteris paribus" (all else being equal) counterfactuals, which are crucial for reasoning about what would have happened if something had been different, while holding other relevant factors constant. The authors develop a novel semantic framework and dynamic operators to handle these clauses, offering a more nuanced approach to counterfactual reasoning that can resolve paradoxes like the "Nixon argument."

What Problem Does This Paper Address?

Counterfactual statements, often expressed as "if P had happened, then Q would have happened," are fundamental to reasoning, prediction, and explanation. David Lewis's influential semantics for counterfactuals relies on comparing possible worlds based on their similarity to the actual world. However, this approach can be problematic when dealing with "unlikely" or "impossible" events, as the similarity order might be skewed, leading to counterintuitive results. For instance, in the "Nixon argument," a world where Nixon launches a nuclear missile but a "miracle" prevents the holocaust is considered more similar to the actual world than one where the missile launches and causes a holocaust, making the counterfactual "if Nixon had pushed the button, there would have been a nuclear holocaust" false.

This paper addresses the challenge of making counterfactual reasoning more robust by incorporating "ceteris paribus" clauses. These clauses, common in scientific laws and economic principles, stipulate that other relevant factors remain unchanged. The authors aim to integrate these clauses into conditional logic to reconcile dissonant analyses of counterfactuals and provide a more principled way to handle "irrelevant alternatives" or "defeaters" that can distort similarity judgments.

Methods

The core contribution of this paper is the development of a ceteris paribus logic for counterfactual reasoning. They extend standard conditional logic (based on Lewis's possible worlds semantics) by introducing a new conditional operator, denoted as [φ, Γ]ψ. Here, φ is the antecedent of the counterfactual, ψ is the consequent, and Γ is a finite set of formulas that are required to remain true (or have their truth value fixed) during the evaluation of the counterfactual.

Key methodological innovations include:

  • Syntactic Apparatus: Introducing [φ, Γ]ψ directly into the object language, allowing explicit representation of ceteris paribus conditions.
  • Semantic Interpretation: Defining the semantics of [φ, Γ]ψ over standard conditional models. The key is how the similarity relation < is modified or restricted based on the Γ set. This involves defining equivalence relations (=r) and restricted similarity relations (<r) that ensure worlds considered for the counterfactual agree on the propositions in Γ.
  • Dynamic Perspective: Interpreting the [φ, Γ]ψ modality as a dynamic operator, akin to those in dynamic epistemic logic. This means updating the model (specifically, the similarity relations) based on the Γ set before evaluating the counterfactual. This "updates" the world's structure to reflect the "all else being equal" condition.
  • Alternative Interpretations: Proposing and formalizing three different interpretations of ceteris paribus clauses:
  • CP (Strict): Requiring exact agreement on all formulas in Γ.
  • NC (Naïve Counting): Relaxing the strict requirement to maximal agreement on Γ, prioritizing worlds that preserve more of Γ.
  • MS (Maximal Supersets): Focusing on worlds that preserve propositional variables in a maximal way.
  • Completeness Proofs: Demonstrating the completeness of their proposed logics (ACP) with respect to the developed semantics, showing that the formalisms are sound and capture the intended reasoning.

Datasets

Not discussed in the paper. The paper focuses on formal logic and semantics, using hypothetical scenarios and models rather than empirical datasets.

Key Insights and Ideas

  • Formalizing "All Else Being Equal": The paper provides a rigorous way to formalize the intuitive notion of "ceteris paribus," which is crucial for understanding scientific laws, economic models, and everyday reasoning.
  • Resolving the Nixon Argument: The ceteris paribus logic offers a principled solution to the "Nixon argument" by allowing the explicit stipulation that certain factors (like the absence of a miracle) must be held constant, thereby making the counterfactual "if Nixon had pushed the button, there would have been a nuclear holocaust" true.
  • Dynamic Nature of Ceteris Paribus: Viewing ceteris paribus clauses as dynamic operators highlights how they actively change the epistemic or similarity landscape, rather than just being static constraints.
  • Multiple Interpretations of "All Else Being Equal": The paper shows that the meaning of "ceteris paribus" can vary, leading to different logical systems (CP, NC, MS) with distinct properties and outcomes, reflecting the nuanced ways we might interpret such clauses.
  • Bridging Logic and Practice: The work connects formal logic with practical reasoning in fields like preference logic and decision-making, demonstrating the utility of these formalisms.

Key Results

  • Formalization of Ceteris Paribus Counterfactuals: The paper introduces a new class of counterfactual operators [φ, Γ]ψ and provides their semantics.
  • Resolution of the Nixon Counterfactual: The proposed framework, particularly with the CP interpretation and the clause {m} (no miracle), makes the Nixon counterfactual true, aligning with intuitive judgments.
  • Completeness Theorems: The paper proves completeness for the developed ceteris paribus logics (ACP) with respect to their respective semantics (CP, NC, MS), establishing their formal soundness.
  • Comparison of Interpretations: A table illustrates how different ceteris paribus clauses and interpretations (CP, NC, MS) can lead to different truth values for the Nixon counterfactual, highlighting the impact of formal choices. For example, p → h is true under CP with {m}, but false under NC with {m,s}.

Limitations and Future Work

  • Iterated Ceteris Paribus Actions: The paper notes that iterated applications of ceteris paribus counterfactuals can lead to different results than a sequence of updates followed by a counterfactual. Understanding this interaction is left for future work.
  • Cardinality Restrictions on Γ: The current framework primarily deals with finite sets Γ. Extending the logic to accommodate infinite sets of conditions is a significant technical challenge. The paper discusses conceptual difficulties in comparing infinite sets and suggests exploring approaches like "naïve counting" or minimizing distance with respect to set inclusion.
  • Uniform Selection of Γ: The paper touches upon the challenge of uniformly selecting the set Γ for ceteris paribus clauses, particularly in light of von Wright's approach. Further research is needed to develop more flexible methods for selecting these clauses.
  • Infinite Γ: The translation of formulas to a comparative possibility language becomes exponential in the size of Γ for finite Γ. Extending this to infinite Γ would require infinitary languages, which is not a complete solution.

Summary updated

BibTeX
@misc{girard2015ceteris,
  author = {Patrick Girard and Marcus Anthony Triplett},
  title = {{Ceteris paribus logic in counterfactual reasoning}},
  year = {2015},
  doi = {10.4204/eptcs.215.13},
}
Figure